In.

This is to certify that the

dissertation entitled

AUTOMATED ANALYSIS OF
TRIPLE QUADRUPOLE MS/MS DATA

presented by

Hugh Ralph Gregg

has been accepted towards fulfillment
of the requirements for

Ph.D. degree in Chemistry

fl»?

Major professor

Date W6

MS U is an Ajfimafl'w Action/Equal Opportunity Institution 0-12771

 

 

MSU

LIBRARIES
.—,_.

 

 

RETURNING MATERIALS:
Place in book drop to
remove this checkout from
your record. FINES will
be charged if 560E is
returned after the date
stamped below.

 

 

 

AUTOMATED ANALYSIS OF TRIPLE QUADRUPOLE MS/MS DATA

By

Hugh Ralph Gregg

A DISSERTATION

Submitted to
Michigan State University
in partial fulfillment of the requirements
for the degree of

DOCTOR OF PHILOSOPHY

Department of Chemistry

1986

ABSTRACT

AUTOMATED ANALYSIS OF TRIPLE QUADRUPOLE MS/MS DATA

By

Hugh Ralph Gregg

A data Dexplosion has accompanied the advent of
microcomputer controlled instrumentation. Many of these
automated instruments are capable of generating more data
than the operator is capable, willing, or able to analyze.
What the analyst wants is not the data, but the chemical
information contained in the data. In this dissertation,
several software tools developed for collecting, storing
and retrieving data as well as methods for extracting the

information contained in mass spectrometry/mass

spectrometry (MS/MS) data are presented.

Triple quadrupole mass spectrometry (TOMS) is a multi-
variate technique with unique data storage and retrieval
requirements. Traditional MS or GC/MS data base systems
are unable to handle the many dimensions of data that make
MS/MS a powerful tool for structure determination and
mixture analysis. A multi-dimensional data base was
developed to quickly and efficiently store all data and

instrumental parameters' produced by TQMS. A program is

presented which is able to extract any arbitrary two-

dimensional plane of data for display and interpretation.

The final tools presented in this dissertation are aids
in the interpretation of MS/MS data. One of the techniques
used in analyzing a mass spectrum of an unknown is to
identify characteristic neutral losses from the ions
present in the spectrum. MS/MS is unique in that if the
collision conditions demonstrate first order fragmentation
of the parent ion, all the daughter ions formed are direct,
single event fragmentation products of the parent. A
program was developed which compares neutral losses (the
difference between parent and daughter masses) with a table
of common losses and presents a list of possible fragments

and/or the substructures giving rise to those fragments.

Mass spectra matching routines have traditionally been
used to identify unknown spectra. The simplicity and wide
variations in intensities in daughter spectra make normal
EI matching techniques unsuitable. The matching algorithm
presented in this dissertation was designed to group

daughter spectra.

Each of the tools presented in this thesis addresses
different aspects of the data/information handling problem
and demonstrates the need for powerful software tools to

complement today’s complex instrumentation.

ACKNOWLEDGEMENTS

I’d like to thank everyone that has helped make my stay
in East Lansing both rewarding and pleasurable, with an
extra thanks to Chris Enke for his guidance and friendship.
Tom Atkinson also deserves special recognition, for as busy
as he was, he always had time to talk about chemistry,
computers or home repair.

Chris Enke’s research group provided an ideal
atmosphere to both learn and make friends - thanks to all
past and present group members. I’d also like to

acknowledge the following people and organizations: the
Physiology Department, Jack Hoffart, Paul Sorrenson (and
Art’s bar), the Milton Hilton (and all former tenants), the
Winded Spartans, Brenda Spiewak, Mike McPherson, the Beak &
Dome Bar & Grill and of course, Milton Webber.

ii

TABLE OF CONTENTS

Chapter 1: Introduction
Introduction
Structure determination by MS/MS
Thesis outline
Conclusions

Chapter 2: Instrument Control

Introduction

Phase 1: Digital strip chart recorder
Digital strip chart recorder: Hardware
Digital strip chart recorder: Software

Phase 2: Single micro control system
Single micro control system: Hardware
Single micro control system: Software

Dr. Memory’s monitor

Dr. Memory’s monitor internal structure
Structured Library Oriented Programming System
SLOPS internal structure

Mass spectrometer control software

Phase 3: Current mass spectrometer control system
Conclusions

Chapter 3: Multi-dimensional Instrument Data Base
Introduction

Scientific versus business data bases
Survey of data base structures
Data base capabilities
An example of multi-dimensional data
A programmer’s view of the data base
Data base file formats
Dictionary file
Header file
Pointer file
Data file
Creation of a data set
Conclusions

iii

mmNo-a

10
10

15
15

19
23

26
30
34
35

36
38
39
42
44
45
46
50
54
59
62
46
66

Chapter 4: Retrieval and Display of Multi-dimensional Data

Introduction
Data retrieval
EXTRACT - the program
How EXTRACT works
EXTRACT internal structure
User interface
Subroutine EXEDIT -
Presentation of EXTRACT results
Subroutine XTRACT
Examples of EXTRACT use
Conclusions

Chapter 5: Extracting the Information Contained
in MS/MS data

Introduction

Information contained in MS/MS data

Neutral spectrum

A simple expert system for neutral loss analysis
ANEUT knowledge base and rules
Examples of ANEUT use

Data groupings
MS versus MS/MS spectra matching
Intensity matching for daughter spectra
Ranking and sorting the data groups
Examples of daughter spectra grouping

Conclusions

Bibliography
Appendix 1: Dr. Memory, SLOPS and control software
Appendix 2: Multi-dimensional data base subroutines

Appendix 3: EXTRACT User’s Guide

iv

69
71
72
73
75
75
78
80
82
83
86

90
93
94
98
98
101
104
105
107
108
110
113

115
122
140
153

NNHQQQMutht-IOIOOON

hhhwwwwwwwwwwuw NH

raw

LIST OF TABLES

TOMS operational modes

Summary of commands for the digital strip chart
software

Dr. Memory’s monitor

Basic SLOPS subroutines and commands

TOMS status display

TOMS control system commands

Capabilities of the multi-dimensional data base

Method to collect 5 dimensions of data

Comparison of sequential and direct access files

Definitions in the TOMS dictionary

Dump of the parameters in the Reader file

Summary of Header file format

Summary of one record in the Pointer file

Dump of one logical record in the Data file

Outline of how EXTRACT functions

User/machine interface types

Numeric representation for EXTRACT

030! :5 wNwthHmmhuNHI-ok wNi-IH

OI 010'. 0| mmmhpppwuwwwcswu NNNH

4

LIST OF FIGURES

Software tools for structure determination

First TOMS control system

A typical single microprocessor system

Relationship between conventional and SLOPS
programming

SLOPS library entry format

Data base structures

Writing to the multi-dimensional data set

Instrument description file format

Use of the dictionary file

Header file format

Pointer file format

Data file format

Menu format for EXTRACT

Example plot of intensity vs. pressure extraction

Example plot of intensity vs. energy extraction

Example plot of three dimensional data extraction

Comparison of daughter and neutral spectra

Entries from the knowledge base for ANEUT

Neutral spectrum and analysis of
1,4-benzenediamine

Neutral spectrum and analysis of
bis-2-ethyl-hexy1 adipate ‘

Characteristics used to match daughter spectra

Example of the daughter spectrum matching
algorithm

Example of the El spectrum matching algorithm

vi

12
16

20
27
41
47
52
53
58
61
63
77

85
87
95
99

102

103
109

111
112

Chapter 1

Introduction

Introduction

In the fall of 1978 a new instrument was just coming
online and being tested in Dr. Enke’s research laboratory.
The triple quadrupole mass spectrometer (TOMS), designed by
Yost and Enke (1-3), was completed and proved capable of
generating daughter spectra (mass spectra of selected
parent ions). The technique of mass spectrometry/mass
spectrometry (MS/MS) was not new (4-6), but this new
instrument was capable of unit mass resolution in both mass
analyzers, and boasted a highly efficient collision
chamber. It was a new instrument - a promising technique

and the basis for many years of fascinating research.

In tandem quadrupole mass spectrometry (7), ions
created in the source are selected by the first mass filter
(quadrupole l) and undergo a fragmenting collision with
neutral molecules in the collision cell (quad 2, RF only:
not mass filtering). The ionic products of this collision
are then mass analyzed (by quad 3) and detected. By this
process, a fragmentation spectrum from each ion in the

normal mass spectrum can be obtained.

2
The TOMS instrument can be operated in several ways by
using the mass filtering quadrupoles in one of three modes:
1) selecting one mass for transmission, 2) scanning a
series of masses, or 3) allowing all masses to pass through
the quadrupole (RF only mode). Table 1.1 shows a summary

of the operating modes and the resulting scans.

The information contained in MS/MS data can be used in
a variety of ways. Daughter spectra are useful in mixture
analysis and screening techniques (2,5,8-11) as well as
structure determination problems (2,5,12-14). I will
present, in this dissertation, tools and techniques
necessary to extract the information present in this type
of multi-dimensional data. To help the reader to gain an
appreciation for how these pieces fit into the overall
goal, I will present the structure determination scheme
developed in our laboratory. This method, and associated
tools, have been developed and refined over the years by

several people (15-16).

Structure determination by MS/MS

The basic premise behind our structure elucidation
scheme is that if many small, "simple" parts or
substructures of a sample are known, its structure can be

determined. Each ion in the source results from some

3
Table 1.1

TOMS operational modes

 

Quad 1 Quad g Quad 3 Description

scan RF only* RF only* Normal mass spectrum
no gas

RF only* RF only* scan Normal mass spectrum
no gas

fixed RF only"I scan Daughter scan: spectrum
gas on of all daughter ions

from a selected parent

scan RF only* fixed Parent scan: spectrum
gas on of all parents that
’fragment to form a

specific daughter ion

scan RF only' seen at a Neutral loss/gain scan:
gas on fixed offset spectrum of those
from Quad 1 parents that lose/gain a
given mass during a

collision
fixed RF on1y* fixed Single or multiple
gas on reaction monitoring

*RF only quadrupoles pass all masses (not mass filtering)

4
substructural feature of the original molecule, and the
fragmentation (daughter) spectra of these ions are often
indicitive of these substructures. Other MS/MS information
(parent scans, neutral loss/gain scans, neutral spectra,
etc.) can also be used to identify structural fragments.
The general approach is to make correlations between a
known substructure and some subset of the complete MS/MS
fragmentation 'map. With a library of these correlations,
an unknown MS/MS fragmentation map can be quickly analyzed
for the indicated structural characteristics. In the first
implementation of this scheme, we will compare an unknown
daughter spectrum against a library of daughter spectra.
The closely matching daughter spectra are assumed to be
related to structural similarities, specifically common
substructures. The structures of the reference compounds
will be compared to determine the substructures they have
in common. Lists of substructure/spectrum correlations
will be made in this way. This process would be
impractical without a group of synergistic software tools
to help automate each step. Figure 1.1 shows in block form
each of the_ major tools needed for a project of this

magnitude.

An unknown sample enters the scheme diagramed in
Figure 1.1 as experimental data which is stored in the

multi-dimensional data base (see Chapter 3). These unknown

______________ ' intelligent :
r i Controller 5
I. ...... 1' _______
I
i

 

Triple Quad
Mass Spectrometer

’ l

 

AL

Reference Speciruv‘klibrary storage I Experimenter's

 

 

 

Data Base Data Base

 

 

 

 

 

 

 

 

 

 

 

t
l
I
3 '.
3 5: i
15. fi'» . l
*5 +5; {2% reference test I
a g : 3 spectra Spectrum spectra :
m o ' l
\xu: ' l
3 54% Matching :
g 75:" r matched lists :
2 is. * v :
a pa- ---------- ‘ ’ '
I | t st structures
: Substructure 5.-.? ---------- (sub)Structure :
. l
1 Searching E'"fib7;,;'gfo';;g;" Data Base :
:- motched identified 3 i i
‘ substructures .1"! r substructures :1
" """""" : Molecular i """""" " :
Structure i< Structure 1 Formula .
-‘ - '
s . .
.---9[9.°.".'.’.°I---.i Generator L---93".°.'f’.*.°f---.i

 

 

t

Identified Structure

Software tools for structure determination

Figure 1.1

6
daughter spectra are compared to spectra in the reference
data base (17-18) by matching algorithms (19-20, also see
Chapter 5). The substructures of the best matched daughter
spectra are obtained (currently these substructures are
manually entered, but algorithms to extract these
substructural features from structures of the parent
molecule are being developed). All resulting substructures
(and .information from other sources) will be used as input
to GENOA (21-23) to determine all chemically possible
structures 'that contain the identified substructures. The
final step involves analyzing GENOA’s output structures for
the major features and determining what experiments will

further reduce the number of candidate structures.
Thesis outline

This thesis describes my work toward automating the
data collection and analysis of a complex, multi-variate
instrument. That this work involves three vastly different
areas of computer science (real-time processing for control
systems, data base management and expert systems for data
analysis) underscores the diversity of techniques needed to
take full advantage of today’s automated instruments. This
dissertation consists of five chapters, including this
introduction. Each chapter is independent of the others

(although Chapter 4 refers back to Chapter 3 for details of

7
the data base) and includes introductory and concluding

remarks.

Control system hardware and software are discussed in
Chapter 2. This chapter describes the first and second
phases of the control system and its operation. The design
considerations and tradeoffs that went into each phase are
discussed. Although these phases of the control system
have been subsequently superseded, they provided an
excellent foundation on which more sophisticated systems

were implemented.

Chapter 3 describes a data_base for multi-dimensional
data. The first version of this data base software was
developed jointly with researchers at Lawrence Livermore
National Laboratory (24-25), but was modified and extended
locally. This data base software is currently running on
three TOMS instruments, and has proven to be extremely

reliable and invaluable for the storage of MS/MS data.

A special data retrieval program is discussed in

Chapter 4. Described is a program that is able to extract
any two dimensional plane of data from the multi-
dimensional data stored in the data base. This is a

convenient and powerful tool for trend analysis and allows
the user to look at a matrix of data from several

orthogonal axes.

8

The final chapter describes some of the chemical
information available in MS/MS data, and presents several
ways of extracting some of this information from the data.
The concept of a neutral spectrum is introduced, it’s
utility is explored, and a simple expert system for the
analysis of these spectra is described. Also presented is
a grouping algorithm designed to cluster daughter and
neutral spectra so that the common structural features of
their parent molecules can be obtained. The substructural
information gained by using these techniques can be used to

help determine the overall structure of unknown compounds.

Conclusions

In this dissertation, the process of automating the
analysis of TOMS data is described, from the analog to
digital converter that samples the ion intensity, through
an expert system which samples the information present in
MS/MS data collected. Throughout this work, I have tried
to show that comprehensive software tools can aid the
chemist in such complex tasks as structure determination.
Someday these tools will be integrated with the addition of
an intelligent controller as shown in Figure 1.1. This
controller will be able to direct the experiments performed
by the instrument. The work that others and I have done in
Dr. Enke’s lab (solid lines and boxes in Figure 1.1) sets

an excellent foundation for these higher level systems.

Chapter 2

Instrument Control

Introduction

In the fall of 1978, the triple quadrupole mass
spectrometer designed by Drs. Richard Yost and Christie
Enke (2) produced it’s first daughter spectra. This
instrument, while designed to be controlled by computer,
was initially operated manually. The early TOMS instrument
used a ramp generator to sweep the mass selected by one of
the quadrupoles and had a strip chart recorder for data
output. To collect daughter spectra, the first quadrupole
was manually set (with a potentiometer) to the mass of a
parent ion (identified by an oscilloscope), the sweep
generator was switched to the third quadrupole, and a scan
was taken. The strip chart recording was then measured and
a mass scale added. As this process indicates, the
collection and analysis of each spectrum was a time

consuming task.

The design of a general purpose, single or multiple
microprocessor system for instrument control had been
initiated in the fall of 1978 (26). One of the goals of
this research was complete control of all instrumental
parameters of the TOMS. We realized that this would not be

9

10
accomplished overnight, and decided to implement computer
control in three phases. The first phase would be a simple
digital strip chart recorder, the second phase would be the
implementation of the new microprocessor system hardware,
and the final phase would be complete computer control of

the instrument with a multiple microcomputer system.
Phase 1: Digital strip chart recorder

The first phase of the TOMS automation was the
implementation of a digital strip chart recorder. This
allowed a computer to generate the sweep signal for mass
selection and to collect the ion intensities. A computer
can then assign the mass values and display the data in

several different formats.
Digital strip chart recorder: Hardware

The INTEL 8085 microprocessor chip was chosen to be the
heart of the general purpose control computer. An 8085
microcomputer evaluation board, the SUE-85, was convenient
for implementing the first phase of the TOMS automation.
The SDK-85 is a microcomputer with a monitor (in read only
memory, ROM), a limited amount of program memory (random
access memory, RAM), a keypad, an eight character display
and an area for custom hardware. In the extra space

provided on the SDK-85, I wire-wapped interfaces to the

11
instrument, a floppy disk drive and a keyboard/display unit
along with some extra RAM and ROM (see Figure 2.1). The
terminal I constructed for this system included two video
memories; a graphics memory which provided a low resolution
display (256 pixels horizontal by 240 pixels vertical) and
a text display which provided 16 lines of 64 characters.
Both of these memories, -a 9 inch monitor, a parallel
keyboard and a disk drive with an intelligent controller
were mounted in a terminal enclosure. This terminal was
connected to a parallel port on the SDK-85 with a ribbon
cable. The instrument was interfaced with a single
digital-to-analog converter (DAC) to control the mass.
selected by one of the quadrupoles, and an analog-to-
digital converter (ADC) to measure the ion intensity. The
ion current was converted to a voltage and amplified by a

Keithley model 18000 programmable current amplifier.

The disk drive (model 270) and intelligent controller
(model 1070) were manufactured by 'PERSCI. The disk
controller had an INTEL 8080 microprocessor which handled
all the disk functions (seek, read, write, etc) and
maintained a file structure. The use of this controller
saved much development time, since all the disk related
functions were already done. I wrote a program for the lab
PDP-ll minicomputer (PIPERSCI) to read and write PERSCI
formatted floppy disk. With PIPERSCI, programs are created

on the PDP-ll (with good editors and cross-compilers) and

12

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

SDK-85
8085 CPU
— RAM J
— ROM J
— A/ D J<-— from current amplifier
:: D/ A F9 to quadrupole controller
' """ t Griffith-3583;53-
TEZI video J

 

 

* graphics 1
board f
keyboard
_ disk
flontroller

 

 

 

 

 

 

——-----------‘-—-- h‘-—--

First TOMS control system

Figure 2.1

13
carried to the microcomputer for use. Data stored on a
floppy by the microcomputer are transferred to the PDP-ll

for post-processing and display.

Digital strip chart recorder: Software

Software for this control system was written entirely
in assembly language. The initial versions of the code
were assembled and typed in by hand; later versions were
cross-compiled on the PDP-ll and transferred to the
microcomputer on floppy disk. ‘ A bootstrap for the
microcomputer disk was burned into a ROM (after interfacing
a PROM programmer to the lab PDP-ll and writing a program

to burn the PROMs).

The final version of the control software for this
phase consisted of only a few commands (summarized in
Table 2.1). These commands consist of only the basics:
load in a new program, set the number of data points to
average, set the threshold level, scan a quadrupole, store
collected data to disk, and display data. This system was
not meant to be the final word in mass spectrometry control
systems, but was designed to replace the strip chart
recording method with a digital method and gain the
experience and tools required to design and implement a
more complete system. With this system, scans could be

initiated and data could be collected and stored on disk.

14

Table 2.1

Summary of commands for the digital strip chart software

CHANGE

FSCAN

SCAN

DISP

WRITE

query the user for the following:
low mass
high mass
mass increment
number of points to average
threshold for saving data

quickly scan from low mass to high mass to
observe the peak on an oscillosc0pe

scan from low to high, collecting data
display the data on the graphics display

write the data to the disk

15
Post processing and display of processed .data were
accomplished on a PDP-ll computer using programs written by

Phil Hoffman (18,27).
Phase 2: Single micro control system

The second phase of instrument control proceeded on two
fronts, the development of the control hardware and the
software systems. The design and development of
microprocessor hardware modules was done primarily by Bruce
Newcome (26,28). My role in the development of the
hardware was as the software consultant. Together, we
would determine whether a hardware design ’feature’ would
make programming the hardware easier, or conversely make
the software much more difficult to code. By designing
both the hardware and software together, we were able to
trade off functions between hardware and software for the
most efficient operation of the resulting system. The
hardware will be briefly described here to enable the

reader to gain appreciation for the control software.
Single micro control system: Hardware

As can be seen in Figure 2.2, the Newcome micro-
processor has bus connections on two levels; each module is
attached to a ’mother-board’, which is in turn plugged into

a backplane. Each ’Bruce-bus’ microprocessor module

16

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

A typical single microprocessor system

 

 

 

 

 

Figure 2.2

l7
performs one specific function. A set of these modules are
joined to comprise a microcomputer system. The ’standard’
microcomputer board consists of an 8085 microprocessor
board, a memory board capable of holding either RAM or ROM,
two serial ports and an interrupt controller. This
hardware modularity allows a great deal of flexibility in

the design of a control microcomputer system.

Specific modules for the triple quadrupole mass
spectrometer control consisted of two digital-to-analog
converters (DACs), one analog-to-digital converter (ADC), a
controller for a Keithley programmable current amplifier,
two video displays and a keyboard. Each of the DACs
controlled one quadrupole power supply, allowing the
software to select masses in both quadrupoles. With
computer control of the amplification range of the Keithley
amplifier, the microcomputer was able to sample the input
signal and adjust the gain for a 10s dynamic range. The
graphics and alphanumeric video memories were split to
separate display units, allowing simultaneous display of
data and of the collection parameters. The terminal
constructed for the first phase was dismantled and the disk

drive was built into the TOMS console.

18

Single micro control system: Software

The development of microcomputer systems from the chip
level necessitated the development of software for testing
both hardware and software. At that time, there were no

affordable, commercially available software packages that

could be adapted to our needs. The first versions of the
home-built microcomputers were debugged with highly
specific software; the first operating processor/memory

combination did nothing but flash a light on the CPU board.
As the hardware matured, so did the software. A software
monitor (called "Dr. Memory"), able to operate on any
standard CPU board, was developed and refined, and used as
the basis of more complex software systems. This monitor
is the equivalent of a simple operating system, complete
with device drivers for the commonly used peripherals (i.e.
the terminal and disk). This frees the user to program
the higher level functions without concern for the details
of the hardware registers. Although this monitor eased the
hardware dependencies of the applications software, a more
complete programming system and command interpreter was

needed.

A library of commands, and a programming system was
developed (the structured library oriented programming
system, SLOPS) as a basis for specific control

applications. Finally, control software for the triple

19
quadrupole mass spectrometer was implemented using the
tools provided by both Dr. Memory and SLOPS. In practice,
all three levels of programming (monitor, library and
control software) were developed concurrently, but they
will be described separately below. Today, one would
choose from a variety of convenient and inexpensive hard-
ware and software modules that are commercially available,
but at this. relatively early time in microprocessor
applications, we were on our own. Figure 2.3 shows the
relationship between currently available software modules

and Dr. Memory, SLOPS and the applications programs.
Dr. Memory’s monitor

A monitor for the home-built microcomputer systems
needs the following characteristics: handle all terminal
I/O, allow display/change of memory locations, execute code
starting at any location, stop and restart execution of
programs and load software. Dr. Memory was designed with
all these features in mind, and the code fit in only
2 Kbytes of ROM, complete with help screens, communication

software and general purpose I/O subroutines.

Dr. Memory is a general purpose monitor used for both
hardware and software debugging. The majority of the
available commands (listed briefly in Table 2.2 and more

fully in Appendix 1) are for examining and changing memory

20

Conventional programming SLOPS based programming

Applications programs Applications programs
subroutine libraries SLOPS
utilities -
operating system Dr. Memory

device drivers

Relationship between conventional and SLOPS programming

Figure 2.3

21

Table 2.2
Dr. Memory’s Monitor. (V2.6 9/3/80)
Commands Description
Break Breakpoint - stops program execution
<ESC> Cold restart - used to clean up user stack
0 Octal entry and display format
H Hexadecimal entry and display
? Prints a summary of this text
a/ Open location a for modification
a\ Opens two bytes for modification
n<CR> Modify open location, close it
n<LF> Modify, close, open next location
nA Modify, close, open previous location
Sr Open register <abcdefhilABDHSP>
aG Start user program at location a
P Procede (at saved PC)
S Start the second EPROM (SLOPS)
T Talk to the PDP-11
L Download from the PDP 11/40
Note: a’s and n’s are optional, defaulting to the last
value.
Restarts fig; Description
RST 0 C7 Cold restart - sets default stack, etc.
EST 1 CF A \
EST 2 D7 BC \ Diagnostics - prints
RST 3 DE DE ) the contents of the
RST 4 E7 HL / indicated registers
RST 5 EF PC /
RST 6 F7 SP, flags printed
RST 7 FF~ Breakpoint (warm restart)
Useful subroutines
Crlf Outputs a <CR><LF> combination
Downld Downloads data from the PDP ll
Efclr Clears an event flag
Getnum Gets a number input
Gettt Gets a character from the USARTs
Lights Sends a predefined light pattern out
Lite Sends a specified light pattern out
Nulljb Boredom routine
Print Prints an ASCII string
Putnum Outputs a number
Puttt Writes a character to a USART
Rhlr Rotates HL right

22
locations, and starting programs. New pieces of hardware
are easily checked out with this tool. The new interface
hardware is connected into the system, the power applied,
and the operator can use Dr. Memory to access the registers

of the new device and manually exercise and test it.

Software (8085 assembly language) entered into the lab
PDP-11 computer is cross-compiled and downloaded to the
microcomputer. Dr. Memory has two commands that facilitate
this process: TALK and DOWNLOAD. TALK is a routine to
connect the user’s terminal to the PDP—ll through a second
serial port. The DOWNLOAD command instructs the
microcomputer to accept a program from the PDP-ll, and load
it directly into memory. Dr. Memory is used to start
execution of these new routines, and stop execution at any

time to examine memory or registers.

This monitor includes a series of software debugging
tools to aid the assembly language programmer. The 8085
restart commands (EST 0 to RST 7) are software interrupts,
used in the monitor as debugging breakpoints and display
routines. Restarts 1-6 print the contents of different
registers and restart 0 initiates a cold restart (as if the
power were just applied). The breakpoint restart (RST 7)
is the most powerful restart for the programmer, returning
control to Dr. Memory. The complete context of the user’s

program is saved, and the programmer is able to examine and

23
change registers and memory locations. When control is
returned to the program under test, the program’s context
is restored and program execution continues as if the

breakpoint never occurred.
Dr. Memory’s monitor internal structure

Internally, Dr. Memory consists of data tables,
interrupt handlers, subroutines and a command processor.
All the terminal I/O’s are interrupt driven, so commands
can be entered and buffered while the microcomputer is
performing another task. These commands are executed when
the current command finishes. A break character from the
console terminal acts as a special command, caught by the
interrupt handler, which causes the microcomputer to stop
executing the current program, saves the program context
and returns control to Dr. Memory. The command processor
is a simple program that accepts input from the terminal
and checks it against the list of available commands.
Valid commands are executed while undefined input returns a

"Say what?" response.

A special subroutine, NULLJB, is called whenever the
microcomputer is not executing a command or when terminal
input is pending. This subroutine continually checks for
an ’event’ to occur (a character input from either serial

port, a timer to go off, or any other interrupt process),

24
at which time control is returned to the calling program if
the program was waiting for that event. While NULLJB is
waiting for an event to occur, it flashes the lights on the
processor board in a characteristic pattern. By watching
the microcomputer, one can instantly tell if a program is
running or waiting for input.

l

Structured Library Oriented Programming System

The structured library oriented programming system
(SLOPS) defines a structure for both commands and
subroutines (which are identical in SLOPS’ view of the
world). The base SLOPS system fits into 2 Ebytes ROM, and
consists of a command processor (more elaborate than Dr.
Memory’s command processor) and a series of subroutines
and/or commands. The basic premise for SLOPS is to create
a library of entries (either subroutines or commands) which
can be built upon to create new, more elaborate entries.
These entries are chained together to form a linked list

which can be quickly scanned.

SLOPS is conceptually similar to the FORTH programming

language (29), in that they are both threaded, interpreting

compilers. When either programming system is started, it
is in interactive mode. In this mode, commands entered are
immediately executed. A new command definition, when

entered, is "compiled", stored in memory, and ready for

25
execution. This compilation phase searches the library (or
dictionary, as it is called in FORTH) for each word in the
definition and replaces it with the address of that
subroutine, or assembles the instruction if necessary.
This process of combining previously defined modules to
form more complex functions is the basis of threaded

programming.

The major difference between SLOPS, FORTH and CONVERS
(a FORTH style language developed by Bonner Denton, et. al.
at the University of Arizona) (30-31) is the way arguments
are handled. Both FORTH and CONVERS are post-fix languages
(similar to Hewlett-Packard calculators), while SLOPS uses
pre-fix notation. In FORTH, arguments for the subroutine or
function are assumed .to be pushed onto an internal stack
followed by the function (i.e. 2 2 +), while SLOPS
functions assume arguments will follow the function name
(i.e. + 2 2). Both of these notations have merits, but
pre-fix notation seemed clearer, especially for single

argument functions (i.e. MASS 41).

At the time we needed to implement the control software
for the TOMS instrument, FORTH was only available for
PDP-11 minicomputers. CONVERS was developed at. the
University of Arizona as a FORTH style language for the
'INTEL 8080 series microcomputers, and was implemented in

our lab by Eric Carlson (32) for an early version of the

26
multi-microcomputer system used to control a stopped-flow
spectrophotometer. Instead of porting CONVERS to the new
microcomputer hardware, we decided to use the experience
and concepts gained from the development of CONVERS to

implement a new, hopefully better, programming system.

SLOPS internal structure

Each library entry consists of a header flag, the name
of the entry, a link to the next entry, and the entry
itself. The header flag byte contains the length of the
entry’s name and a two bit flag describing the type of.
entry. Only three types of entries are currently used.
Figure 2.4 shows each of the entry formats. Two of the
formats have variable length entries; hence they have link
fields pointing to the next entry in the library. The
other two types of entries have fixed lengths and do not

require link fields, saving a few bytes of RAM per entry.

The subroutine or command library entry is the most
common type of entry. The code for the subroutine starts
immediately after the link field and can be as long as the
available RAM. The macro library entry is presently
unused. The label and assembler library entries are used
for _the built in assembler/linker. The data field of the
label entry is two bytes long, and can be either a value or

an address (absolute or relative). The data field for the

Entry type

Subroufine

Macro

Label

Assembler

flag byte format:

27

Entry format

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

IflagI label IlinkI subrofii‘ne
l l l:L L l l l—E
IflagI IabEI‘ IlinkI macFo
Iiiiyiitiix,
flagI lobar
l 1%: l
label's value
Iflag |ab3l I
l L h
op code
Ldent? length I
depen ant of label.
0 0 Subroutine entry
0 1 Label entry
1 0 Macro entry
1 l

Assembler entry

SLOPS library entry format

Figure 2.4

28
assembler entry is one byte long, and is the microprocessor

operational code for this mnemonic.

The command processing routine of SLOPS prompts the
user for one or more lines of input, and breaks the input
stream down into ”words”. These ”words" are defined as
alphanumeric strings delimited by non-alpha characters
(spaces, commas, quotes, etc.) The input words are kept in
a forward-chaining stack, and the first word in a line is
assumed to be a command. The command processor searches
the library for a match, and if a subrbutine/command is
found, it is executed. If the executing procedure needs
input parameters, it gets them from the word stack; if no
words are available, SLOPS prompts the user for the
required input. Any executing task is able to call any
subroutine in the system, recursively if necessary. When
any procedure is finished, it simply RETURNs to the program
that called it. The top level command, when done, returns
to SLOPS. If more words are in the word stack, they are
then executed in turn until the stack is empty, and SLOPS

prompts the user for more commands.

The base SLOPS system contains a number of useful
subroutines as well as the command kernel. Table 2.3
contains a list commands and brief descriptions of each
subroutine or command, and Appendix 1 is an abbreviated

form of the user’s manuals for both Dr. Memory and SLOPS.

Subroutines

Addr
Ascii
Blank
Brkdwn
Check
Cvtext
Cvtint
Dcmp
Delay
Link
Number
Search
Ttyin
Word

29

Table 2.3

Basic SLOPS subroutines and commands

Returns code address of library entry
Converts binary to and from ASCII
Clears the screen

Breaks input line into words

Compares word with library entry
Converts a number to ASCII string
Converts ASCII string to binary number
Double compare

Software time delay

Links to next library entry

Get number from word stack

Search library for a match

Gets a line of input

Get word from word stack

(also see subroutines listed under Dr. Memory)

Arithgetic subroutines

Ddiv
Div
Dmult
Dsub
Mult

Commands

Convert

Downld
Drmem
Talk

Double divide
Divide

Double multiply
Double subtract
Multiply

Converts a number to any base

Loads from the PDP 11 (See Dr. Memory)
Puts Dr. Memory in control

Talk to the PDP 11 (See Dr. Memory)

30
As can be seen from this list, about half of the
subroutines are used by SLOPS to keep track of library
entries, etc. The other half are general purpose
subroutines (double multiply, convert binary to/from ASCII,
etc.) An extended SLOPS system includes an 8085 assembler

and many more general purpose subroutines.

The extended SLOPS system is a powerful tool for
software development. The user is able to create and test
small subroutines and build more complex systems using
these building blocks. Execution speed of the resulting
commands is excellent due to the extensive use of assembly
language and the low overhead of the command processor.
When a library of low level commands is built, this system
behaves like a higher level programming language, allowing

the programmer to use structured programming techniques.
Mass spectrometer control software
SLOPS provided a good foundation for control software

for the triple quadrupole mass spectrometer. The new

control hardware featured two video displays, one for

graphics and another for status display and user
interaction. The graphics screen could be logically
divided into two halves termed the upper and lower
displays. Each of these displays were equivalent, but

could be programmed independently. The status display also

31
had two ’halves’ to control each half of the graphics
display. The status display, shown in Table 2.4, displayed
the current status of the instrument, the mass range for
each quadrupole to scan and the threshold for gathering and

storing intensity data.

Table 2.4

TOMS status display

UPPER LOWER

Mass: Quad 1 xxx-xxx xxx-xxx

Quad 3 xxx-xxx xxx-xxx

# pts to avg xxxxx xxxxx

threshold xxx.x:-x xxx.xz-x

min/max range -x/-x -x/-x
Quad 1 Quad 3 Intensity
DC:xxx.x RF:xxx.x xxx.x:-x

The two halves of each screen effectively allow two
separate experiments to be run simultaneously. This
feature, as well as much improved graphics and the ability
to change the gain on the amplifier during data collection,
made this control software much more convenient and

versatile than the ’digital strip chart’ of phase 1.

The commands available to the operator of this system
are listed in Table 2.5. The control software allows four
sets of parameters to be stored (with the SAVE command) and
retrieved (with the GET command). This allows the operator

to set up several different experiments and quickly switch

INI
GET
SAVE

PARAM

MANUAL

FSCAN

SCAN

WDATA

32
Table 2.5
TOMS control system commands
Initializes the system, clearing the displays,
setting default mass values, etc.

Gets a set of parameters from one of four stored
setting areas.

Saves the current parameter table in a bank of
parameter tables, for future recall.

Sets general parameters. It acts in a ’single
character input’ mode for the parameter to change,
i.e. ’O’ not ’OUAD’ for changing a Ouad’s mass
range. The parameters for either graph, the UPPER

or the LOWER, must be changed individually, and the
commands U and L identify which graph is being
changed. -

The following commands are currently supported:

O n low~high Changes Ouad n’s mass range
T int range Changes the threshold
R min-max Changes the min and max range
A n Changes the # of points to average
G low,n Changes the low mass on the graph
F Changes flags, each flag queried
M text Puts message on the status display
H text . Puts a header (title) on the graph
? Prints a summary of the options
AZ exits (control Z)

Implements ’manual’ control. Using the keypad of

the terminal, the keys 1, 2 and 3 are for quad l; 4,
5 and 6 are for quad 3; and keys 7, 8 and 9 are for
the range. Keys 1, 4 and 7 decrease the current
value by 1, keys 2, 5 and 8 set the current value
into the parameter table, and keys 3, 6 and 9
increase the current value by l.

Increments the mass, checks for a typed character,
and if none entered, leaps about.

Scans the set mass ranges, collects data, updates
the screens, records the data and loops until a
character is typed.

Writes the data from RAM to a disk file.

33
between them. Setting parameters is accomplished by using
the PARAM command. The parameter changes are immediately
displayed on the status and/or graphics displays. The
MANUAL command allows the operator to step the quadrupole
masses by single DAC units to accurately find the maximum
of a peak. For example, when set up for a daughter scan,
the MANUAL command allows the operator to tune the first
quadrupole mass to the maximum intensity (which might not

be an integral mass due to calibration inaccuracies).

Two commands, FSCAN and SCAN sweep one or both (as
defined on the status display) quadrupole masses from the
preset minima to the maxima. The FSCAN command sweeps
quickly, allowing the spectra to be displayed on an
oscilloscope. The SCAN command is necessarily slower since
data are being collected and averaged. Each of these
commands sweeps the entire specified range of one or both
displays, and restarts the sweep unless told to stop by any
keystroke on the keyboard. One last command, WDATA, writes
the data collected by a SCAN command to the disk for post

processing by the lab PDP—11 computer.

This control software is written as a series of modular

subroutines allowing for independent testing of each
module. Appendix I briefly describes the library of
subroutines created for and used by this system.

Enhancements to the control software, such as special

34
purpose commands, are easily constructed using these

routines as a foundation.
Phase 3: Current mass spectrometer control system

The single microcomputer system described above was a
temporary solution to the control problem. When that
system was implemented, we were aware that the system could
not be operated at the high speeds we desired. The third
and current phase of instrument control was accomplished
with a multiple microprocessor system, designed and
implemented to allow several tasks to run concurrently. A
full description of the control hardware can be found

elsewhere (26,33).

Commercial software tools and systems were being
developed and becoming available during the implementation
of the SLOPS based system. These tools were now
sufficiently powerful, flexible and affordable that we
decided to implement the third phase of software control
using these tools. The control software for the multiple
microprocessor system was written in the FORTH programming
language, now available for microprocessors. FORTH is a
high level programming language commercially available that
is well suited to control systems. The experience gained
from the SLOPS based control system was invaluable in the

design and implementation of the current FORTH based

35
software. Complete descriptions of the current control
software can be found in Carl Myerholtz’s thesis with

modifications by Adam Shubert and Mike Kristo (11,34-39).
Conclusions

The triple quadrupole. mass spectrometer is a complex
and experimental instrument, and automation could not occur
in a single step. The three-phase process described in
this chapter worked very well, allowing us to become
familiar with the advantages and disadvantages of various
techniques. The transition from a completely manually~
operated instrument to a completely computer controlled
instrument was arduous both in implementing the
harware/software as well as in convincing the operators

that a computer can reliably operate a complex instrument.

Chapter 3

Multi-dimensional Instrument Data Base

Introduction

The evolution of GC/MS brought with it the development
of computer data systems designed to handle the additional
dimension of time as well as the mass and intensity axes.
However, even three dimensions are inadequate to cope with
the measurement capabilities of MS/MS instruments and many
other modern computer-controlled systems. In the case of a
totally computer-controlled MS/MS, a large number of
variables can be scanned, either singly or jointly. Each
such scan will produce one plane of information in a multi-
dimensional data base. As an example, the axial energy can
be scanned at fixed masses or the mass (in either quad 1 or
quad 3) can be scanned with incremental changes in axial
energy. Other variables (dimensions) include direct inlet
probe temperature, collision gas pressure, ionization
voltage, chemical ionization gas pressure, lens voltages
(both in the source and between quadrupoles), collision gas
type and others. Thus, a new, more versatile data base

management system was required.

36

37

Fortunately, the storage of all these data has become
affordable using computers with large capacity disks and
magnetic tapes. Now the analyst can store all data as they
are acquired on disk, and subsequently put them on magnetic
tape for archival storage. Once these data are stored, the
problem becomes one of rapidly and flexibly accessing the
data. An additional problem in multi-dimensional
instrumentation is the need to extract any plane of
information in the data base, even if that plane does not
correspond to the types of scans that produced the data
base. The concerns of data retrieval are explored in the

next chapter.

The chemical literature contains many examples of
systems for creating and searching libraries (40-49). There
are also suggestions for using pattern recognition (50-55)
and artificial intelligence or heuristic (21,56-59) means
of (interpreting data. Little is said, though, about rapid,
versatile and efficient initial storage of raw data in real
time. We have developed a system that provides these
storage characteristics, provides for storage of multi-
dimensional data, and incorporates mechanisms for rapidly

accessing the data.

38

Scientific versus business data bases

In many ways, both business and scientific data bases
are similar, but there are significant differences (60-74).
Business data bases must be able to handle text ("warehouse
X"), integers (52 widgets left in stock), and floating
point numbers ($42,000) (60-72). In addition to these
requirements, ’scientific data bases may also have to
accommodate bit strings, vectors, arrays, graphs and multi—
dimensional data (63-74). Both types of systems must have
query languages, or methods to retrieve the data stored in
the data base. Business systems usually require a small
set of specific answers (all employees with salaries
greater than $30,000). Scientific data base retrieval
languages or programs must .be able to handle the
uncertainty (error) present in many scientific measurements
(68). Queries must be made in the form ”retrieve all
energies between -20 volts and -18 volts". Both data base
query systems must be able to join several queries ("names
of employees older than 62" AND "been with the company
longer than 15 years"), and display the results as lists,

tables or graphs.

The data formats of business data bases must be defined
before, use by the data base administrator using a data
definition language, and the data are often entered by

keypunch operators. The structure of the data base must be

39
carefully considered and analyzed, for the resulting data
base may be in use for many years. This data base
definition process can take weeks or months, and has
spawned an industry that provides data base definition
templates for the popular data base programs. Scientific
data bases, especially those created by during an
experiment, must be created quickly (possibly
automatically) and be able to accept data from the ongoing
experiment in real time. A scientist does not have the
time to spend weeks organizing a data base for one
experiment, and then redefine the data base for the next!
A scientific data base must be easily adapted to new and

rapidly changing problems.
Survey of data base structures

There are essentially 4 major structures for data
bases: flat file, hierarchial, network and relational
(so-61,72). or these, the flat file is the simplest, and
is used in the majority of the ”spread sheet” programs for
micro- and mini-computers (75-76). These flat files are
two-dimensional tables consisting of columns for each
variable or attribute, and rows for each observation or
‘instance. If the data to be recorded are amenable to this
form with minimum duplication of values, this format is

efficient and can represent a natural working order of data

40
from right to left and top to bottom. Example 1 in

Figure 3.1 illustrates a simple flat file data base.

Hierarchial data bases provide 1 to N links between
records, such that one root record may have one or more
"child” records, each of which may have one or more
"grandchild” records, and so forth. An illustration of an
hierarchical data base can be seen in example 2 in
Figure 3.1, where Prof. Smith teaches 3 classes and Prof.
Jones teaches two others. By providing the links between
the records, redundant information (Prof. Smith teaches
math 101; Prof. Smith teaches math 103, etc.) is.

eliminated.

Network model data bases provide for N to M linkages
between records, and have all the features and advantages
of hierarchial data bases. In addition, they provide more
flexible ways to interconnect records, allowing almost all
redundant information to be eliminated. Example 3 in
Figure 3.1 shows a simple network data base.. As is
expected, this increased flexibility comes at a price;
network model data bases are more complex than hierarchial
models, and often require special training for data base

administrators.

Relational data bases make a radical departure from the

linkage schemes used in both hierarchial and network

41

 

F’rof. Smith I math 101
IProf. Jones I math 413

room 42

r room 123
l

1
1

 

Example 1: flat

   
    
    
   
   

Prof. Smith

Prof. Jones

file

moth 101
math 103
math 142
math 413
math 417

   

   

Example 2: hierarchial model

 

 

 

 

 

]< moth 101 student Miller
[me' Smith math 103 student Johnson
I Prof. Adamk .

. physncs 101 student Webber

 

 

 

 

 

 

 

 

 

Example 3: network model

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

student Miller moth 101
Prof. Smith moth 101 student Miller physics 101
Prof. Jones math 413 student Johnson moth 101
Prof. Adams physics 101 student Webber math 103
I student Webber physics 101
1 relational link j .
Exampie 4: relational model

Data base structures

Figure 3.1

 

42
models. In the relational model, all data are stored in
flat file format, and "relations" are established between
various tables. In this manner, implicit links are made
between relations as opposed to the explicit links needed
in hierarchial and network models. The data base system
creates these links as needed by "joining" tables together
on the fly. Example 4 in Figure 3.1 shows an example of a
simple relational data base. This type of data base is
easy to maintain (the flat files are conceptually simpler),
but require more computational power to join relations for
all but the simplest queries. For example, from
Figure 3.1, a query ”list all students of Prof. Jones"
requires the course number from the two tables to be
joined, and only students in one of Prof. Jones’ classes to

be listed.

Data base capabilities

We have included in our system several capabilities
that have been either overlooked or not necessary in past
data base management implementations for scientific

instrumentation (see Table 3.1).

These extra capabilities of the multi-dimensional data
base system are essential for use with our triple
quadrupole mass spectrometers. There is a large degree of

interdependence among the TOMS parameters, and the effects

43

Table 3.1
Capabilities of the multi-dimensional data base
1) The ability to store all instrument parameters
2) The ability to store a variety of changing parameters
3) The ability to store any X/Y data pairs (such as
mass/intensity and energy/intensity pairs)
4) The ability to add comments before, during or after

the experiment.
5) One compound or experiment for each dataset.

of these dependencies are not yet completely understood.
Storage of all instrumental parameters along with the data
is required if we are to fully understand these parameters.
To study these interdependencies effectively, the operator
may vary several parameters incrementally to obtain a
dataset with multiple dimensions of data. These data can
later be searched in a variety of ways to study the effects

of individual instrumental parameters.

For many research applications, complete mass spectra
or daughter spectra are not required. For an energy
dependence study, the operator may want to vary the axial
energy for a certain parent/daughter fragmentation. This
data base is able to handle these data (intensity vs. axial
energy) as well as the more usual mass spectral data

(intensity vs. mass). The additional capability ‘Df

allowing the operator to enter comments at any time clur‘i-r‘g

the experiment allow the dataset to be used as a notebook ’

 

44
storing information about that experiment with the

experimental data.

An example of multi-dimensional data

The control systems for the triple quadrupole mass
spectrometers at MSU have the ability to devise "methods"
which allow the rapid collection of multiple dimensions of
data. This is useful for characterizing both the
instrument as well as a chemical sample. A method to
collect data for a study of axial energy and collision gas

pressure would follow the outline presented in Table 3.2.

Table 3.2

Method to collect 5 dimensions of data

1) LOOP: collision gas pressure: from low to high.

2) LOOP: axial energy: from low to high.

3) LOOP: quad 1 mass: set to parent in the EI spectrum
4) scan quad 3. _

5) continue to the next parent mass

6) continue to the next energy

7) continue to the next pressure

This method creates a five dimensional dataset:
collision pressure, axial energy, quad 1 mass, quad 3 mass
and intensity. This- dataset contains a wealth of
information about the compound, and the data are collected
automatically (except for setting the collision gas

pressure). The operator is then able to search through

45
this dataset and extract a plane of information, even if
that plane wasn’t specifically scanned. For example, with
this dataset, we can extract data to produce a plot of the
intensity vs. axial energy for a certain parent/daughter
pair at a specific collision pressure. Details of data
retrieval and extraction from a dataset are presented in

the next chapter (24-25,77).
A programmer’s view of the data base

The data base definition is complete when the variables
to be studied are defined. Since many experimental.
parameters will not change, an hierarchial model was chosen
for this application. At the top level or root of the
hierarchy are the parameters that do not change during the
course of the experiment. A second level was implemented
in order to store a traditional mass spectrum as one record
in the data base. This structure was choosen to reduce
redundancy in the data base, and hence conserve disk space.
The details of the data base structure are presented later

in this chapter.

The portion of this system used to write and read data
to and from the data base consists of a series of FORTRAN
subroutines. These subroutines are easy to incorporate
into any computer-controlled instrument. Only a few 11088

of code are needed to implement this data base system into

46
the control software (see Figure 3.2). All the work of
managing the data in a dataset is accomplished by a set of
subroutines which frees the programmer to concentrate on
other aspects of the project. Appendix 2 provides more
information regarding the subroutines used for data storage

and retrieval.

Because the subroutines are short and data storage is
efficient, this system could be added to many existing
dedicated instrument control computers. It could also be
put on a separate, time-shared computer system, with a high
speed data link. Both of the triple quadrupole mass
spectrometers at MSU are controlled by one or several
microprocessors linked to a minicomputer which handles the
data storage and retrieval. This allows for the efficient
separation of tasks, the data acquisition and the data
storage and retrieval. By using separate computers, the
instrument can be collecting data while other operators

analyze their previously collected data.
Data base file formats

Computer files are generally organized in one of three
ways: sequential, indexed or direct (60—62,71,78). A
sequential file is the simplest, and as the name implieS.
is a sequentially ordered collection of information, placed

in the file as received and packed in an unstructured WHY-

000

10

47

PROGRAM SAMPLE

INCLUDE ’MDDB.CMN’ ! include the commons
REAL*4 X(500),Y(500) ! arrays for data

LOONTI = 5 ! define the terminal LUN
CALL MSINIT ! initialize things

IF (IERR .NE. 0) THEN ... 9 check for errors

IFILE = 1 ! define the dataset number

CALL MSOPEN(IFILE,’New dataset name? ’,’NEW’,2,3,4)
! open new dataset, using

2 LUNs 2,3,4.
Now define the variables
NUMSTC(IFILE) = 3 ! three static variables
ISTATC(1,IFILE) = 23 2 the first is code #23
RSTATC(1,IFILE) = 2.0 9 and it’s value is 2.
NUMVAR(IFILE) = 6 ! six variable parameters
IVAR(1,IFILE) = 42 2 first variable is #42
CALL PUTPRM(IFILE) ! and write to the dataset

(fill up the X, Y arrays with data)

RVAR(n,IFILE) = value ! record the variables
NUMDAT(IFILE) = xx ! number of data pairs
CALL PUTDAT(IFILE,X,Y) ! and store the data

IF (more) GOTO 10

CLOSE (UNIT=2) ! close all the files
END ! and all done

Writing to the multi-dimensional data set

Figure 3.2

48
If these data are written in such a way that they can be
printed or displayed directly without any processing, the
file is called a sequential, formatted ASCII file. This
type of file is almost never used for large amounts of
data, as it requires a great deal of space on the disk and

is cumbersome to access.

A variation, 'of an unformatted sequential file, is
again sequential, but the data are written in internal
(unformatted or non-printing) format. This is a compact
form of storing data. It has been used in some instrument
data systems, but has several drawbacks, the most severe
being access speed. As with any sequential file, the only
way to find any particular piece of data is to start at the
beginning of the file and read all entries until the
particular datum is encountered. For large datasets, this

is a very slow process.

A second type of file format is an indexed file. In
this type of file, a "key field" is associated with some
part of the data record. the keys from all the records are
collected in one or more structured indicies, and access to
the records are provided through these indicies. This is a
powerful file format, but unfortunately is not part of the
FORTRAN-77 standard. It was not used in order to remain
within the standard. Another drawback is the time required

to write a record; after the data are written, the file

49
system must update the indicies which could slow the data

collection process down.

A third type of file is a direct access file. The file
is broken into fixed length records, and any one of these
records may be accessed almost immediately. An elementary
direct access file, consisting of one record per scan,
could be used by an instrument that always puts out a fixed
amount of data. For mass spectrometry, this format would
waste a great deal of space, as the fixed record length
would have to be long enough to hold the maximum number of
mass/intensity pairs that might be recorded in any one
scan. Table 3.3 shows the advantages and disadvantages of

sequential, indexed and direct access files.

Table 3.3

Comparison of sequential and direct access files

 

Sequential Indexed Direct
Record length variable fixed fixed
Time required fast medium fast
to write record
Access speed to slow fast fast
specific datum
Memory/storage very wastes wastes

space compact space sapace

50

In our system we have combined elements of several file
types to create a fast, efficient system. An unformatted,
sequential file, the header file, is used to store a
variety of information that describes the instrument, the
experimental conditions, and the variables to be recorded.
This file also contains the instrumental parameters that
did not change during the given experiment, and these
variables become the root of the data base hierarchy. A
second file, the pointer file, is written with short,
direct access records. The major function of this file is
to record the starting record number (pointer or index) of
the variables and data pairs located in the third file.‘
This file also contains some redundant data for fast data
retrieval. The third file, the data file, is also direct
access and contains the values for variable parameters and
the X-Y data pairs acquired. Another file, the instrument
description or dictionary file, is also direct access and
each record contains a definition of a variable or

parameter. Each file is described in more detail below.
Dictionary file

The instrument description file, or dictionary,
contains complete descriptions ' of the instrumental
parameters that may be varied or recorded. To reduce
computer storage, all variables and parameters are assigned

a code number. These code numbers are the record numbers in

51
a direct access file which point to the descriptions of
each code number. Two descriptions are stored for each
code number, a short one (less than 20 characters) and a
long one (up to 57 characters), and one flag integer (see
Figure 3.3). The short descriptions are used for speed when
the operator is entering them, or when display space is at
a premium. The long descriptions are more suitable for
tables and graphs. The general purpose nature of the data
base is thus maintained by conferring the instrument-
specific parameter lists to this file, which is, in effect,
a kind of conversion table or template. Each different

instrument can have it’s own dictionary file.

The flag associated with each description in the
dictionary tells how that description is used. If the flag
is greater than zero for a particular code (say the code
for ”collision gas type”), the value associated with the
variable also needs to be looked up in the dictionary. For
example, if we were looking at code 81, "collision gas
type”, and had a value of l, we would display ”Argon", not
the numeric value 1 (see Figure 3.4, example 1). In this
case, the flag is a pointer into the dictionary, and the
value is an offset from this pointer. However, if the flag
were zero, we would use the definition itself; in this
case, there either isn’t a value (as in the "Argon" example
above), or the value has no physical meaning. In the final

case, when the flag is negative (or more precisely, a

52

Record Number Record Contents (80 byte direct access records)

 

 

 

 

 

 

 

 

1 A20 A58 l2
QSDICT OLDICT IPTDIC
QSDlCT Short description of instrument parameter
QLDICT Long description of instrument parameter
lPTDlC Flag: -1 ==> The value is the answer
(i.e. "70.0" eV)
0 ==> No values
. (i.e. ”Argon")
>0 ==> Must look up the value
(i.e. gas #1 = "Argon”)

END

More definitions

Instrument description file format

Figure 3.3

Example 1:

Parameter 3'81 has a value at 3

record __]

81

m+1
m+2
m+3

 

 

 

 

 

 

 

 

 

 

 

'CAD Gas Collision Gas m
me no
AR Argon 0
N2 Nitrogen O
HE Helium 0

 

 

Result: Collision Gas: Helium

Example 2:

Parameter #fo has a value of 70.0

record

 

 

 

20

 

eV

 

Electron Energy

 

 

 

Result: Electron Energy: 70.0

Use of the dictionary file

Figure 3.4

54
"-l"), the value itself has physical significance, such as

70.0 eV electrons (see Figure 3.4, example 2).

The instrument description file is a powerful feature
of this system. By changing this dictionary, the data base
management software can be used by almost any analytical
instrument. Table 3.4 shows a portion of the dictionary
for our triple quadrupole mass spectrometer. Modifications
to an instrument often add new variables, and many data
base systems would need to be fundamentally modified to
account for the new parameter. This software only requires

the addition of a new definition in the dictionary.
Header file

The header file serves primarily as a notebook, storing
a variety of numerical and textual descriptions of the
analysis. Included in this file are the values of all
instrument parameters that will not change in the
experiment. Examples of these static parameters are the
operator’s name, positive ion mode, EI spectra, etc. (see
Table 3.5, a sample dump of a header file). Other
parameters that will change in the course of the experiment
are called variables, and their codes are listed in this
file. The contents and order of data in the header file

are diagramed in Figure 3.5 and summarized in Table 3.6.

5)
9)
10)
13)

14)

18)
19)
20)
21)
22)
23)
24)
43)
44)
45)
46)
47)

43)
51)
52)
73)
74)
75)
76)
30)

55

Table 3.4

Definitions in the TOMS dictionary

Operator
1) John
2) Milton

Date

Scan type
1) lscan
2) 3scan
3) Dscan
4) Pscan
5) Nscan
6) Sweep

7) Stable ion
Neutral Loss Mass
Parent mass
Daughter mass
Source

1) CI
2) EI
Ions
1) Pos
2) Neg
SP
CI Gas
1) CH4
2) H2

3) CH4+N20
FC

~EC

eV
Ion Volume
Repeller
CI Drawout
EIV
Ql Lens 1
Q1 Lens 2
Q1 Lens 3
Q1 Offset
Ql Mode
1) RF
2) DC
3) Scan
Q1 Mass
01 Delta M
Q1 Res.
02 Lens 1
02 Lens 2
02 Lens 3
02 Offset
02 Pressure

Operator
John 0. Public
Milton Webber
Date of Experiment
Scan type
Quad 1 scan
Quad 3 scan
Daughter Ion Scan
Parent Ion Scan
Neutral Loss (gain) Scan
Potential Sweep
Stable ion Scan
Neutral Loss (gain) Mass (amu)
Parent mass
Daughter mass
Source Type
Chemical Ionization
Electron Impact
Ion Type
Positive Ions
Negative Ions
Source Pressure (Torr)
Chemical Ionization Gas
Methane
Hydrogen
Methane + Dinitrogen Oxide
Filament Current ,(Amperes)
Emission Current (Milliamperes)
Electron Energy (Volts)
Ion Volume (Volts)
Repeller Potential (Volts)
CI Drawout-Potential (Volts)
EI Ion Volume (Volts)
Pre-Quad 1 Lens 1 Potential (V)
Pre-Quad 1 Lens 2 Potential (V)
Pre-Quad 1 Lens 3 Potential (V)
Quad 1 Offset Potential (Volts)
Quad 1 Mode
RF Only (No mass filtering)
DC (Mass Filtering)
Scan
Quad 1 Mass (amu)
Quad 1 Delta M
Quad 1 Resolution
Pre-Quad 2 Lens 1 Potential (V)
Pre-Quad 2 Lens 2 Potential (V)
Pre-Quad 2 Lens 3 Potential (V)
Quad 2 Offset Potential (Volts)
Quad 2 Pressure (Torr)

81)

104)
105)
106)
107)
108)

109)
112)
113)
114)
135)
136)
140)
141)
142)
143)

Table 3.

CAD Gas

1) AR

2) N2

3) RE

4) SF6

5) 002

6) CH4

7) NZ

8) CH4+N20
Q3 Lens 3
Q3 Lens 2
Q3 Lens 3
Q3 Offset
Q3 Mode

1) RF

2) DC

3) Scan
Q3 Mass
Q3 Delta M
Q3 Res.
EM Lens 1
Conversion Dynode
EM Voltage
Peak Finding Thres
Min Peak Width
Max Peak Width
Scan Rate

56

4 (cont’d.)

Collsion Gas

Argon

Nitrogen

Helium

Sulfur Rexaflouride

Carbon Dioxide

Methane

Hydrogen

Methane + Dinitrogen Oxide
Pre-Quad 3 Lens 1 Potential (V)
Pre-Quad 3 Lens 2 Potential (V)
Pre-Quad 3 Lens 3 Potential (V)
Quad 3 Offset Potential (Volts)
Quad 3 Mode

RF Only (No mass filtering)

DC (Mass Filtering)

Scan
Quad 3 Mass (amu)
Quad 3 Delta M
Quad 3 Resolution
Pre-Electron Multiplier Lens 1
Conversion Dynode Potential (V)
Electron Multiplier Potential
Peak Finding Threshold
Minimum Peak width
Maximum Peak width
Scan Rate

57

Table 3.5

Dump of the parameters in the Reader file

11 variable parameters:

Homms‘lmmbwnw

l-u-o

Modification History
Registry Identification Number

Quad 1 Mass
Quad 3 Mass
Scan type

Quad 1 Mode

Quad 3 Mode

(amu)
(amu)

Neutral Loss (gain) Mass (amu)

Source Type
Ion Type

Quad 2 Pressure (Torr)

24 static parameters:

coooslozouswml-a

l3
14
15
16
17
18
19
20
21
22
23
24

Comments:

70.0
14.9
13.5
10.1
-19.1
21.6
0.000
-53.8
4.20
-11.6
-4.90
-25.1
-5.60
0.144E+04
-12.5
~14.0
0.000
0.000
0.000
0.150E+04
4.00
25.0
4.00
0.129E+05

TRIAL RUN.

Electron Energy (Volts)

Repeller Potential (Volts)

CI Drawout Potential (Volts)

EI Ion Volume (Volts)

External Potential (Volts)

Pre-Quad 1 Lens 1 Potential (Volts)
Pre-Quad 1 Lens 2 Potential (Volts)
Pre-Quad 1 Lens 3 Potential (Volts)
Quad 1 Offset Potential (Volts)
Pre-Quad 2 Lens 1 Potential (Volts)
Quad 2 Offset Potential (Volts)
Pre-Quad 3 Lens 1 Potential (Volts)
Quad 3 Offset Potential (Volts)
Electron Multiplier Potential (Volts)
Quad 1 Delta M

Quad 3 Delta M

Quad 1 Resolution

Quad 3 Resolution

Quad 2 Pressure (Torr)

Peak Finding Threshold

Minimum Peak width

Maximum Peak width

Scan Rate

Date of Experiment

Record Number

 

END

58

Record Contents (unformatted binary variable length)

 

 

 

 

 

l2 I2 A10 A8

 

 

 

NUMSTC QDATE QTIME
NUMVAR

NUMSTC Number of static (fixed) parameters
NUMVAR Number of variable parameters
QDATE Date the dataset was created (DD-MMM-YY)

' QTIME Time the dataset was created (HH:MM:SS)

I L

 

 

 

4. ‘_

 

 

 

 

 

 

:2 R4 l2LR4 “12 R4

 

 

ISTATC(I) ISTATC(Z) j . ISTATC(NUMSTC)
RSTATC(I) RSTATC(Z) RSTATC(NUMSTC)

ISTATC(i) Static parameter code number
RSTATC(i) Static parameter value

 

 

 

 

 

 

 

 

 

 

 

 

 

I l—F—-'
l2 l2 4 2— l2 ’
lVAR(l) IVAR(NUMVAR)
IVAR(2-)
lVAR(i) Code number for variable i
fir a--—»

'A10 A8 480
QDATE arm: ocoTANT
QDATE Date this comment was entered (DD-MMM-YY)
QTlME Time this comment was entered (HH:MM:SS)

QCOMNT ASCII comment

More comments

Reader file format

Figure 3.5

59
Table 3.6
Summary of Header file format
1) The number of static or fixed parameters (all
computer readable instrument settings that will not
be changed in this experiment)
2) The number of variables (those parameters which are
likely to be changed during the analysis)
3) Date and Time
4) The code numbers and values of static parameters
5) The code numbers of variable parameters

6) Comments as needed, with time of comment entry in
front of each.

The header file is constructed so the comments are
stored at the end. This allows comments to be entered
before, during and after an experiment, thus providing
exceptional archival value. The computer clock time is
prefixed to each comment so that the coincidence of
comments and particular sections of the collected data can

be established during post-collection analysis.
Pointer file

There exists one record in the pointer file foreach
scan performed on the instrument, and each record has as
it’s first entry a pointer into the data file. This allows
one to access any data from an experiment rapidly. The
other elements in the record are either for housekeeping,
or to enable faster display of the data. The last entries,
’the fast access variables, are especially important for

multi-dimensional work. These are copies of variables

60
stored in the data file. They are redundant, but this
redundancy gains speed when searching the dataset for

specific results (see the next chapter for more details).

Selected variables, kept in the pointer file, save
significant amounts of time by not accessing the data file.
For example, if we want. to examine the behavior of a
daughter ion from a specific parent, and one of the fast
access variables is quad 1 mass, we can quickly determine
if we must retrieve the data for this scan. If quad 1 mass
were equal to the parent of interest, we would get the data
for this scan. However, if quad 1 mass didn’t equal the.
parent of interest, we don’t have to access the data file,
and we’ve saved the operator some time. The time savings
can become significant (from seconds to minutes) for medium

to large datasets.

The contents and order of data in the pointer file are
shown in Figure 3.6. Each scan produces one record in the

pointer file. Each record contains the elements listed in

Table 3.7.

The pointer file was designed as a separate file for
several reasons. First, since it is a small, direct access
file, it can be quickly accessed. Each read of this file
immediately brings the important aspects of this scan to

light, i.e. the dependent variable, number of data pairs

Record Number

 

END

61

Record Contents (32 byte direct access records)

 

 

I2 I2 R4 R4 l2 I2 R4 l2 R4 l2 R4

 

 

 

 

 

 

 

 

 

 

 

 

 

IPTDAT arms men vaeru) RVARF(2) mam) I
NUUDAT RSUMY mam) IVARF(2) mam)

IPTDAT Index (pointer) into the data file
NUMDAT Number of X,Y pairs in this scan

RTIME Time this scan was taken (seconds since start)
RSUMY Sum of Y values for this scan
IVARD Code for the dependant variable

IVARF(i) Code for a selected variable
RVARF(i) Value for a selected variable

Identical to above

Pointer file format

Figure 3.6

62
Table 3.7
Summary of one record in the Pointer file

1) The record number (pointer) where the data for each

scan begins in the Data file
2) The number of data pairs (for example, mass/intensity

or energy/intensity) in the scan.
3) Elapsed time (in seconds) from experiment start to scan

start

4) The sum of the Y values (for MS, the total ion current)
for the scan

5) A code number for the dependent variable in the scan

6) Copies of three selected variables

and most importantly, three selected variables and their
values. The second reason the pointers-were placed in a
separate file was for extensibility. More scans can easily
be added without wasting any space caused by preallocation

of space within the data file.

Data file

The data file contains one or more records per scan.
The contents and order of data in this file are shown in
Figure 3.7. Each record or set of records associated with
a particular scan contains 1) the values of variable
parameters, such as probe temperature, lens voltage, etc.
and 2) the X-Y data pairs (intensity vs. dependent

variable).

The logical records in the data file are variable

length; one scan may have 10 X,Y pairs and the next may

Record Number

 

om\I m 0' # (N N

END

63

Record Contents (64 byte direct access records)

 

 

Values for variables, scan I

 

___l

X,Y data pairs. scan I

 

~777 /~’ ,«71
WéZfiv/g

 

Values for variables, scan 2

 

 

 

X,Y data pairs, scan 2 %/

 

 

L

F

C

 

 

.... .5. @ZfiW/Z

 

 

 

Data file format

Figure 3.7

64
have 1000 pairs. In this scheme, variable length logical
records have been imposed onto one or more direct access
physical records. The direct access records allow us to
jump quickly to a specific scan. One disadvantage of
imposing variable length logical records onto fixed length
physical records is that some space following the end of
the logical record is wasted. With small physical records,
this wasted space is small, and is outweighed by the
convenience of direct access to the data. Table 3.8 is a

display of one logical record in the data file.

The physical record number that is the beginning of
each logical record, or scan, is stored in the pointer
file. Each logical record consists of the values of the
variables (in the order specified in the header file), and
the X,Y data pairs. Each of these values is stored
unabridged as real numbers. Since the data are stored in a
separate file, adding new scans of data is simply a matter

of extending the file.
Creation of a data set

The multi-dimensional data base described follows the
hierarchial model. Those variables, instrument parameters
and miscellaneous data that do not change during the
iexperiment, form the foot segment of a l:N tree and are

stored in the header file. As data are acquired by the

65
Table 3.8

Dump of one logical record in the Data file

Scan number 1
Time (seconds) 0
Sum of the Y’s 0.600E+05
Variable parameters:
1 0.000 Modification History
2 1.00 Registry Identification Number
3 -l.00 Quad 1 Mass (amu)
4 0.000 Quad 3 Mass (amu)
5 lscan Scan type
6 Scan 01 Mode
7 RF Q3 Mode
8 0.000 Neutral Loss (gain) Mass (amu)
9 EI Source
10 Pos Ions
11 0.000 Quad 2 Pressure (Torr)
Ql Mass intensity 01 Mass intensity
18.100 4433.000 32.000 6607.000

28.000 19263.000

66
instrument control computers, they are stored in scan
records consisting of one record in the pointer or index
file, and one or more records in the data file. Again, the
hierarchial model is followed and those variables not
scanned are stored together with a link to the scanned data
(actually, the variables and data are stored together in
the data file as described earlier, and the link is

implicit).

Business style data bases are designed with ”data
definition languages" (60-62) which define the records and
fields available. A multi-dimensional dataset is defined
by creating the header file, specifying the static and
variable parameters. This process is usually done by the
control computers and the user is not required to learn a
”data definition language” or become a data base
administrator. This ability to create a dataset tailored
to an individual experiment allows for very fast data
storage and retrieval. when an instrument is modified and
a new parameter added, the dictionary file is simply

updated, and new datasets may be created containing the new

variable.
Conclusions
The data base described here has several unique

features: 1) It has the ability to store multi-

67
dimensional data in real time; 2) Datasets can be
automatically created and tailored to individual
experiments; and 3) The data base can be extended with a
simple addition to the dictionary file. These features set
this data base system apart from those used in business (no
data definition language, no complicated structures or
programming required). Since each dataset holds
information about only one compound or experiment, the
overhead and complexity of a complete laboratory
information management package (66,67,70,73-74) are

eliminated.

The subroutines that comprise this data base are all
capable of handling multiple datasets. This is necessary
for matching spectra and doing various "massaging"
functions on the data, such as averaging spectra. Various
programs for matching and data manipulation have been
written by Kevin Cross, of MSU. Examples of the use of
this system, and a program for extracting orthogonal planes

of data are presented in the following chapter.

The instrument data base system described here is an
efficient, extensible and modular set of routines to store
multi-dimensional data (24—25) in a hierarchial data base.
The dictionary file is an extremely powerful mechanism for
adapting this system to ever changing instrumentation.

This system has been in routine use since 1982 on three

68
triple quadrupole mass spectrometers, each with a different
set of ion optics and features. The three file formats for
the data allow a dataset to expand almost without limit.
The provisions for the rapid retrieval of data make the
system easy to use. The ease of programming and modularity
of the subroutines has been proven by the variety of

applications for which it has been adapted.

Chapter 4

Retrieval and Display of Multi-dimensional Data

Introduction

The ability to store and retrieve data is essential to
its utility. New instrumentation, capable of creating and
collecting more data than can be practically analyzed or
far more than is ultimately needed is constantly being used
to collect unwieldy amounts of difficult-to-access data.
Any new computer-controlled instrument can be told to
blindly collect and store data, and a disk quickly becomes
full of data, much of which will never encounter human
observation. However, with the right tools, one might be
able to automate a search through this sea of data to
extract trends, obtain minimum or maximum values for
parameters, or otherwise gain some appreciation for the
collected data. With the ability to automatically collect
grand amounts of data comes the need to automatically sort
and analyze them to extract the reduced set of trends or

conclusions we seek (79-80).

In triple quadrupole mass spectrometry, large amounts
of multi-dimensional data can be collected by the control
computers. This instrument is capable of collecting much

69

70
more data in one hour than the operator could possibly.
analyze in a month. This is not to say that large
quantities of data should not be taken, Just that tools to
look at these data must be utilized. If such tools are not
available, and cannot be developed, a more selective
approach to data collection is required to simplify the
analysis. The latter requires the ability to anticipate
over which range of experimental variables the needed
information will be found. Since the triple quadrupole
MS/MS technique is relatively new, the knowledge needed to

anticipate the role of the various parameters is not yet

available. In fact, copious amounts of data must be
collected in order to assess the effects of these
parameters. Since we must take large quantities of data,

and since suitable software tools for data analysis were
not available, we developed our own tools, notably a

program called EXTRACT.

The previous chapter described a multi-dimensional data
base suitable for storage of data from MS/MS experiments.
Fast storage of data into this data base is essential while
the instrument is operational; fast retrieval of data from
the data base is also a requirement. In the first case,
fast storage is required since the sample may have a
limited life. In the second case, fast retrieval of data
from the data base is required due to an operator’s limited

patience. The program we have developed, EXTRACT, provides

71
for the quick presentation of results from the dataset, as
well as several other ’user-friendly’ features. In this
chapter, I will discuss the considerations that went into
the design of EXTRACT; how it works, the user interface,
examples of use, internal configuration of the program and
possible future directions for this and other data
retrieval programs. Appendix 3 is a copy of the EXTRACT

User’s Guide for details on the operation of the program.

Data retrieval

The data retrieval methods for business and scientific

data bases have the same goal: to extract a user defined

subset of the dataset for display or a report. The query

languages for business systems often involve data
manipulation languages, query languages and report
generators (60-62). While these full report generation

systems are extremely powerful and flexible, programmers or
data base administrators are often needed to create the
templates required for even simple reports. Since the data
base needs of the business community are relatively static
(i.e. same inventory form used every week), the weeks or
months required to tailor a report generation template are

justified.

Full laboratory information management systems (LIMS)

often have full reporting capabilities (73-74), but our

72
needs dictated a much more simple interface. LIMS packages
are more akin to business systems and generate standard
sample analysis reports. A scientific data base is needed
for raw experimental data, tools are needed to view these
data from a variety of viewpoints and scientists often need
to interactively search the data base for the desired

information. The retrieval program must include provisions

for: 1) the uncertainty (error) present in our
measurements; 2) ever changing instrumentation; and 3)
simplicity of use. Simplicity of use is a key point;

scientists generally don’t want to learn a full query
language or design report templates to see the results of

their experiments.

EXTRACT - the program

EXTRACT is a general program for the retrieval of data
from a multi-dimensional dataset. This program is
completely generalized, and can be used with any instrument
that uses the multi-dimensional data base described in
Chapter 3 (25). The only element that links the data in a
multi-dimensional dataset to a specific instrument is the
correlation between a parameter code and a physical or
logical parameter of an instrument, as defined in the
instrument description (or dictionary) file. EXTRACT is
instrument independent; it makes extensive use of the

dictionary, and all displays are derived from this file.

73
In this way, the report generation displays are
automatically created from the contents of the dataset and

definitions from the dictionary.
How EXTRACT works

Data, as stored in the multi-dimensional data base, are
stored in scans (intensity vs. something) as recorded by
the instrument. If the operator wishes to see these data
in this format (i.e. one of the scans performed by the
instrument), it is a simple matter to retrieve that scan.
However, if the operator wishes data presented along an
axis that wasn’t scanned, EXTRACT must search through the
dataset extracting the data that the user wishes to see.
To do this, the operator sets the limits of the search and
instructs the program to extract data matching these

criteria.

The base rule of EXTRACT is to exclude as much of the
data in a dataset as possible, and then to present the user
with all data that remains. In this way, we are assured
that no datum will go unnoticed unless we specifically
reject if from further consideration. Specifically,

EXTRACT follows the steps outlined in Table 4.1.

74
Table 4.1

Outline of how EXTRACT functions

1) EXTRACT loops through each scan in the dataset.

2) The variables stored with this scan are
examined; if any variable is outside of the
user set limits, this scan is rejected.

3) The data in this scan are examined; if the
user requested values that don’t exist, this
scan is rejected.

4) The values the user wished displayed are
extracted and saved.

5) The next scan is examined.

By following these steps, we are certain that all the
data that fall within the user-specified limits are
extracted and saved for later display. When a datum is
retrieved from a scan (step 4 above), the minimum and
maximum values of the variables throughout the extraction
process are saved. These values are then presented to the
user after the entire dataset has been searched, and serve
to inform the user .of the status of the extraction just
performed. Say that the user extracted some data, but did
not set 'any limits for the CAD gas pressure. If the
minimum and maximum extracted values indicate a wide
pressure range, the operator may wish to set limits on the

pressure and extract the data again.

The user sets the acceptable limits for variables,
extracts the data, and the program displays the actual

minimum and maximum values for all the variables. In this

75
way, the user may interactively modify the extraction

limits while keeping an eye on the resulting data.

EXTRACT internal structure

EXTRACT consists of three main subroutines, a user
interface (EXEDIT), a graphics interface (MSPLOT) and the

extraction subroutine (XTRACT).

User interface

Various types of user interfaces were considered for
this program. There are four main types of user/machine

interfaces (see Table 4.2).

Table 4.2

User/machine interface types

1) question/answer or prompting
2) command or switch oriented
3) menu driven

4) icon driven

The first type of interface, prompting, is the simplest
to code, but the worst to use. Prompting, or question and
answer, is also the easiest for a novice computer user to
understand and use; the computer asks questions and the

user answers them. However, this type of interface is the

76
most difficult for an experienced user of the software, for
he or she too, must answer each question in turn, and this
is a time consuming process. On the other hand, a command
oriented interface doesn’t prompt the user at all. With
this style interface, the user is assumed to be an expert,
and has all the commands or switches memorized, and they
are entered as needed. For a novice, or any user that
doesn’t use the software routinely, this type of interface
requires the user to have a copy of the user’s guide beside

the terminal.

The fourth interface, icon driven, is a special
graphical form of a menu-driven system, and requires a
graphics terminal to operate the software. The menu style
interface, not requiring a graphics terminal, was selected
as appropriate for EXTRACT. As described in the theory
section above, the user must set the limits of variables,
and EXTRACT displays the actual limits found. To be a
useful, interactive program, all these limits must be
presented to the user, the user must be allowed to change
the extract limits, and see the results of these changes.

A menu format, on a video terminal, fulfills these

requirements.
The format selected for the menu is shown in
Figure 4.1. This display consists of four sections: two

lines for the header or titles, up to 20 lines for the

... a.=u..

ao<ma~m mom amiaoh ass:

77

5.38.6: 55m
.535. 3...: o... .3... sous .5055. 3...: 8. «9:5... mm
.5935. 3...: a... 38.8 8 .5935. 3...: o... m 8.... 8
.585. 3...: 9.. poor... 9 .5935. . 3...: o... . 23 No
.5935. 3...: o... m 2.3 «a .5055. 3...: o... . 2.3 .o
.5835. 3...: a... 353.0 .5055. 3...: 9.. >3
REC-3.5.; afln§un 9.: 039.3(3— HU. nsocxgv nvuiau 8v (an—0&0“
Asp-*3: .nflu'u— 0.: “SH age-SH Avaqu 06v 003
.5935. 3...:.8. 3 .3335: .5055. 3...: on. one: 8
.585. 3...: o... one: 8 .5555. 3...: o... «.3. 50m
.585. 3...: o... .3: no .5345. 3...: o... 8a: 8
.585. 3...: o... .- Sc .5555. 3...: o... 3...... .8:
.5935. 3...: o... 3.3 .9 .5355. 3...: o... > so Sm
.5055. 3...: o... a...» .5055. 43:318.. 55mm
xdl Cal XI! .3! 0—8073) yam. Cum. XII Cal Quads)
3.3... .....u 3...... 38.5.0 3...:... ....u 3...... 33.3.0

78
display of the variables, a line for the commands and a
last line for expanded descriptions. Each of the variable
displays consists of 40 character positions, including 10
characters for the variable description, and 4 numeric
fields of 6 characters each. This crowded display is
necessary to display all the variables, the user set limits

and the resulting extraction limits.

These extraction displays are generated from parameters
stored Tin the dataset. The parameter code numbers (form
the header file) are looked up in the dictionary file to
produce text for the display. All text displays (except
the two header lines) are generated this way, making
EXTRACT a fully generalized program able to retrieve data
from any dataset as described in chapter 3. iThis automatic
creation of customized report generation screens makes

EXTRACT easy to use.
Subroutine EXEDIT

EXEDIT is a screen oriented editor, allowing the user
to enter and change the limits of the extraction. EXEDIT
is a driver that keeps track of the values and their
locations on the screen, and calls subroutines to update
the screen and parse command input. The command input
'subroutine is special because it accepts single character

input from the terminal, decides if the user is starting to

79
type a command or a number, and prompts the user
accordingly. By doing this, the advanced user may reduce
the number of keystrokes required to enter extraction
limits. This input subroutine also allows the user to
enter both the minimum and maximum limits separately, or

enter them at the same time.

Upon entering the EXTRACT program, none of the
variables have extraction limits set (as in Figure 4.1);
the user must set these limits as desired. To change a
variables limit, the user moves the highlighted or active
area using the cursor keys on the terminal. Once the-
variable to change has been selected ("Scan” in
Figure 4.1), the user simply types in a new value, and that
value is inserted into the display. Since each variable
position on the screen is only six characters long, some

significant figures may not be displayed (see Table 4.3).

Table 4.3

Numeric representation for EXTRACT

Full number Truncated number
integers and floating 42 42
point numbers less than 123.4567 123.45
six characters long *63.7 —63.7
floating point numbers 1.23E+07 .123+8
greater than six digits -2.34E-17 -.2-16

The worst case truncation of the numbers is for small
negative numbers, where the number is truncated to one
digit plus the exponent. Note that EXTRACT displays
the full number on the last line of the display.

80
To alleviate this problem, the last line on the screen is
an enhanced description of the active area, including the
full. dictionary definition, and the full value, or the

value’s description, if there is one.

EXTRACT is based on the principle of retrieving a plane
of data from a multi-dimensional data base. The retrieved
plane of data can be visualized as a plot of Y vs. X.
These X and Y variables are chosen by positioning the
active area to a variable, and typing either X or Y (or Z
for three dimensional plots, see description below). The
selected X and Y variables become the axes titles for the

resultant plot.
Presentation of EXTRACT results

Once the user has entered all the limiting values for
the extraction and specified the X and Y axes, the program
is told to extract all data that are not outside of the
given limits. This extraction process can take from microf
seconds to several minutes, depending on the size of the
data base and the type of extraction desired. When the
extraction is complete, the screen is updated, showing the
user-requested limits and -the actual extraction limits
found. The actual limits will always fall within the given
limits, but may indicate a larger range than acceptable.

For example, if no limits were set for a variable, say

81
quad 2 pressure, an extraction performed, and the resulting
display showed that the actual limits of quad 2 pressure
ranged from zero to 1.0x10" torr, this would indicate that
several points were extracted at various pressures, and the
resulting plot would be meaningless. In this case, the
user should limit the pressure range, and extract the data

again.

When an extraction has yielded limits within an
acceptable range, the data may be filed and plotted (using
the graphics program MULPLT). The link to the graphics
program has been automated (subroutine MSPLOT) and can be
called up in two keystrokes. This automated link sets up a
variety of scaling factors, axes limits and other
parameters for the user, in one of four formats: point
plot, line plot, bar graph and spectrum plot. The first
three of these are options of MULPLT, but the fourth, a
spectrum plot, is a specialized bar graph that normalizes
the data to the base peak. By calling MULPLT (81) directly
from EXTRACT, either simple or publication quality
graphics, on a variety of graphics devices, can be
obtained. (MULPLT was originally written by Dr. Tom
Atkinson, but I have extensively modified it in recent
years, adding such features as color, new commands, new
device support, shading of bar graphs, speed enhancements,
a post processor for raster printers and multiple plots,

etc.)

82
EXTRACT is capable of generating pseudo three
dimensional plots by extracting a series of two dimensional
planes of data, each offset by a specific ’2’ variable
increment. These planes of data are extracted and filed,
and a post processing program (PLOT3D) is used to generate~

a series of MULPLT commands to produce pseudo three

dimensional plots. These plots are not true three
dimensional plots; they are just two dimensional plots
offset with X and Y increments. True 30 plots are

projection plots, and usually have the ’hidden’ lines
removed. They are "prettier”, but the hidden data are
inaccessible, a feature considered undesirable for most of

our applications.
Subroutine XTRACT

The subroutine XTRACT was written with speed in mind.
Since one complete pass through the entire dataset is
required for a thorough extraction of data, hooks are built
in to extract more than one plane of data in one pass
through the dataset. While the ability to extract several
planes of information in a single pass would dramatically
speed up three dimensional extractions, it would have no
effect on standard two dimensional extractions. Currently,
three dimensional plots are semiautomatically made by
EXTRACT calling the subroutine XTRACT many times. Each

time XTRACT is called, one more two dimensional plane of

83
data is added to a three dimensional .file for future
plotting. Although the multiple plane extraction feature
has- not yet been utilized, there is no reason to believe
that it will not work. To implement this feature would
require modifications to the EXTRACT main program and

possibly to the EXEDIT subroutine.
Examples of EXTRACT use

EXTRACT has proven to be a very useful program for
looking at data collected by a triple quadrupole mass
spectrometer. By using this tool, we are able to collect
large amounts of data from one sample, and present them in
a variety of ways. This allows us to easily pick out
trends in the data, and identify the effects of instrument
parameters on the collected data. By using EXTRACT, we
have been able to quickly gain new insights into the

processes going on in the TQMS instrument.

A five dimensional matrix (intensity, quad 1 mass,
quad 3 mass, collision energy and collision gas pressure)
of data is often obtained for a compound on the TQMS. An
example of such a dataset is used for Figures 4.2 and 4.3,
of the compound cyclohexyl-acetic acid. The upper portion
of Figures 4.2 and 4.3 shows the extraction limits used to
produce the plots shown in the lower portion of the figure.

In the case of Figure 4.2, a pressure plot for a specific

 

 

84

.. N .v 9:3...

sowuosmuxo ousuuomn .s> unwasous« «o uoam oflmlsxm

_ 3380.... «no co_m=_oo
Norm. .o .6th ...- m: .o

 

 

 

ri- - u b . u n h b n n n n - - - O
. \tll
[comm
No ~\E .53.. Eat. so ~\E .3395“.
2.... 0.3801335203

ml ms nu ml 0004.3 ma
. 3...: o... .5. N w 3...: o... 330m
TEN. 7.8.. 3...: 2.. 53...... 8a m m 3...... oc. . one: 8
m 3...: e... on... «e m m 3...: o... 83. Sum
Nib Nfih mfim mdw one: 8 «.8 «.mm mdm mém use: «a
8 3...: o... o as s s . 3...: o... 3.: so:
one 3...: 05 03s) >H 8” man. . 3.8: 05 > ....o lam
8 ENE. 3...:. o... a...» m... 8 8m 19.1 50m
5! xsl 5m 030....) xs... 5.. .xs... SI 030......)

0330.1 .0...“ 03!... 0095...” minnow ..wa 0.3!...— 0093..“

Kigsuazug

85

_ n.v omsm«h

so«aosmaxo humans .n> >a«asous« Mo uo«m o~mlaxm

33>. .8... N no.5

 

 

ON 0. o of: owl on... owl on!
p —y b — L n p - P - n — b — p o
m.
. ...
5.
Icons
No N\E “cocoa 50.. so u\E 5:333.
28 2.8... 3.2663
.mm mm: 3...: a... .836 8a
« . 3...: o... 28 N N 3...: o... 3.38
mg. mn3«. n.0««. 7.3m. 3000...... me n m 3...: 05 one: 8
N N 3...: a... can: «a m m 3...: o... 253 zoom
N35 so mfim m6... .3... 8 «.Nm ..8 R8 m.«m one: «0
av. 8. 3...: o... a ..«m s s 3...: 8.. 3a.: .8:
.8 8.. 3...: o... 3.3 .9 vmmmN Eh 3...: o... > 4.. ....m
E «8.. 3...: 9.. ...... D: 8. SN Infill. 50m
.3. Cu. XII Cal Qua...) XI. .3. XI .3! O—iaaxo

9:3 .3. .¢«I«J 98.3.0 00—3001 3a...“ 0.3-1.— «01.5.0

86
parent/daughter combination at a specific collision energy
has been extracted. Figure 4.3 is from the same dataset,
showing an energy plot for the same parent/daughter at a

specific pressure.

EXTRACT is capable of generating pseudo three
dimensional plots, enabling the user to see trends in
several dimensions at once. Figure 4.4 is an extraction
from the same dataset used in the above examples, but now
three dimensions of data are displayed. In this example,
if we were looking for the energy/pressure that resulted in
the largest daughter intensity, we would see that quad 2
offset needs to be about 5 volts, and the collision

pressure about 2x10'4 torr.

Conclusions

EXTRACT has proven to be a valuable tool in the
analysis of multi-dimensional data. The ability to extract
data in a plane other than that scanned gives the operator
more flexibility designing an experiment. For many
samples, collecting a full three dimensional map
(intensity, quad 1 .mass, quad 3 mass) at a variety of
collision pressures and energies allows us to determine the

optimum collision parameters.

87

\

 

\\

\ \
\\ \
\_ acoeod E9; mm N\E coacoooo
Eco osmoo.;>xmzos>o

¢.e ensues

newuoshuxo menu aoooweoolwv conga no «can oumlaxm

scammed mom co_m_=ou

 

\.\\ M“\\W\.MUV
.\\

\

 

.1 Doc:

"slug

....
.->u

(

88

EXTRACT is an instrument-independent. program; it is
capable of working with any multi-dimensional data base (as
described in the last chapter). This is due the use of an
instrument description (or dictionary) file. EXTRACT has
no instrument specific parameters coded in; instead the
dictionary file provides translation for the variables and
other parameters associated with the multi-dimensional data

base.

The generality of EXTRACT and the multi-dimensional
data base was proven by their use with simulated
electrochemical data. Simulated data was generated and
stored in the data base. EXTRACT was used to retrieve and
display these data in ways that would have been difficult

or impossible to simulate directly.

The use of the dictionary file for EXTRACT display has
the added advantage of easing report generation. With
business data bases, the field names must be known and
entered with each query (”list all employees with salary GT
35000”). The menu format for EXTRACT displays all the
parameters in a dataset, and the cursor keys provide an

easy way to select the variables.

EXTRACT provides for range limits and feedback for each
variable to minimize the impact of data uncertainty. Being

able to select the extraction limits for a variable and see

89
the actual extracted range for all variables make EXTRACT a

powerful tool for retrieving scientific data.

Chapter 5

Extracting the Information Contained in MS/MS Data
Introduction

The triple quadrupole mass spectrometer is capable of
generating large amounts of data. The analysis of these
data, in a reasonable time would be impossible without the
use of the software tools presented in the last several
chapters. These and other tools allow the operator to
examine the collected data and display them in a variety of
formats. These displays present the operator with the raw
data as they were collected (or possibly normalized), and
allow the operator to interpret the results. The next
logical. step in computer-assisted problem solving is to
automate the extraction of the information present in the

data.

There are two predominant ways that a computer can help
interpret data from a mass spectrometer: expert systems and
pattern recognition. Artificial intelligence (AI) is a
field of computer science dealing, generally speaking, with
making computers "think”. A subfield of AI is the study of
expert systems. An expert system (82-84) is a computer
program that mimics a human expert in a specific field of
expertise. This is currently possible only for small, well

90

91
defined problems for which there exists a well-defined
”knowledge base" of information on which the program bases
its decisions. Essentially, these programs apply a set of
"rules" to the data, presumably the same rules that human

experts use, to extract the information present in the

data.

Several researchers have applied expert systems (85-93)
to mass spectrometry. The DENDRAL project, from Stanford
University (21,85-93) has shown that expert systems could
be used to help interpret mass spectra. The original
DENDRAL algorithm, developed by J. Lederberg, was able to
identify all possible acyclic molecular structures given a
set of constituent atoms. Heuristic DENDRAL achieved the
same objective in less time by using mass spectrometric
data and rules to infer constraints on the structure gen-l
eration. Meta-DENDRAL was then'developed to automatically
generate the rules for Heuristic DENDRAL. CONGEN was later
developed to replace the older DENDRAL algorithm, and was
able to generate cyclic structures. DENDRAL has three
functional units, Plan, Generate and Test. The planning

phase uses rules to constrain the generate phase (using

CONGEN). The test phase then uses another set of rules to
”fragment" the generated structure and compare the
resulting mass spectrum with the unknown. In this way, the

generated structures could be ranked, and top ranked

structures are often indicative of the unknown’s structure.

92

One program from the DENDRAL project, GENOA (22-23), is
a program to generate all chemically possible structures
from the molecular formula, constrained by any additional
chemical information presented to it. We are currently
using this program in our laboratory as one part of our
structure elucidation scheme (see Figure 1.1). Another
expert system applied to mass spectrometry is the system
developed at. Lawrence Livermore National Laboratory
(95-98). Currently this system is capable of tuning a
triple quadrupole mass spectrometer based on both signal

intensity and peak shape.

Pattern recognition is a field of study based on the
assumption that data can be clustered or grouped into
distinct sets (50-54,85-87). Each of these sets is
presumed to have one or more characteristics that
distinguish it from other sets. A subset of this field is
routinely used in mass spectrometry: spectrum matching
(85-87). Any MS data system capable of searching a library
for reference spectra that "match" an unknown is performing
a simple pattern recognition task. The criteria for
separating normal EI mass spectra into groups, with top

ranked matches, have been studied extensively (40-49).

In this chapter, I will present two tools that help a
chemist extract a small portion of the information present

in MS/MS data. MS/MS data are different (i.e. more

93
specific) than "normal" MS data, and .new rules for
information extraction are required. The first tool
presented is a knowledge-based program to aid in the
analysis of neutral losses from parent ions. The second
tool is an example of a pattern recognition technique, and

is used to match daughter (or neutral) spectra.
Information contained in MS/MS data

Mass spectrometers have traditionally been used to aid
chemists in determining the structure of unknowns. The
information contained in a typical mass spectrum comes from
several basic sources, including 1) the absolute mass to
charge ratio of an ion; 2) the relative m/z value relative
to another ion in the spectrum; 3) the intensity of an ion
peak relative to the base peak intensity, total intensity
or another ion’s intensity; and 4) the absence of ions at
certain m/z values. The absolute mass of a peak shows the
presence of a relatively stable charged species, giving an
indication of it’s chemical makeup. The relative masses of
two or more peaks gives the mass(es) of non-charged species
that may have been lost from the higher mass ion. The
relative intensities of the ion peaks in a spectrum yield
information about which fragmentation pathways are most
predominant. The data present in a mass spectrum are
interpreted by analyzing these absolute and relative masses

and intensities according to'well-defined rules (99).

94

A set of all daughter spectra from a sample contains
all the information present in an El spectrum from the same
sample. Each daughter spectrum has the advantage of being
simpler than the E1 spectrum of the molecule, but the
concatenation, or overlapping, of all the daughter spectra
produces a spectrum similar to the E1 spectrum. The
relative simplicity of a daughter spectrum is due to the
selective nature of the first mass filter and the less
severe nature of the second collision. Each daughter ion
contributes evidence of the composition of its parent ion,
and can be used to help interpret the structure of the
parent. Due to these differences, new information.
extraction rules or algorithms are needed to analyze MS/MS

data.

Neutral spectrum

A neutral spectrum is similar to and derived from a

daughter spectrum. A neutral spectrum presents the

relative amounts and masses of neutral fragments lost in

the fragmentation of a parent. These scans are easily
derived from a normal daughter spectrum by simply
subtracting the daughter masses from the parent mass. The

resultant spectrum contains the same information present in
a daughter spectrum (less the parent ion, now at mass 0),

but is presented in a different form. Figure 5.1 shows

intensity

intensity

95

2—methyl-4—pheny|—2-butonol
daughter spectrum of m/z 147

 

 

 

 

 

 

 

 

1oo -
50 -4
.4
O l
I r I r T! I r I l I I I1 rfi I I r T I I I I I rj I [j 1
o 20 40 so so 100 120 140
' m/z
neutral spectrum of m/z 147
100 -'
so -
° rTrrTrTT'er'r'Trrrr'r'rrrfl
o 20 40 so no 100 120 140

m/z

Comparison of daughter and neutral spectra

Figure 5.1

96
both daughter and neutral spectra for parent m/z 147 from

2—methyl-4-phenyl-2-butanol.

One of the techniques for identifying an unknown from
its EI mass spectrum is to characterize the neutral losses
from ions in the spectrum. The characterization of these
losses is complicated by the richness of an El spectrum, in
that it is difficult to tell exactly which ions are formed
from neutral losses from the molecular or any other ions.
Losses are most easily observed from the molecular ion,
looking backwards down the spectrum. ‘ If the next ion
present is 28 daltons less, the parent lost either CO or
Czflc. However, as we proceed down the spectrum, it becomes
unclear where a specific loss occurred from. This limits
the usefulness of this technique to confirming postulated

structures.

In tandem mass spectrometry, we can determine exactly
which daughter ions are formed from which parent ions. If
we set up the collision conditions (CAD gas pressure, axial
energy) so as to ensure only a single collision for the
parent ion and little or no chance that the resultant
daughters will further fragment, we know that all the
daughter ions formed are direct, single-event fragmentation
products of the parent. If the collision gas pressure were
too high, some of the daughter ions formed would again

collide with the target gas. These second and higher order

97
fragmentation processes give the same types of uncertainty
in a daughter spectrum as in an E1 spectrum. Therefore, if
we keep the collision conditions conservative (CAD gas
pressure low, medium energy), we produce spectra of first-
order collision products. These spectra are not as rich as
spectra produced at higher pressures since only first-order
products are generated. However, first-order daughter and
neutral spectra give direct, definitive information about

the composition of parent ions.

Daughter spectra and neutral spectra present
essentially the same information, but this information can
be used in a complementary fashion. Daughter spectra
present those species that retain the charge during
fragmentation, while neutral spectra present the uncharged
fragments. Deriving neutral spectra gives spectra of a
form amenable to the same matching techniques used for ion
spectra. If the spectra were left in daughter ion form,
different pattern recognition techniques (such as sliding
correlations) would have to be used to group spectra by
neutral loss information. By analyzing both of these
spectra, we are able to deduce some information about the

parent ion’s composition and structure.

98

A simple expert system for neutral loss analysis

The low mass neutral fragments are simple structures,
generally corresponding to simple, common neutral losses.
I have written a program (ANEUT) to aid in the
determination of these losses. ”Expert knowledge" may be

encoded in a knowledge-based system, such as an expert

system, in two primary ways: as rules and as base of
knowledge upon which rules can draw information
(85-87,90-93). The knowledge base for this program was

derived from a table of possible losses from appendix A.5
in McLafferty’s Interpretation of flags S ectra, and was
enhanced as we gained experience with the program. The
conclusions drawn from this deductive system are simply a
list of possible neutral losses - it is the chemist’s
responsibility to utilize the resulting information and

update the knowledge base as required.
ANEUT knowledge base and rules

The knowledge base is a text file which allows for the
easy extension of the knowledge available to the program.
This file consists of one entry for each possible loss, or
series of losses. Figure 5.2 shows several entries.from
the current file, and the format of each entry. For series
'of losses, such as CnH2n+1, an alkyl loss, limits are

placed on the values of n (for this case, n is greater than

99

«.m mousse

9=m2< new moon ouvouzosx one loam nomsaom

 

 

 

 

 

mEmEEoo

\\ll. :JILHHHHHUIII ||t|t\\\/lllr rtl|HHHHtlltis 1!!!!
2282

35:2

38 s. 233 stops

.8. .32

+.o..no!.~_ £2.38. 2.93 .muczanoo 523 33:82 :38

 

: so.

2352 emcee
.3/1. .\/4.:.\.\-- ,I. \.7.>.\/
o «:5 .o

m: 2.0 .o

o o A_+=sz co.m .o
A_+=~V= =0.o~._
Co co.m ._

100
0). These limits, the formula, and a comment comprise one
record. The comment describes the conditions or

conformation that lead to a specified loss.

There are currently only three ”rules" for the program
ANEUT. One rule is used to match the actual losses (found
in the spectrum) with the possible losses (generated from
the knowledge base). Another rule is used to allow higher
mass neutrals to be a sum of lower neutrals, and the final
rule adds a "quality" to a possible loss. The primary rule
is simple: if the mass of an actual loss is within 10.3
daltons of the mass calculated from the knowledge base,
that loss/knowledge base entry correlation is retained.
The mass deviation of 10.3 allows for instrumental effects
(the instrument may be slightly off calibration, etc).
This rule searches the entire knowledge base generating a
list of correlations. It should be noted that this rule
does not guarantee a complete and exhaustive check of all
possible neutral losses, instead it relies on the knowledge
base to contain a fairly complete list of the common

neutral losses.

The second rule allows several of the neutral losses to
occur, and effectively adds their masses together. The
third rule adds a ”quality" factor to certain of the
correlations made by the first rule. If the unknown

daughter or neutral spectrum contains a series of losses

101
(such as 15, 29, etc. from CnH2n+1), each of these entries
is flagged as being the likely loss. If just one of the
elements in a series is present, it is not flagged: This
is based on the assumption that if a series of losses is

possible, the fragmentation process will generate the

entire series.

Example of ANEUT use

Examples of ANEUT’s use are seen in Figures 5.3
and 5.4. Figure 5.3 shows the neutral spectrum from parent
m/z 108 (the molecular ion) of 1,4-benzenediamine and the
output from ANEUT. As can be seen from this list of
possible neutral losses, the peaks at l and 27 daltons are
well accounted for (H and HCN), while the peak at 28
daltons is the loss of both H and HCN. Figure 5.4 shows
the neutral spectrum from parent m/z >147 (probably
HOOC(CH2)4COOH2*) from bis—2-ethylhexyl adipate and the
results from ANEUT. After ignoring the Cl, N and S
containing structures, we are left with fragments

indicative of acids and alcohols.

The comments used for ANEUT’s output are directly from
McLafferty’s table which was derived from normal EI
spectra. These comments represent the possible
conformation of the original molecule and take into account

the extensive rearrangements possible in the source. As a

102

m.m mosses

ucwlowoososconlv.~ mo mwahaoso moo lsauoomm accuse:

Ow.

 

 

 

 

«Amuse to Amuse so noes saw-uuaateaut owesoos. .enuuoosaua «a so «.mm
pa + H cusses mo cowuscwnlou m.mm
+oru0I|z .nucouox awashu .svssoAlou souaxo ewes-os< o o «.mu
.2"... 0:26 .8..qu .dsrzuzuz?‘ «2 New
A=I=V+A:~I.mv so Ammlmv mo wood use-oucssseus uwmwouom .moooa n: we m.~v
nonsense .nvcsoamou omdo>uos0uoz souosuwz z o a m.Nv
. «oeseoesa can noessaose pesos .uoeseaso ssxsm u: masses m e.mm
sumo-loo o~sssoa omen
~\:c
on ow o.
—[PF-Pht-b|—- p—Pbbp-IPP-hnnrph- P-r C

a.

e
I on
I oo.

 

Aco_ 3300.95 no. ~\E acucoa

mc_Eo_UmcmNcmml¢._

U1

\
O

{usua

mm
mm
mm
an

hm
am

so:

103

e.m unsung

oaomwos amxomtfizsaulwlnma mo mwmhnmsm moo assuuumm Hmsusuz

 

monomauhs< .mNomm «O m “Um ham
53830850.. .3381:— = o as 0 m 0 mm. 0 5n :28
a canned new: mooo.m sate “mosuo.e none ace mo.aaos a o as no m.m m.me
sumac no use“ .55qu new: no: he 003 N: o n o m: o m.m one
00003035305 aloe .eocfifiss «and. Tamas: m m: 0 m6 m.mv
whousoouuwc .suwumIOHeouuwz. No 2 m.m m.mv
Iomuuol 03050 5 333 not; :89: lo...“ :8 + 00 mo 003 a o m o o m.m mdv
3 + S cusses .«o 5325430 En cdm
noun" 8 a Tm odn
maonooamhxosuu! .waoao0~m«n a o m m o : H.m +o.mm
z 0:43 :33 moo?"— moum £085.". some now mo :3 m o m mdm and
, 3230: o as now 93
ease-loo ads-Lou Oahu use:
N\c. .
on on on on at on ON Op 0
Fun-b—nnnn-pnnan-nb-muhn-Prbn—nbhn-mb-b-nmpb—Pan—bh-n-h-mb-n mn-bbbn-m-nb-nnbb
N M“
I on m.
r S.
r ,M
m
I oo—

A+NIOOU¢ANIQVUOOIV C; u\E “cocoa

 

Bogus Exogiésmimrma

 

we
we
me
mo
9.

mm
mm
mm

ma
m~

loz

104
result, these comments may only be useful for analyzing
neutral species lost during the fragmentation of the

molecular ion.

Future extensions to this system include constraining
the matching process to include or exclude certain atoms.
For example, if we know that the compound doesn’t include
nitrogen, several of correlations will not be displayed,
shrinking the list of possible losses. Alternatively, we
may know that the unknown contains only carbon, hydrogen

and oxygen, effectively eliminating all other elements.

This same program, used with a different knowledge
base, could be used to aid in the identification of peaks
in a daughter spectrum. This program or similar expert
systems can be (and will be) used in an iterative manner
with others, including GENOA. The more information that an
expert system is given about an unknown, the more certain
its analysis of the unknown. For example, output from
ANEUT can be used as constraint input into GENOA, limiting

the number of structures generated.
Data groupings
The data content of a daughter spectrum is not as great

as that in an El spectrum; however, the information content

may be greater if the conclusions are more certain. The

105
sum of the information contained in all the daughter
spectra for one compound could then greatly exceed the'
information available from the E1 spectrum. Many
information extracting techniques have been applied to E1
spectra to deconvolute them into simple patterns (52-54).
Some of these pattern recognition techniques have not fared
well, being overwhelmed by the overlapping data present.
These same techniques applied to the daughter spectra from
a compound don’t reveal any new information, because the
daughter spectra are already well grouped subsets of the E1

spectrum.

A daughter spectrum of the molecular ion bears a great
similarity to the E1 spectrum, however the daughter spectra
from other than the molecular ion are relatively simple.
The normal EI spectrum contains fragments from all
substructures in the compound, while a daughter spectrum
contains fragments from one (or a small number) of
substructures of the molecule. If we are able to group
sets of these daughter spectra from different compounds, we
expect, and find, that the closest matching spectra are

derived from compounds with similar substructural features.

MS versus MS/MS spectra matching

The traditional spectrum matching programs designed for

matching EI mass spectra take one of several general

106
approaches to this problem of grouping spectra. Each of
these methods relies on a library of spectra which may be
abridged in a variety of ways. Biemann searching
techniques abridge the spectra to the two most abundant
peaks in a 14 dalton window (42). Other techniques reduce
the intensity to a binary value, i.e. either there is a
peak at a mass or not (40-41). Kevin Cross of MSU has
implemented a matching system for E1 spectra that is based
on the unabridged spectra (20,47). This matching program,
patterned after a variety of other matching techniques,
assigns weights to the masses and the intensities of each

peak, and produces good results for El spectra.

The simplicity and wide variations in intensities in
daughter spectra make normal EI matching techniques less
suitable for daughter spectra. Better daughter spectra
groupings (and) hence substructural feature groups) can be
achieved by doing minimal intensity screening for peaks at
similar m/z values. I have studied a variety of intensity
matching algorithms to determine the importance of the
relative intensities of daughters at the same mass. The
problem becomes more complicated when you must match a
daughter spectrum from a weak parent ion to a spectrum from
an intense parent ion. Simply normalizing the spectrum to
the parent peak or to the total ion current may result in
the noise peaks becoming too prominent. The threshold

level for recording a peak on the spectrometer may have

107
been set too high, and low intensity daughter ions may not
be seen. These and other problems require special

attention when matching daughter spectra.
Intensity matching for daughter spectra

The approach I took to determining the importance of
daughter ion' intensity was to try several intensity
compression algorithms on daughter spectra from the same
parent structure taken under different conditions. Also
included in the reference library were vastly different
compounds with the same parent ion masses. The first
attempt ignored intensities altogether, ire. a binary
compression. If any peak existed at the same mass in both
the unknown and reference spectra, it scored one point. On
this basis, the top scoring matches proved to be an
unselective sample of the data base. This was primarily
due to scoring ’matches’ of intense daughter peaks with
very small peaks, and neglecting to account for missing
peaks. This indicated the need for some intensity

weighing, and several methods were tried.

Reducing the intensity range from six orders of
magnitude available from the instrument to a number
representative of the magnitude (i.e. LOGIO) gave a total
dynamic range of zero to six. Even with this dramatic

reduction of scale, and counting intensities that matched

108
within :1 unit, the resultant groupings of spectra were not
as good as desired. The intensity differences, from
similar spectra, were just too great for this 6 level

approach.

The current grouping algorithm uses only three levels
of intensity information: strong, medium and weak. First,
any peaks that are "weak" (less than an arbitrary 800
counts) are marked as being weak. These small peaks must
be classified as weak before normalizing the spectrum to
identify them as peaks not much greater than the instrument
background noise. Next, the daughter spectrum is
normalized to the parent ion, and those peaks that are less
than one percent are marked weak, between 1 and 10 percent
are medium, and greater than 10 percent are strong. Two
peaks are considered to have matched intensities if they
are in the same or adjacent groups. ' This algorithm
produces good results, allowing for wide intensity

variations present in daughter spectra.
Ranking and sorting the data groups

The last step in a grouping program is to have the
computer rank the results (Figure 5.5). This involves
assigning ‘quality factors to matching peaks, and deducting
points for mis-matched spectra. Peaks in an unknown that

intensity match peaks in the reference are given full

109

Intensities reduced to: Weak 800 counts or < 12)

 

 

( <
Medium ( 1* g intensity 5 10%)
Strong (10% g intensity 5 100%)
mass comparison intensity comparison weighting factor
ref = unknown intensity matches + 1.0 U
(a mass in both) (within i one group)
ref =.unknown intensity doesn’t + 0.5 U
(a mass in both) match
mass in unknown weak intensity : 0.0
not in reference (ignored)
mass in unknown medium, strong - 0.33 U
not in reference intensity
mass in reference weak intensity - 0.05 R
not in unknown
mass in reference medium, strong - 0.1 R

not in unknown intensity

The weighting factor is based on the number of peaks in
either the reference or unknown spectra. R and U each
represent the percent of the total spectrum (based on the
number of peaks, not their intensities) in the reference
and unknown spectra.

Characteristics used to match daughter spectra

Figure 5.5

110
credit, while those peaks at the same mass but different
intensities gain only one half credit. Each "credit" is
the percentage of the number of peaks in the unknown. For
example, if an unknown has 5 peaks, and 4 peaks match in
mass and intensity and one matches mass but not intensity,
the rating would be 90% (4*100/5 + 1*50/5). Medium and
strong peaks in the unknown that are not in the reference
spectrum are negative 1/3 credits. Weak peaks from the
unknown, not in the reference, are given a zero credit
(i.e. they are ignored). The last of the ranking credits
are based on reverse searching (comparing the reference to
the unknown). These negative points are based on the
number of peaks in the reference spectrum (here one credit
is a percentage of the number of peaks in the reference
spectrum). Unmatched strong and medium peaks in the
reference are a -lOX credit, while weak peaks not in the

unknown spectrum deduct 5* of a credit.
Examples of daughter spectra grouping

An example of this grouping algorithm is shown for a
group of 21 compounds, each containing a parent at m/z 149.
In both Figures 5.6 and 5.7, the phthalate ion
(CsH4(CO)2OH*) of di-ethyl-phthalate is matched against all
the daughter spectra from these 21 compounds (which include
other phthalates and a variety of other compounds). The

grouping. technique correctly clusters the spectra from the

111

Matching scan 6 (0149.0) from dataset R15444 (diethyl phthalate)

Group factor Scan

100 6
97 10
87 10
66 13
64 10
63 7
58 4
58 5
55 7
54 13
53 13
50 2
49 5
35 6
34 2

Parent

(0149.
(0149.
(D149.
(0120.
(0149.
(0121.
(0149.
(0149.
(0121.
(0121.
(0120.
(0164.
(0177.
(0 93.
(0164.

0)
0)
1)
8)
0)
0)
0)
0)
1)
0)
6)
0)
1)
0)
0)

Dataset

R15444 (diethyl phthalate)

R27399 (dioctyl phthalate)

R20688 (dibutyl phthalate)

R27399 (dioctyl phthalate)

R22855 (dipentyl phthalate)

88248 (p-t-butylbenzyl alcohol)
R8239 (2-t-butyl-6-methyl phenol)
R12776 (lO-undecenoic acid, methyl ester)
R15444 (diethyl phthalate) .

R20688 (dibutyl phthalate)

R22855 (dipentyl phthalate)

R8248 (t-butylbenzyl alcohol)
R15444 (diethyl phthalate)

R8416 (1,3 benzenedicarboxylic acid)
R8239 (2-t-buty1-6-methyl phenol)

Example of the daughter spectrum matching algorithm

Figure 5.6

112

Matching scan 6 (0149.0) from dataset R15444 (diethyl phthalate)

PT PC NC NS NR IS
100 100 4 0 0 0
68 55 3 l 0 61
67 51 2 2 0 25
66 59 3 l 1 61
59 58 3 1 2 61
58 60 3 '1 2 61
57 55 3 1 2 61
56 61 3 l 3 8
53 52 2 2 2 25
50 55 4 0 l 0
49 40 2 2 1 70
49 32 2 2 0 75
46 59 3 1 3 8
45 33 2 2 l 75
44 34 2 2 1 75
*1

s2
*3

351555313

IR Scan Parent

no hi
caaucoca-scncncococococ:

NH
030!

14

6
7
3
13
13
13
4
7
2
10
10
6
5
16
9

(0149.0)
(0121.1)
(0149.0)
(0121.0)
(0120.6)
(0120.8)
(0121.0)
(0149.0)
(0164.0)
(0149.1)
(0121.0)
(0 93.0)
(0177.1)
(0 92.8)
(0 93.2)

1,3 benzenedicarboxylic acid
tetra ethyl silicic acid
2-t-butyl-6-methyl phenol

overall match factor

Dataset

R15444
R15444
R8416

R2060!
R22855
R27399
88416

R13923
R8239

R20688
Rl3923
R8416

R15444
R22855
R15444

(diethyl phthalate)

(diethyl phthalate)

t

(dibutyl phthalate)

(dipentyl phthalate)
(dioctyl phthalate)

*1

*2

x3

(dibutyl phthalate)

t1

(diethyl phthalate)
(dipentyl phthalate)
(diethyl phthalate)

pattern correspondence - intensity based fit
‘ 3 common

number of peaks
number of peaks

the sample (unknown) not matched to ref.

number of peaks ;n the reference not in the unknown
percent total ion current of unmatched sample (unknown)
percent total ion current of unmatched reference

Example of the EI spectrum matching algorithm

Figure 5.7

113
phthalates, and other samples are distributed lower on the
scale (Figure 5.6). Figure 5.7 is a comparison of the
match factors produced from the EI matching technique. As
can be seen from these figures, the EI matching algorithm
did not successfully group the phthalates together, while
an algorithm designed for matching daughter spectra

performed well.

Conclusions

These two techniques extract only a small fragment of
the information present in MS/MS data. The neutral-
spectral studies have shown that the first-order losses
from parent ions give direct, definitive information about
the composition of the parent ion. The expert system

presented in this chapter is a simple program that shows

the promise of this technique. The grouping method
presented here demonstrates the differences between
daughter/neutral spectra and EI spectra. In order to

effectively group daughter or neutral spectra, we need to

use algorithms with minimal intensity screening.

The technique of MS/MS is a powerful tool for structure
determination. The tools presented in this thesis are an
introduction to the kind of tools, rules and algorithms
needed to elucidate the structure of unknowns. The

combined research efforts of many members of the Enke group

114
have and will continue to lead toward the development and
implementation of a variety of tools and techniques useful

in determining the structure of unknown samples.

B IBLIOGRAPHY

10

11

12

13
14

115

BIBLIOGRAPHY

Yost, R.A., Enke, C.G., J. Am. Chem. Soc., 1 0, 2274
(1978).

 

 

Yost, R.A., Enke, C.G., Anal. Chem., 5;, 1251A (1979).

Yost, R.A., Ph.D. Dissertation, Michigan State
University, E. Lansing, MI (1979).

McLafferty, F.W. in ”Tandem Mass Spectrometry", F.W.
McLafferty, Ed., John Wiley & Sons, New York, NY,
1983, Chapter 1.

Levsen, E., Beckey, H.D., Org. Mass Spectrom., g, 570
(1974).

Bozorgzadeh, M.H., Morgan, R.P., Benyon, J.H.,
Analyst, 103, 613 (1978). .

Yost, R.A., Enke, C.G., in "Tandem Mass Spectrometry",
F.W. McLafferty, Ed., John Wiley & Sons, New York, NY,
1983, Chapter 8.

Yost, R.A., Fetterolf, 0.0., Hass, J.H., Harvan, 0.J.,
Weston, A.F., Skotnick, P.A., Simon, N.M., Anal.
Chem., 56, 2223 (1984).

Maquestian, A., et. al., in "Tandem Mass
Spectrometry”, F.W. McLafferty, Ed., John Wiley &
Sons, New York, NY, 1983, Chapters 21-26.

Bauer, M.H., Masters Thesis, Michigan State
University, Lansing, MI (1983).

E.
Myerholtz, C.A., Ph.D. Dissertation, Michigan State
University, E. Lansing, MI (1983).

Yost, R.A., Enke, C.G., Org. Mass Spectrom., 16, 171
(1981).

Yost, R.A., Enke, C.G., American Lab, June 1981.
Levsen, K. in ”Tandem Mass Spectrometry", F.W.

McLafferty, Ed., John Wiley & Sons, New York, NY,
1983, Chapter 3.

15

16

17

18

19

20

21

22

23

24

25

26

27

28

29

116

Giordani, A.B., Gregg, H.R., Hoffman, P.A., Cross,
R.P., Beckner, C.F., Enke, C.G., presented at the 32nd
Annual Conference on Mass Spectrometry and Allied
Topics; San Antonio, TX, May 27—June l, 1984.

Cross, R.P., Palmer, P.T., Giordani, A.B., Beckner,
C.F., Hoffman, P.A., Gregg, H.R., Enke, C.G., ACS
Symposium Series, in press.

Hoffman, P.A., Enke, C.G., presented at the 3lst
Annual Conference on Mass Spectrometry and Allied
Topics; Boston, MA, May 8-13, 1983.

Hoffman, P.A., Ph.D. Dissertation, Michigan State
University, E. Lansing, MI, (in preparation).

Cross, R.P., Enke, C.G., presented at the 32nd Annual
Conference on Mass Spectrometry and Allied Topics; San
Antonio, TX, May 27-June l, 1984.

Cross, R.P., Ph.D. Dissertation, Michigan State
University, E. Lansing, MI (1985).

Lindsay, R.K., et. a1., "Applications of Artificial
Intelligence to Organic Chemistry: The Dendral
Project", McGraw Hill, New York, NY (1980).

Carhart, R.E., Smith, D.H., Gray, N.A.B, Nourse, J.G.,
Djerassi, C., l. Org. Chem., 15, 1708 (1981).

Carhart, R.E., Varkony, T.H., Smith, D.H., ACS
Symposium Series 54, 126 (1977). -

Crawford, R.W., Brand, H.R., Wong, C.W., Gregg, H.R.,
Hoffman, P.A., Enke, C.G., presented at the 30th
Annual Conference on Mass Spectrometry and Allied
Topics; Honolulu, HI, June 6-11, 1982.

Crawford, R.W., Brand, H.R., Wong, C.W., Gregg, H.R.,
Hoffman, P.A., Enke, C.G., Anal. Chem., 56, 1121
(1984).

Newcome, B.H., Ph.D. Dissertation, Michigan State
University, E. Lansing, MI, (1984).

Hoffman, P.A., Private Communication.

Newcome, B.H., Enke, C.G., Rev. Sci. Instr., 55, 2017
(1984).

Brodie, L., "Starting FORTH, An Introduction to the
FORTH Language and Operating System for Beginners and
Professionals", Prentice—Hall, Englewood Cliffs, NJ
(1981).

 

 

30
31

32

33

34

35

36

37

38

39

40

41

42

43

44

45

46

117
Denton, M.B., Private Communication.

Tilden, S.B., Denton, M.B., Technical Report #18, for
ONE contract #N00014-75-C-05l3 (1979).

Carlson, E., Ph.D. Dissertation, Michigan State
University, E. Lansing, MI (1978).

Gregg, H.R., Chai, J.W., Chakel, J.A., Hoffman, P.A.,
Latven, R.E., Matthews, R.S., Myerholtz, C.A.,
Newcome, B.H., Enke, C.G., presented at the 29th
Annual Conference on Mass Spectrometry and Allied
Topics; Minneapolis, MN, May 24-29, 1981.

Schubert, A.J., Ph.D. Dissertation, Michigan State
University, E. Lansing, MI (in preparation).

Myerholtz, C.A., Newcome, B.H., Enke, C.G., presented
at the 31st Annual Conference on Mass Spectrometry and
Allied Topics; Boston, MA, May 8-13, 1983.

Kristo, M.J., Enke, C.G., presented at Rochester FORTH.
Conference, June 1985.

Kristo, M.J., Myerholtz, C.A., Schubert, A.J., Enke,
C.G., presented at Rochester FORTH Conference, June
1985.

Myerholtz, C.A., Schubert, A.J., Kristo, M.J., Enke,
C.G., accepted for publication in Instruments and
Chemistry.

Myerholtz, C.A., Schubert, A.J., Kristo, M.J.,‘Enke,
C.G., accepted for publication in Instruments and
Chemistry.

Grotch,.S.L., Anal. Chem., 42, 1214 (1970)

Wangen, L.E., Woodward, W.S., Isenhour, T.L., Anal.
Chem., 35, 1605 (1971).

Hertz, H.S., Hites, R.A., Biemann, K., Anal. Chem.,
11. 681 (1971).

Naegli, P.R., Clerc, J.T., Anal. Chem., 6 739A
(1974).

 

Gronneberg, T.O., Gray, N.A.B., Eglinton, 0., Anal.

,Chem., 31, 415 (1975).

Pesyna, G.M., Venkataraghavan, R., Dayringer, R.E.,
McLafferty, F.W., Anal. Chem., 15, 1362 (1976).

Blaisdel, B.E., Anal. Chem., 55, 180 (1977).

 

47

48

49

50

51

52

53

54

55

56

57

58

59

60

61

62

63

118

Damen, H., Henneberg, 0., Wiemann, B., Anal. Chim.
Acta, 103, 289 (1978).

 

 

Van Marlen, G., Van Den Hende, J.H., Anal. Chim. Acta,
112, 143 (1979).

 

 

Lebedev, H.S., Tormyshev, V.M., Derendyaev, B.G.,
Koptyug, V.A., Anal. Chim. Acta, 133, 517 (1981).

Jurs, P.C., Kowalski, B.R., Isenhour, T.L., Anal.
Chem., 5;, 21 (1969).

 

Jurs, P.C., Kowalski, B.R., Isenhour, T.L., Rielley,
C.N., Anal. Chem., 3;, 1949 (1969).

Bender, C.F., Shepherd, H.0., Kowalski, B.R., Anal.
Chem., 35J 617 (1973).

 

Lam, T.F., Wilkins, C.L., Brunner, T.H., Saltberg,
L.J., Kaberline, S.L., Anal. Chem., 35, 1768 (1976).

Hitter, C.L., Isenhour, T.L., Computers Chem., A, 243
(1977).

Lowry, S.B., Isenhour, T.L., Justice, J.H.,
McLafferty, F.W., Dayringer, R.E., Venkataraghavan,
R., Anal. Chem., 55, 1720 (1977).

Buchs, A., Duffield, A.M., Schroll, G., Djerassi, C.,
Delfino, A.B., Buchanan, B.G., Sutherland, C.L.,

Feigenbaum, E.A., Lederberg, J., 5. Ag. Chem. Soc.,
5;, 6831 (1970).

Delfino, A.B., Buchs, A.B., Helv. Chip. Acta, 55, 2017
(1972).

Buchanan, B.G., Smith, D.H., White, W.C., Gritter,
R.J., Feigenbaum, E.A., Lederberg, J, Djerassi, C., J.
5;. Chem. Soc., 55, 6168 (1976).

McLafferty,F.W., Anal. Chem., 55, 1441 (1977).

Cardenas, A.F., Data Base Management §ystepg, 2nd ed.,
Allyn and Bacon, Inc., Boston, MA (1985).

Gillenson, M.L. Database, Wiley-Interscience, New
York, NY (1985).

Wiederhold, 0., Database .Design, McGraw-Hill, New
York, NY (1977).

Borman, S.A. Anal. Chap, 51, 983A (1985).

 

64

65

66

67

68

69

70

71

72

73

74

75
76

77

78
79

119

Hampel, V.E. in Data for Science and Technology,
Glaesen, P.S. (ed.), Pergamon Press, Oxford, (1981).

Chen, C.C. and Harnon, P. (ed.) Numeric Databases,
Ablex Publishing, Norwood, NJ (1984).

Glaeser, P.S. (ed.) Data for Science and Technology,
Elsevier Science Publishers, Amsterdam, The
Netherlands (1984). -

Glaeser, P.S. (ed.) Data for Science and Technology,
North-Holland Publishing, Amaterdam, The Netherlands
(1983).

Shoshani, A., Olken, F. and Wong, H., pg. 349 in The

Role 5: Data 15 §cientific Programs, Glaeser, P.S.
(ed.) Amsterdam, The Netherlands (1985).

Sobel, y., Dagane, I., Carabedian, M., and Dubois, J.,
pg. 395 in The Role 5: Data 15 Scientific Programs,
Glaeser, P.S. (ed.) Amsterdam, The Netherlands (1985).

Smith, F.J. and Hughes, J.G., pg. 435 in The Role 5:
Data 55 Scieptific Programs, Glaeser, P.S. (ed.)
Amsterdam, The Netherlands (1985).

 

Rumlble, J.R. Jr. and Hampel, V.E. (eds.),Database
Management ip_Scienca and Technology, Elsevier Science
Publishers Amsterdam, The Netherlands (1984).

Rumlble, J.R. Jr. in Database Management 15 Science
555 Technology, Rumlble, J.R. Jr. and Hampel, V.E.
(eds.), Elsevier Science Publishers Amsterdam, The
Netherlands (1984).

Kipiniak, W. and Finnerty, W. pg 17 in Computers 55
the Laboratory, Liscouuski, J.G. (ed.), ACS Symp.
Series 265 (1984).

Baumann, F., Lewis, R.A., and Brown, A.C. III, pg 23
in Computers 55 the Labopgtory, Liscouuski, J.G.
(ed.), ACS Symp. Series 265 (1984).

 

RS-l, BBN Software Products.
Lotus 1-2-3, Lotus Development Corp.

Gregg, H.R., Enke, C.G., in preparation for submission
to Computers and Chemistry.

VAX FORTRAN, Digital Equipment Corp.

Enke, C.G., Science, 215, 785 (1982).

 

 

80

81

82

83

84

85

86

87

88

89

90
91

92

93

94

95

96

120
Perone, S.P., ACS Symp. Series, 265, 99 (1984).
Atkinson, T.V., Gregg, H.R., "MULPLT: A Multiple
Dataset, Fila Based Plotting Program", submitted to
the DECUS software library, Marlboro, MA (1984).

Hayes-Roth, F., Waterman, D.A., Lenat, D.B., "Building
Expert Systems", Addison-Wesley, Reading, MA (1983).

Hayes-Roth, F., Com uter, 11, 263 (1984)
Barr, A., Feigenbaum, E.A., "Handbood of Artificial

Intelligence", William Kaufman, Los Altos, CA, Vol. 1
(1981), Vol. 2 (1982).

Jurs, P.C., Computer Software Applications 15
Chemistry, Wiley-Intarscience, New York, NY (1986).

Rames, L.S., et. a1., Anal. Chem., 58, 295R (1986).

Martinsen, D.P. and Song, B.H., Mass Spectrom. Rev.,
4, 461 (1985).

Barr, A. and Feigenbaum, E.A., The Handbook 5:
Artificial Intelligence Vol. 5, William Kaufmann, Los
Altos, CA (1982).

 

Buchanan, B.G. and Feigenbaum, E.A., Artificial

Intelligence, 11, 5 (1978). 6"
Dessay, R., Anal. Chem., 56, 1200A (1984).

Dessay, R., Anal. Chem., 56, 1312A (1984).

Addis, T.H., Designing Knowledge-Baseg S stems, Kogan
Page Ltd, London (1985).

Pierce, T.H. and Hohna, B.A. (eds.), Artificial
Intelligence ;5_ Chemistr , ACS Symp. Series, 306
(1986).

Mclafferty, F.W. and Stauffer, D.B., 55_Chem. Inf.
Comp. Sci., 25, 245 (1985).

Wong, G.M., Crawford, R.W., Kunz, J.C., Kehlar, T.P.,
IEEE Transactions 55 Nuclear Sciences, NS-31, 804
(1984).

Wong, G.M., Crawford, R.W., Lanning, S.M., Brand,
H.R., presented at the 32nd Annual Conference on Mass
Spectrometry and Allied Topics; San Antonio, TX, May
27-June 1, 1984.

97

98

99

121

Wong, G.M., Lanning, S.M., Crawford, R.W., Brand,
H.R., presented at the 32nd Annual Conference on Mass
Spectrometry and Allied Topics; San Antonio, TX, May
27-Juna 1, 1984.

Brand, H.R., Lanning, S.M., Wong, C.M., submitted to

Int. Joint Conf. on Artificial Intelligence, August
18-24, 1985.

McLafferty, F.W., "Interpretation of Mass Spectra, 3rd

Ed.", University Science Books, Mill Valley, CA
(1980).

APPEND ICES

122

Appendix 1

Dr. Memory, SLOPS and control software

This appendix contains brief descriptions of each of the
commands and subroutines used in the TQMS control software.
For the purpose of brevity, the complete user’s manuals for
each of the three software packages (Dr. Memory, SLOPS, and
the control software) will not be reproduced. Each command
or subroutine in each user manual consisted two or more
pages. A sample of the format used is presented (for the
subroutine that checked with the user, then set/reset
specified bits, BITCHK), followed by only the description
section of the documentation for the rest of the
subroutines.

Appendix 1, part 1: Bit check example
part 2: Dr. Memory
part 3: SLOPS

part 4: TQMS control software

123

Appendix 1, Part 1

BITCHK

Description

Version

This routine, bit check, sends out a message, gets
a response, and clears, sets or leaves alone the
indicated bit in the indicated byte. A YES
response sets the bit, NO clears is, and nothing
(<CR>) does not change it.

hg 1.0

Required

Returns

Modified

B
DE
HL

bit position (mask)
pointer to the byte to change
pointer to query text

none

none

Pushes/Pops

Calls

Bugs

0/0 - register A modified

PRINT
TTYIN

BITCHK:

BITCHI:

BITCHZ:

8088
CALL
CALL
LDA
CPI
JZ
CPI
JZ
CPI
JZ
CPI
RNZ
MOV
CMA
MOV
LDAX
ANA
STAX
RET
LDAX
ORA
STAX
RET

BITCHK
PRINT
TTYIN
CHRBUF
’Y
BITCH2
’y
BITCH2
’N
BITCHl

124

-0 .6 -m -0 as. -0 use -0 -0 .0 -o as. .0 U. -0 -0 .. \ss -0 -0 -0 -0

ask the question

get the answer

get the first character
yes

or no

ignore ...
get
and reverse the mask

clear the bit

and head home
set the given bit

and head home

125

Appendix 1, Part 2

0r. Memory summary

 

Subroutines

Globals Definitions, memory locations, etc
Lowcore Restart and trap jumps

Interupts Interupt handling routines

RSTO Cold start, sets defaults

RSTl A \

RST2 BC \ Diagnostics - prints
RST3 DE \ the contents of the
RST4 HL / indicated registers
RST5 PC /

RST6 SP+f1ags/

RST7 Software breakpoint

Kernal Dispatcher for the monitor
Commands Processes several commands

Go Go and Procede commands

Quest Prints a summary of the commands
Regstr Open a register for modification
Talk Connect terminal to PDP 11

Chrchk Checks and converts ASCII to binary!
Close Closes an open location*
Crlf Outputs a <CR><LF> combination
Downld Downloads data from the PDP ll
Efclr Clears an event flag
Getnum Gets a number input
Gettt Gets a character from the USARTs
Lights Sends a predefined light pattern out
Lite Sends a specified light pattern out
Modify Modify a location in memory*
Nulljb Boredom routine
Open Opens a memory locations*
Print Prints an ASCII string
Putnum Outputs a number
Puttt Writes a character to a USART
Rhlr Rotates HL right
Stkbit Stacks 3/4 bits of HL on stackt
* these routines are useful only to Dr. Memory, and

probably will not help the casual programmer.

126

GLOBALS This is a list of mnemonics that assign values to
ASCII characters, memory locations, etc.

LOWCORE These are the hardware vector jumps. This routine
just transfer control to the processing routines.
The restart interrupts are transferred to RAM where
RST 7’s await (unless user modified).

INTERRUPTS These routines process interrupts from devices.
The USART interrupt routines fetch and store one
character. The break interrupt routine does a warm
restart.

RSTO This cold start routine initializes several
parameters needed for the monitor to operate.
Breakpoint instructions are placed in RAM where
unused interrupts transfer control, jump tables set
up, flags set to default values, default user stack
set up as well as a default program counter.

RSTl Debugging aid: prints the contents of register A
RST2 Debugging aid: prints contents of register pair BC
RST3 Debugging aid: prints contents of register pair DE
RST4 Debugging aid: prints contents of register pair HL
RST5 Debugging aid: prints the user’s program counter

RST6 Debugging aid: prints the value of the stack
pointer and flags

RST7 Warm start or breakpoint entry. This routine saves
all of the user registers and status. Control is
returned to either Dr. Memory or Slope, depending
on who is in control. '

KERNAL This is the heart of Dr. Memory. It accepts
numbers (addresses and values) and commands and
dispatches to the appropriate routine.

COMMANDS These are routines that process Dr. Memory’s
commands. Those processed in this module are
slash, backslash, carriage return, line feed, up
arrow, Slops, Rex, and Octal.

GO Two of Dr. Mamory’s commands live here: Go and
Procede. Procede is useful - it restores all user
status and registers, including program counter.
This gets us back to ’user mode’.

QUEST

REGSTR

TALK

CHRCHK

CLOSE

CRLF

DOWNLD

EFCLR

GETNUM

GETTT

LIGHTS

127

This is the ? command of Dr. Memory. It prints a
summary help message.

This is another of Dr. Memory’s commands, the S.
It is used to translate a register mnemonic to an
address, and open one or two bytes for
modification.

This is a routine that ’connects’ the two USARTs

together. Used to communicate to the POP 11 thru
the micro. Note that there is no exit to this
routine. As such, it is both a command and a

subroutine. A break is needed to exit.

Character check. Numbers (actually digits, in
register A) are converted to binary (numbers either
octal or hex) and the result is left in register A.
Non-numbers cause the carry bit to be set.

Sets the open/closed flag to closed, and write out
a carriage

This routine sends a carriage return/line feed
combination to the terminal.

This subroutine talks to the PDP 11, and expects a
defined protocol for loading the micro’s memory.
Binary records are loaded directly into memory,
while ASCII records are sent to the terminal.

This routine clears the event flag specified in
register A.

This routine gets characters from the keyboard,
echoes them, checks them, and if numeric, builds a
number and loops for more. The number, if entered,
is built in HL and the ’entered number’ flag is
set. This routine returns when a non-numeric is
entered.

These are the get character routines. GETTTO gets
a character from USART 0, while GETTTl gets a
character from USART 1. Each routine waits until a
character is entered. The character is returned in
register A.

This routine displays a characteristic light
pattern each time called. The pattern displayed is
stored at location LITES.

LITE

MODIFY

NULLJB

OPEN

PRINT

PUTNUM

PUTTT

RHLR

STKBIT

128

This routine turns lights on/off as defined by a
mask in register A. Each of the low 4 bits specify
a light as follows:

0 ==> bottom PS light
1 ==> middle PS light
2 ==> top PS light

3 ==> SOD light

This routine will modify a memory location if there
is an open location and a number has been entered.

This routine waits for the occurence of specified
event flags. Register A contains a mask of the
flags to wait for. While waiting, this routine
plays with the lights.

These routines set the open location flag and
writes the contents of the open byte(s) to the
terminal. OPEN opens a one byte location, OPEN2
opens a two byte location.r

This subroutine prints a string of ASCII characters
which ‘are ended by a null (0). HL point to the
start of the string.

This routine writes out a number, one or two bytes
long, octal or hex, to the terminal. HL contain
the number to output, and B contains a flag byte: 1
for byte output, 0 for word output.

These routines output a byte to a USART. PUTTTO
sends a byte to USART 0, while PUTTTl send it to
USART l. The character to send is in register A.
These routines wait until the USART is ready to
send.

This routine rotates register pair HL right one
bit. The most significant bit is always set to 0,
and the least significant bit is returned in the
carry flag.

This routine takes the least significant 3 or 4
bits, depending on the octal/hex flag, of HL and
pushes them on the stack under the return address.

 

1F -..-5&7. i" ‘i

129

Appendix 1, Part 3

SLOPS summary

Globals

Globals Mnemonics, address, etc for Dr. Memory
Sglobals Globals for Slops

Init Initialization routine

Kernal Driver routine and dispatcher
Subroutines

Addr Returns code address of library entry
Ascii Converts binary to and from ASCII
Blank Clears the screen

Brkdwn Breaks input line into words

Check Compares word with library entry
Cvtext Converts a number to ASCII string
Cvtint Converts ASCII string to binary number
Dcmp Double compare

Delay Software time delay

Link Links to next library entry

Number Get number from word stack

Search Search library for a match

Ttyin Gets a line of input

Word Get word from word stack

(also see subroutines listed under Dr. Memory)

Arithmetic gpbroutineg

Ddiv
Div
Dmult
Dsub
Mult

Commands

Convert

Downld
Drmem
Talk

Double divide
Divide '
Double multiply
Double subtract
Multiply

Converts a number to any base
Loads from the PDP 11 (See Dr.
Puts Dr. Memory in control
Talk to the PDP 11

Memory)

130

GLOBALS This is a list of mnemonics, address assignments,

SGLOBALS

INIT

KERNAL

ADDR

ASCII

BLANK

BRKDWN

CHECK

CVTEXT

etc. for Dr. Memory. Needed for proper assembly.

Slops globals. Mnemonics for Slops use,
definition of memory locations and subroutine
addresses.

This routine initializes a few parameters, and sets

Dr. Memory to look directly at Slops, not here.

This is the heart of Slops. It gets a command
line, searches the library, and if the command if
found, starts the named routine.

This routine returns a pointer to the code of a
library subroutine entry, given a pointer to the
entry. e.g. the pointer is incremented past the
flag byte, the name and the link field, if present.

These are two routines, TOASCI and UNASCI. TOASCI
adds either 48 or 55 to the number in register A to
make it either numeric or alpha, as needed. UNASCI
checks the character in register A to see if it is
a legal digit, with radix given in register C. If
legal, the number is converted to binary. Carry
flag is set otherwise.

Blanks the screen and resets the cursor to the
’homa’ position.

This utility routine breaks a line of input
(assumed to be in the terminal buffer) down into
’words’. A word is defined as alphanumeric
characters delimited by non-alpha characters.
These non-alpha characters are also words of length
l. Spaces are defined as non-alpha delimitars, but
do not count as words. Text after a semicolon is
treated as a comment, and not broken down into
words.

Compares the current word (eg an entered ’command’)
with an entry in the library. Sign flag set if the
words don’t match.

Converts a number (DE) to an ASCII string (radix
(C)). The resultant string is ’pushed’ to an area
pointed to by HL. The end of the string is flagged
by a null (0). Note that the string built is built
backwards, so on entry, HL must point to the TOP of
a buffer. On return, carry set implies a bad radix
was given.

CVTINT

DCMP

001V

DELAY

DIV

DMULT

DRMEM

DSUB

LINK

MULT

NUMBER

131

This routine converts an ASCII string (radix (C))
to a number (DE). The carry flag will be set on
return if a bad character was discovered. On
return, carry set implies a conversion error (bad
character in string).

Double compare, patterned after the 8085 CMP
instructions. Here, flags (S, CY, 2) set according
to BE - HL.

Double divide. The following takes place:
BC = HLBC / DE, remainder in HL

Software timing loop generate approximate time
delays. HL contain the count, each count is about
10 microseconds. Note that HL=1 is the shortest
time, and HL=0 is the longest.

Divide routine. This acts as follows:
DE = DE / A, remainder in A

Double multiply. This routine acts as follows:
DE = DE 3 A

This routine sets Dr. Memory as the kernel in
charge, and returns to it.

Double subtract. This routine acts as follows:
HL = HL - DE
This is the counterpart to the 8085 instr DAD D

This routine links from one library entry to the
next. e.g. BC pointing to a library entry is
redirected to point to the next entry. The sign
flag is set if no more entries exist.

Multiplication. This routine acts as follows:
DE = D t E

This routine tries to find a number as the next
element on the word stack. If found, conversion to
binary is attempted. The conversion radix may be
specified with the number or defaulted. Radix
qualifiers follow the number, as follows:

’ octal
decimal
hexadecimal
radix n, where n is any valid
number.
The carry flag tells of errors, which might include
any of the following: no next word, next word not
a number, or the radix qualifier (number after :)
is not a number.

0
fl

VVV

:n

 

SEARCH

TTYIN

WORD

CONVERT

132

This routine searches through the library for a
match with the current word. If‘a match is found,
the entry is passed back along with its flag byte.
The sign flag is used as an error indicator for no
match.

Keyboard driver. Accepts a line of input, ended by
a control character (ASCII value < 32), stores the
line in the input buffer and echoes each character
as typed. This routine also processes the
following special characters:

delete deletes the last character

escape blanks the screen

cntl/U deletes the current line of input

Pops pointer to and length of next word off the
word stack. If no more entries exist, the sign
flag is set.

This command converts a number from one base to
another. To invoke this routine, type:

CONVERT number base
where ’number’ is the number to be converted to the
base ’base’. For a description of valid numbers,
see NUMBER.

 

133

Appendix 1, Part 4

Control system summary

Commands

INI Initialize the system

GET This routine gets a set of stored parameters

SAVE This routine save the current parameter table

PARAM This is the general parameter setting routine

MANUAL Routine for ’manual’ control

FSCAN Fast scan - see the spectra on the oscilloscope

SCAN Scan, collect data and display it

WDATA This routine writes the data from RAM to a disk
file

Graphics gpbroutinag

CHRLIB
CLEAR
DRAW
GRFAXS
GRFCHR
GRFLAB
GRFVEC

Display

UPDATE

UPSTAT
UPSNUM
UPINT

UPMASS
UPGRAF
UPAXES
DLAST

Character library

This subroutine/command clears the graphics screen
Etch-a-sketch command '

This is the axis drawing routine

This subroutine draws a character on the screen
This routine draws a string on the graphics screen
Draws a vector on the graphics screen

update routines

This routine updates the status
displays

This routine updates the status display

This routine updates numbers on the status display
Format and display the intensity on status display
Format and display the mass on the status display
This routine updates the graphic display

This routine draws the axes

Draws the last collected intensity

and graphic

 

134

Other ggbroptineg

BITCHK
FMASS
FORMAT
GETGRF
GETMAS
GETPM
GETPRM
INCMAS
INTENS
INT16
MSADC
MSDATA
MSINIT
MSMASS
PVRAM
SMASS
SMGL
SRANGE
STARTP

Sets/clears bits in a flag based on user response
Formats the mass as follows: aa:xxx.x

Format a number for output

This routine returns a flag byte with graph info.
This routine gets a mass from the word stack

Get X coordinate for current mass for display

Get a new set of parameters

Increment the quads to the next mass

This routine collects the data

This routine returns the sum of 16 intensities
Gets one ADC value

Store current intensity in memory

Initialize the control hardware

Send the quad controller a mass via DACs

Put text to the status display video RAM

Set current mass

Send message to user, get line of response

Set range of Kiethley amplifier

This is a file of the startup (default) parameters

135

Commands

INI

GET

SAVE

PARAM

MANUAL

This is a routine to clean things (mainly the
display) up. The data RAM is zeroed, graphics and
status screens are cleared and updated, range and
mass reset.

This routine gets a set of parameters from a stored
bank. The number of the bank where the parameters
are stored is requested, and the entire bank is
transferred into the current parameter table.

This routine save the current parameter table in a
bank of parameter tables, for future recall. Use
is as follows:

SAVE n
where n is a number between 1 and 4 inclusive.

This is the general parameter setting routine. It
acts in a ’singla character input’ mode for the
parameter to change, i.e. ’0’ not ’QUAD’ for
changing a Quad’s mass range. At this time, little
checking is done of the parameters, and each
parameter must be set individually, i.e. changing
the’ mass range does not change the graph
parameters.

The parameters for either graph, the UPPER or the
LOWER, must be changed individually, and the
commands U and L identify which graph is being
changed.

The following commands are currently supported:

Q n low-high Change Quad n’s mass scan

T int range Change the threshold

R min-max Change the min and max range

A n Change the # points to average

G low,n Change low mass on the graph

F Change flags, each queried

M text' Puts message on status display

H text Puts a header on graph

? Prints a summary of the
options

“2 to exit (control 2)

Routine for ’manual’ control. Using the keypad of
the terminal, the keys 1, 2 and 3 are for quad l;
4, 5 and 6 are for quad 3; and keys 7, 8 and 9 are
for the range. Keys 1, 4 and 7 decrease the
current value by 1, keys 2, 5 and 8 set the current
value into the parameter table, and keys 3, 6 and 9
increase the current value by l.

 

FSCAN

SCAN

WDATA

136

This routine increments the mass, checks for a
typed character, and if none entered, loops about.

This routine scans the set mass ranges, collects
data, updates the screens, records the data and
loops until a character is typed.

This routine writes the data from RAM to a disk
file.

Graphics subroutines

CHRLIB

CLEAR

DRAW

GRFAXS

GRFCHR

GRFLAB

This is not a subroutine, but a library of vector
moves that draws the ASCII character set on the

MATROX graphics board. A list of pointers, in
ASCII“ order, points to the entries for each
character.

This subroutine/command clears the graphics screen.
The actual clearing takes a maximum of 34 msec, but
this routine does not wait.

This command allows the user to draw, dot by dot,
on the MATROX graphics board. The numeric keypad
is used to determine the direction, 5 ==> don’t
move, 9==> northwest, 4 ==> east, etc. The 0 key
tells the routine to clear dots in the drawn path,
the period (.) tells the routine to draw points.
To exit, type a slash (/).

This is the axis drawing routine. It will draw an

axis either vertical or horizontal. Large and
small tick marks can also be drawn, and large tick
marks can be labeled. This routine works from a

table of data the defines the axis to be drawn.
The table (an axis description block) is defined
below. '

This subroutine draws a character on the graphics
screen. The ASCII code for a character is passed
in register A, the graphics board must be aware of
the address to plot at (lower left corner of the
character space), and this routine will draw the
character from the library (CHRLIB).

This routine draws a string of characters on the
graphics board. The supported characters are all
the printing ASCII characters. The string can be
left, center or right justified. Registers D and E
contain the coordinates of the lower left (center
or right) for the string to be drawn. HL points to
the string, terminated by a null.

 

GRFVEC

Display
UPDATE

UPSTAT

UPSNUM

UPINT

UPMASS

UPGRAF

UPAXES

DLAST

137

Draws a vector, of length (C), in the direction (B)
[Note that the direction mnemonics (north, south,
east, west, ne, se, nw, sw) should be used],
starting at physical coordinates X and Y (D,E).

update routines

This routine updates the status and graphic
displays.
This routine updates the status display.

Individual parts of the display can be updated, or
the entire screen can be blanked and redrawn.

This routine updates the numbers on the status
display for either the upper or the lower scan.
Not updated are the descriptions, the current info,
or the messages; just the numbers under the "UPPER"
or "LOWER" headers. -

This routine scales the given intensity and range
to the format xx.x -xx and sends it to the status
display, at location pointed to by HL.

This routine updates a mass range on the status
display, in the following format: xxx-xxx

This routine updates the graphic display. The
entire graph may be updated, or only the current
information can be plotted.

This routine draws the axes, taking into account
single or double plot, linear or log scale, low
mass for display, and # pts/AMU.

This routine updates the last mass, and the current
intensity. This is done to allow time for the
quads to settle, ions to be selected, etc. before
the next intensity is collected. While waiting,
graphing the last displayed point seemed like the
thing to do.

Other subroutines

BITCHK

FMASS

This routine sends out a message, gets a response,
and clears, sets or leaves alone the indicated bit
in the indicated byte. A YES response sets the
bit, NO clears is, and nothing (<CR>) does not
change it.

This routine formats the given mass as follows:
aa:xxx.x, where as is either RF or DC, and xxx.x is
the mass.

FORMAT

GETGRF

GETMAS

GETPM

GETPRM

INCMAS

INTENS

INT16
MSADC

MSDATA

138

This routine converts a number from registers DE
into ASCII in a predefined format, into a buffer
pointed to by HL. This routine converts numbers to
decimal, adds the required number of trailing
zeros, inserts a decimal point, and pads the from
of the string with spaces. The user than cells
PRINT to output the formatted string.

This routine returns a flag byte with graph
information.’

This routine attempts to get a valid mass from the
word stack. If a number was entered, it is scaled
for the DAC’s in use (MASS = MASS * 8.). If no
number given, the carry flag is set.

This routine returns the X-coordinate on the
graphics display corresponding to the current mass.
Also returned are some of the graph parameters.

This routine changes the entire parameter table for

Mass Spec control. A table, set up exactly like
the parameter table, is pointed to by HL, and its
contents ‘ are moved to the current or active

parameter table.

This routine checks the parameter table, finds
which quads want their mass incremented, and
increments them. Checks are made if any of the
masses increment past the and of their allowed
scan, and if so, the masses are reset.

This routine collects the data. A small sample of
intensities (16) are collected, and this set is
used to determine if autoranging is required. If
so, the Keithley is ranged, and time is allowed for
it to settle. This is repeated until the signal is
in range, or no more auto-ranging is possible. The
signal is then sampled and averaged the required
number of times.

This routine returns the sum of 16 intensities. It
is intended to give an idea of the current ion
intensity.

This is the lowest level data acquisition routine.
It requests and receives a datum from the analog to
digital converter.

This routine stores the current intensity in the
data RAM, at a location corresponding to the DAC
value (mass).

MSINIT

MSMASS

PVRAM

SMASS

SMGL

SRANGE

STARTP

139

This routine sets up the parallel ports used for
the DACs, ADC. It also gets the default
parameters, sets masses to reasonable values, and
cleans up the displays.

This is the lowest level routine to send a mass to
the DACs controlling the quad’s mass.

This routine prints the text pointed to by BC to
the VRAM pointed to by HL, offset by DE. The
display is written into during vertical or
horizontal flyback, minimizing flicker.

This routine gets the "current masses" from the
parameter table, and sends them to the DACs.

This routine sends a message (pointed to by HL),
and gets the response from the user. The message
is broken down into words, and an attempt to get a
number is done.

This routine sends a new range to the Keithley, and
delays a bit to allow the amplifier to settle and
become reasonably stable.

This is a file of the startup (default) parameters.
0n startup, this file is copied into the parameter
table.

140

Appendix 2

Multi-dimansional data base subroutines

This appendix contains descriptions of the subroutines used
to interface with the multi-dimensional data base format.

 

141

—~_*

SUBROUTINE MSINIT

This subroutine (along with the PARAMETER statements in
MDDB.CMN) set up the characteristics of this system.
All possible datasets are initialized, and system-wide
globals are set up, and the dictionary file is opened.

C

C

C

C

C

C

C calls CONCAT (STRING)
C SCOPY (STRING)
C DSINIT (MDDB)
C EXIT (RSX)
C

C

C

C

C

h gregg jul-82

SUBROUTINE MSOPEN(IFILE,QPRMPT,QTYPE,LOONl,LOON2,LOON3)
BYTE QPRMPT(1),QTYPE(1)
INTEGER*2 IFILE,LOON1,LOON2,LOON3

This subroutine open (close) the files associated with a
dataset. New or Old files may be opened, and filenames
can be prompted or defaulted.

where IFILE the number of the data set to open

negative means close that dataset.

QPRMPT prompt for asking for a dataset name
if LENGTH(QPRMPT) = 0, use QDSFIL as
the dataset name (in common).

QTYPE is either ’NEW’, ’OLD’ or ’READONLY’

LOONn are the logical units to use for the
dataset

returns (if ’OLD’, see GETPRM,GETCOM)

calls CONCAT (STRING)
LENGTH (STRING)
GETCOM (noon)
GETPRM (MDDB)

h gregg ju1—82

OOOOOOOQOOQQQOOOOOOOOOOOO

142
Subroutines 55 write into 5 dataset

SUBROUTINE PUTPRM(IFILE)
INTEGERIZ IFILE

This subroutine writes parameters to the header file.
with the current time and data.

where IFILE is the dataset number

uses NUMSTC the number of static variables
ISTATC code for the static parameters
RSTATC value of corresponding parameter x
NUMVAR the number of static variables
IVAR code for the variable parameters

calls DATE (RSX)
TIME (RSX)

h gregg jul—82

OQOOOOOOQQOOOOOOGOO

SUBROUTINE PUTCOM(IFILE)
INTEGER¥2 IFILE

This subroutine writes a comment (title, words of wisdom,
etc.) to the header file, including the current date and
time, if needed.

where IFILE is the dataset number

uses QDATE is the date to write out.
If LENGTH(QDATE) = 0, get current date.
QTIME is the time to write out.
If LENGTH(QTIME) = 0, get current time.
QCOMNT an array holding the comment
The comment is assumed to and with a null
and be 80 bytes or less long.

calls DATE (RSX)
TIME (RSX)
LENGTH (STRING)

h gregg ju1*82

OOOQOOQOOOOOOOOOOOOOOOOO

 

00QQOOOQOQOOQQQQOOOOOOOOOO

143

SUBROUTINE PUTDAT(IFILE,X,Y)
INTEGERXZ IFILE
REAL34 X(l),Y(l)

This subroutine puts a scan of data into the pointer and
data files. All pointers are kept internally, and some
values are calculated. .

where IFILE is the number of the dataset to use
X,Y are the X,Y data pairs

uses NUMDAT number of data points
IVARD code for X of X,Y pair
RTIME time the scan was taken
RVAR values of the variables
IVARF codes for fast variables
RVARF* values of fast variables, from RVAR.
RSUMY* sum of Y values, done here
ISCAN* + l is where this scan will go

* ==> calculated or kept internally, do not change!

calls PTDATM (MDDB)
PUTPTR (MDDB)

h gregg jul-82

S

OOOOOOOOOOOQOOOO

OOOOOOQOOOOOOOOOOOO

144

ubroutines to read from a dataset

SUBROUTINE GETCOM(IFILE)
INTEGER#2 IFILE

This subroutine gets the next comment in the header file,
along with its time and date of entry.

where IFILE is the dataset number
returns QDATE the date of the comment

QTIME the time of the comment
QCOMNT the comment (or title) itself

calls none

h gregg ju1—82

SUBROUTINE GETDAT(IFILE,JSCAN,X,Y)
INTEGERtZ IFILE,JSCAN
REALt4 x(1).Y(l)

This subroutine gets all information associated with
the current scan. All error information is from the
subroutines.

where IFILE is the dataset number
JSCAN is the scan number to retrieve
X,Y are arrays for the X,Y data pairs

returns (sea GETPTR, GETVAR, GETXY)
calls GETPTR (MDDB)
GETVAR (MDDB)
GETXY (MDDB)
h gregg jul-82

 

000000000000000000000

00000000000000

145

SUBROUTINE GETPTR(IFILE,JSCAN)
INTEGER*2 IFILE,JSCAN

This subroutine gets JSCAN’s pointer variables out
of the pointer file.

where IFILE is the dataset number
JSCAN is the number of the pointer to get

returns IPTDAT pointer into the data file
NUMDAT number of data points
RTIME time this scan was taken
RSUMY sum of the Y values
IVARD code for the dependent variable
IVARF codes for the fast variables
RVARF values of the fast variables

calls none

h gregg ju1-82

SUBROUTINE GETVAR(IFILE)
INTEGER¥2 IFILE

This subroutine get the variable parameters for the
current scan (i.e. GETPTR must have been called).

where IFILE is the dataset number
returns RVAR variable values for the current scan
calls none

h gregg ju1-82

 

0000000000000

000000000000000

146

SUBROUTINE GETXY(IFILE,X,Y)
INTEGERtZ IFILE
331114 X(l),Y(l)

This subroutine gets the X,Y pairs for the current scan
(i.e. GETPTR and GETVAR must have been called).

where IFILE is the dataset number
X,Y are arrays for the X,Y pairs

calls none

h gregg ju1—82

SUBROUTINE GETEND(IFILE)
INTEGER*2 IFILE

This subroutine gets quickly to the and of the dataset,
and sets up the commons for the next write.

where IFILE is the dataset number
returns all variables necessary for the next write

calls GETPRM (MDDB)
GETVAR (MDDB)

h gregg may-85

 

 

“1.1

 

147

User interface subroutines

00000000000000000000000000000000000000000000

SUBROUTINE EXEDIT(INSTR,TRACE,IXY)
INTEGERIZ INSTR,IXY(6)

REAL*4

TRACE(4,1)

This subroutine edits the ’traca’ matrix used by the
extract subroutine. This trace matrix is the mask for
the extraction program, and this subroutine allows
manipulation of the input parameters (1&2).

where

INSTR is a return variable for the main program:
( 0 == 30 plot: -INSTR is the parameter
for 3rd dimension
= 0 ==> exit
= l ==> start an extraction
= 2 ==> file the data
= 3 ==> call MULPLT
= 4 ==> extract and file the data
= 5 ==> file and plot the data
= 6 ==> extract, file and plot the data

TRACE is the matrix of min/max values
(l,n)&(2,n) are the min/max limits allowed —
edited here
(3,n)&(4,n) are the min/max found in the
data base - from extract
IXY(l) is the code for the independent variable
IXY(2) is the code for the dependent variable
IXY(3) is the number of pairs extracted
IXY(4-6)aren’t used here

The numbers on the screen have the following format (see

NUMDIS):

integer - if it fits

real - if its fits

exponential: .xx-ee, .xxx-e, .xx+ee or .xxx+e
if negative, one decimal place is lost

The screen format is as follows:

lines 1&2 are for titles
lines 3-22 are for the numbers: .
3 variable 1 . . . . variable 2

4 variable 3 . . . . variable 4
etc
22 variable 39 . . . . variable 40

line 23 is for input
line 24 is for descriptive information

 

C Each variable location consists of a 40 character area as

C

000000000000000

000000000000000000000000000000

148

follows:
bytes 3-12 are a description of the variable
bytes 13,20,27,34 are spaces
bytes 14-19,21-26,28-33,35-40 are the numeric
fields 1-4 '
calls: GETPOS (MDDB)
VTxxxx (string)
EDISPL (MDDB)
NUMDIS (MDDB)
INPUT (MDDB)

SUBROUTINE INPUT(ICMD,
INTEGER*2 ICMD
REAL*4 VALUE

This subroutine works

h gregg 7/84,8/84

VALUE)

for EXEDIT, getting the user input,
correcting mistakes and translating the input into an

index for the main program.

calls VTxxxx (string)
returns: ICMD VALU what the user typed
l -- Extract, <PFl)Extract
2 -- File, <PFl>File
3 -- Plot, <PFl)Plot
»4 -- EF, <PF1>EF
5 -- FP, <PF1>FP
6 -- Go, <PFl>Go
7 -- Quit, AZ, <PFl>Quit
8 -- Z, <PFl>Z
9 -- Help, (PF2), <PFl>He1p
10 -- <up-arrow key)
11 -- (down-arrow key)
12 —— (right-arrow key>
13 -- <1eft-arrow key>
l4 -- (carriage-return)
15 value value, <PF3>va1ue
16 -- X, <PF1>X
l7 —- Y, <PF1>Y

h gregg may—85

 

 

 

149

Main extraction subroutine

00000000000000000000000000000000000000000

SUBROUTINE XTRACT(IFILE,NPLANE,IXY,TRACE,ARRAY,MAXARY)
INTEGER*2 IFILE,NPLANE,IXY(6,l),MAXARY
REAL*4 TRACE(1),ARRAY(1)

This subroutine extracts data from the multi-dimensional
data base. Several ’planes’ of data may be extracted at
once, dependent on the size of the data ARRAY. As many
planes of data as can fit into the ARRAY will be
extracted. The extraction parameters are passed in two
arrays - IXY and TRACE. IXY contains the codes for the X
and Y values to be retrieved, while TRACE contains the
constraints on the constants (min and max values forming
range). On return, TRACE will also contain the range of
values selected for every variable; this helps determine
if the extraction parameters were specified in enough
detail.

where IFILE is the dataset number
NPLANE is the number of X,Y planes to extract
IXY are codes and pointers for each plane

(l,m) pointer to code for X
(2,m) pointer to code for Y
(3,m) number of X,Y pairs extracted
(4,m) pointer to first X value in ARRAY
(5,m) pointer to first Y value in ARARY
(6,m) max number of X,Y pairs, no shuffle
Note that an additional plane is used intern’ly
and space for it must be allocated by caller.
TRACE extraction parameters
[equivalent to TRACE(4,NUMVAR(IFILE)+4,NPLANE)]
(l,n,m) minimum allowed value
(2,n,m) maximum allowed value
(3,n,m) minimum found value
(4,n,m) maximum found value
(i,l,m) scan number (ISCAN)
(i,2,m) time scan was taken (RTIME)
(i,3,m) sum of Y values (RSUMY)
(i,4,m) the Y values
(i,k,m) the variables [IVAR(ifila,j)]
where k=j+4
m is the extraction plane
ARRAY is the data array
MAXARY is the maximum size of ARRAY

 

000000000000000

calls ASHUFL
SHUFFL
GETPTR
GETVAR
GETXY
STORXY
TRACER
IT
ITRACE

(MDDB)
(MDDB)
(MDDB)
(MDDB)
(MDDB)
(MDDB)
(MDDB)
(MDDB)
(MDDB)

150

h gregg aug-BZ

 

151

 

Common area - variable definitions

C

C MDDB.CMN - common areas for the MDDB data base routines
C

00000

000

000

PARAMETER MXFIL=1

PARAMETER MXSTC=40
PARAMETER MXVAR=40

Variables common to all

for the dictionary:

open datasets
static params
variable params

number of
number of
number of

maximum
maximum
maximum

datasets:

BYTE QSDICT (20) 2 short dictionary description

BYTE QLDICT (58) ! long dictionary description
INTEGER*2 LUNDIC ! logical unit for the dictionary
INTEGERXZ IPTDIC ! dictionary pointer: no/val/point
misc. variables:

BYTE QSYSTM (20) ! device and UIC for common files
BYTE QDATE (10) ! the date \ either current

BYTE QTIME ( 8) ! the time / or from the dataset
BYTE QCOMNT (82) 2 comment lines from the header
INTEGER¥2 LOONTI ! the logical unit for TI: (errors)
INTEGER*2 IERR ! error return flag

INTEGERXZ MAXFIL ! == MXFIL

INTEGERXZ MAXSTC ! == MXSTC

INTEGERt2 MAXVAR ! == MXVAR

variables for each data set:

BYTE QDSFIL ( 30,MXFIL) ! dataset file name
INTEGER*2 LUNHDR ( MXFIL) ! LUN for the header file
INTEGERXZ LUNPTR ( MXFIL) 2 LUN for the pointer file
INTEGER*2 LUNDAT ( MXFIL) ! LUN for the data file
INTEGEth ISCAN ( MXFIL) 2 current scan number
INTEGER*2 IPTDAT ( MXFIL) ! pointer to current scan
INTEGER*2 JPTDAT ( MXFIL) ! pointer to memory record
INTEGER*2 IDAT ( MXFIL) ! pointer into RDAT

REAL*4 RDAT ( .16,MXFIL) ! one data record buffer
REAL*4 RTIME ( MXFIL) ! time this scan was taken

 

INTEGER*2
INTEGERIZ
REAL34

INTEGERtZ
INTEGERXZ
REAL*4
INTEGER*2
REAL¥4

INTEGEth
INTEGERXZ
INTEGER*2
INTEGERtZ
INTEGERtZ
REAL*4

COMMON
QSDICT,
LUNDIC,
QDSFIL,
NUMSTC,
NUMDAT,
RSTATC,

152

NUMSTC ( MXFIL) !
ISTATC (MXSTC,MXFIL) !
RSTATC (MXSTC,MXFIL) !
NUMVAR ( MXFIL) !
IVAR (MXVAR,MXFIL) !
RVAR (MXVAR,MXFIL) !
IVARF ( 3,MXFIL) !
RVARF ( 3,MXFIL) !
MAXDAT ( MXFIL) !
NUMDAT ( MXFIL) !
IVARD ( MXFIL) !
IPTX ( MXFIL) 3
IPTY ( MXFIL) !
RSUMY ( MXFIL) !
/MSFILE/
QLDICT, QSYSTM, QDATE,
IPTDIC, MAXFIL, MAXSTC,
LUNHDR, LUNPTR, LUNDAT,
ISTATC, NUMVAR, IVAR,
IVARD, MAXDAT,
RVAR, RTIME,

number of static params
code for static params
value for static param

number of var. params
code for variable params
value for variable param
code for the fast vars
values of fast vars

max number of data pairs
number pairs this scan
code for dependent var
pointer for X, this scan
pointer for Y, this scan
sum of Y, this scan

QTIME, QCOMNT,
MAXVAR, IERR,
LOONTI,
ISCAN, IPTDAT, JPTDAT,
IPTX, IPTY, IDAT,
RSUMY, RVARF

C __________________________________________________________

C

 

153

Appendix 3

EXTRACT User’s Guide

Hugh Gregg
June 24, 1985

EXTRACT is a program for interrogating a
multi-dimansional dataset and extracting
any two dimensional plane of data from
it. The resulting data can then be
formatted and presented to MULPLT for
immediate graphics presentation, or the
user may choose different extraction
parameters and extract another plane of
data.

 

154

EXTRACT, the program

The program EXTRACT was created to extract any two
dimensional plane of data from a multi-dimensional dataset.
This is accomplished by specifying a set of extraction
limits, and letting the program search through the dataset
extracting all the data that matches the specified
constraints. The actual limits of the extraction are then
presented to the user. If these limits are acceptable, the
user may plot (using MULPLT) the extracted data.

Multi—Dimensional Dataset

Datasets created by multi-dimensional instruments, such
as Triple Quadrupole Mass Spectrometers (TOMS), may contain
more than two dimensions of data (i.e. more than mass vs.
intensity pairs). A TOMS has many instrumental parameters,
each of which, when varied, creates an orthogonal dimension
of data. Examples of these dimensions include quad 1 mass,
quad 3 mass, the axial energy, collision pressure, and ion
intensity. Varying more than two variables results in a
dataset with multiple dimensions of data. A multi-
dimensional data base system (MDDB) was created to handle
these data.

An MDDB dataset is a set of three files (*.HDR, *.PTR,
*.DAT) that contain all the instrumental parameters as well
as the data. The data is stored in this dataset in a
series of "SCANS”. Each scan is the result of one
instrumental operation (i.e. an intensity vs. quad 1 mass
scan, or an intensity vs. axial energy scan at specific
parent/daughter masses). The values of all variables at
the time each scan was recorded are stored with the scan.
In this way, each scan is a two dimensional slice into the
multi-dimensional dataset with all other variables held
constant.

How extraction works

EXTRACT extracts a plane of data based on the
constraints specified by the user. The user is presented
with a list of all the variables in the dataset and must
choose what data is to be presented. The abscissa and
ordinate (X and Y) of the resulting plane are selected. If
the user wishes to limit the extraction process, the limits
of the variables must be set. When EXTRACT is told to do
it’s stuff, it extracts all data that satisfy the

constraints. When this is accomplished, the user is
presented with the actual limits found during the
extraction. If these limits are acceptable, the data may

be formatted for MULPLT and displayed. However, if the
actual limits of the extracted data indicate a variance
larger than acceptable, the limits of the offending
variable may be narrowed and the extraction tried again.

 

155

Using EXTRACT

To use EXTRACT, you must first convince RSX to let you
use it. This is done by typing ”EXTRACT" or "EXTRACT
datasetname". If you don’t specify the dataset name on the
command line, EXTRACT asks you for the name. If the
dataset exists, EXTRACT displays all the variables and
waits for the user to type something. One section of the
screen is highlighted - this is the "active" area. The
user may move to a new active area by using the cursor
control (arrow) keys on the terminal. To change a specific
value, move the active area to the position of the value to
change, then just type in the new value (followed by a
<CR>). Commands can be entered at any time, and take
immediate action. For more information about the commands
and how to enter values, see the sections on COMMANDS and
VALUES.

Screen Display Format

The screen format of EXTRACT displays all the user
variables stored in the dataset. Each variable has four
values associated with it: the minimum and maximum
extraction limits (set by the user) and the minimum and
maximum limits found by EXTRACT in the dataset.

Each value is displayed in only six character positions
and may appear truncated. If the value conveniently fits
within the six character window, it is displayed in full.
A problem arises when displaying large or small numbers

(i.e. 2.3456E7 counts). These numbers are displayed in
exponent format with implicit "times 10 raised to" (i.e.
.234+8). In the worst case, only one digit of the number

plus the exponent is visible (-.3-16).

The bottom line of the screen displays an enhanced
version of the active area. A number truncated in the main
portion of the screen is displayed in full along with the
full name of the variable. This status line is also used
to display definitions of certain values. For example, a
value of "2" for the variable "CAD gas type" produces the
description "CAD gas type: Nitrogen".

Values

Values may be entered in a variety of ways, the simplest
of which is to type in the new value. While this method
will always work, there are a few short cuts available.
When the active area is at either the minimum limit or the
maximum limit, the following may be used:

 

Minimum limits

(CR)

N1,N2

Maximum limits

(CR)

156

A single carriage return, when the minimum
limit is the active area, resets the
minimum and maximum limits to "no limit".
This allows EXTRACT to use any values for
this variable.

Entering a number N results in that value
being inserted into both the minimum and
the maximum value slots. Note that the
input of exponents must be done in
computer notation (i.e. 1.23E7), and that
a carriage return must be typed after the
number.

This format enters the number N1 into the
minimum value slot, and the number N2 into
the maximum value slot.

This format enters the number N into the
minimum value slot, and the number N+i
into the maximum value slot. This format
allows easy entry of a range of limits;
you specify the minimum value and the
range or increment. ‘

This vrmat enters the number N into the
min1n.m value slot, and the number N+l.0
into the maximum value slot. This format

is identical to the above format, but
defaults to a range of 1.0 (useful for
mass selection).

This format enters the number N into the
minimum value slot, and sets the maximum
value to infinity (1.0E36).

A carriage‘ return at the maximum limit
sets the maximum equal to the minimum
limit.

Entering a number N results in that value
being inserted into the maximum value
slot.

 

157

Command Summary

This is a list of the available commands and a brief
description of their use. All commands may be abbreviated
to their first letter, and are explaned in greater detail
below.

Single keystroke commands:

arrow keys move the active area
(PFl) get the "Command?" prompt
<PF2) Help: display the known commands
<PF3> get the "Value:" prompt
(OR) at first limit: sets "no limits"
at second limit: set second value to first
AZ Exits EXTRACT

anything else starts either a command or value

Other Commands (in response to the ”Command?" prompt):

X set the abscissa

Y set the ordinate

Z three-dimensional extraction and file
Extract Extract a plane of data

File File the extracted data

Plot call MULPLT

EF Extract and File a plane

FP File and Plot extracted data

Go Extract, File and Plot extracted data
Save Save the current extract limits

Read Read in saved extract limits

Help Display the command summary

Quit Exits EXEDIT

Description of Commands

X or Y This command sets the active area to be
either the abscissa or ordinate (X or Y)
for the resulting extract plane. A

highlighted letter (X or Y) is placed in
front of the variable name to indicate the
current X and Y variables.

2 This command executes a three-dimensional
extract. Position the active area to the
variable that will be the third axis (note
that this variable’s limits must be set).
Enter the 2 command, and you will be asked
for the incremental value for the 2
variable. This procedure does a series of
two dimensional extracts, starting with 2
equal to the minimum limit, and subsequent
extraction with the 2 value incremented
until the maximum limit is reached. Data

 

Extract

File

Plot

EF

FP

Save

158

from each two dimensional extract are
filed. An auxiliary program (PLOT3D) is
used to read this file and produce MULPLT
files for pseudo three-dimensional plots.

This command extracts a plane of data from
the dataset and displays the resulting
extraction limits. The data extracted are
not automatically filed or plotted.

This command files previously extracted
data. The filename is derived from the
dataset name (*.BIN), and each File
command in one extraction run uses the
same file (the MULPLT tag for each file
command starts at "AA" and is incremented

for subsequent file commands: "AB", "AC",
etc). A MULPLT command file is also
created (*.PDL). Before the first file

command is executed, the user is asked
whether the plot will' be a point plot,
line plot, bar graph or spectra plot.
Point plots and bar graphs are just that;
*the data for line plots will be sorted;
spectra plots are normalized to the base
peak, and the maximum intensity is put
into a MULPLT special features file
(*.SPF).

This command go directly 'to MULPLT,
passing it the name of the command file
generated by the File command. The user
is left in MULPLT, ready for a 00 command,
or whatever. To exit MULPLT, type Halt or
AZ, and you will be returned to EXTRACT.

Extract and File command for those times
that you know the extraction will work,
but you don’t want the data plotted yet.

File and Plot command. Used after doing
an extract command, and you wish to plot
the data.

Extract, File and Plot extracted data,
this command ”Goes for it", and does it
all.

This command saves the current extraction
limits in a file (the default file name is
derived from the dataset name, *.EXT).
This is a normal file that may be edited,
displayed, etc. For information on the
file format, see the Read command below.

 

159

Read This command reads in a file of saved
extraction limits (the default file name
is derived from the dataset name, *.EXT).
The file contains lines with the following
form: "variable = N1 to N2" for each of
the variables with limits set. The
variable name must exactly match the short
dictionary definition, and the two
keywords "=" and ” to " are not optional.

Help This command produces a short list of
available commands.

Quit , This command exits EXTRACT.
The mechanics of EXTRACT

Internally, EXTRACT consists of three major phases:
extraction parameter editing (EXEDIT), extraction (XTRACT)
and formatting for plotting (MSPLOT). Both the editing and
plotting phases are described in the COMMANDS section. The
extraction routine itself basically operates as two units.
The first unit processes the simplest of the extractions;
i.e. when the extraction plane is a scan. In this case,
the requested plane of information is stored.in the format
that the user wants, and all that is required is to
retrieve it.

The retrieval of a plane of data that wasn’t scanned is
more complicated. A large loop is initiated where each
cycle of the loop looks at one scan in the dataset. The
limits specified by the user are used to try to reject a
scan from further consideration. If the scan cannot be
rejected, the X and Y variables that the user specified are
studied. If there are multiple X values, all X values are
retrieved along with their corresponding Y values and are
stored in the extraction plane. If there are multiple Y
values and only a single X, the Y values are averaged and
the resulting X, Y pair is placed in the extract plane.
Finally, if no X or Y values match their extraction limits,
no data from this scan is saved. Once this scan is
finished, EXTRACT loops back and gets the next scan until
the entire dataset has bee examined.

The dimensions of the internal arrays in EXTRACT are not
fixed as in conventional programs. This allows quite a bit
of flexibility in the size of the scans and planes
extracted. Space for the data in each scan (from the
dataset) and the extract plane (under construction) is
dynamically allocated from a pool of 16000 bytes (room for
2000 X,Y pairs). Therefore, as the extract plane is built
(more data pairs added to it), the room available to hold
the next scan diminishes. In practice, this space is more
than enough. If we assume a fixed length scan of 1000 X,Y

 

160

pairs, we still have room for an extraction plane of 1000

X,Y pairs.

GLOSSARY of terms used

Active area

Extract

Plane

Scan

Value slot

(CR)

(PFI)

<PF2>

<PF3>

describes the area currently highlighted
on the screen.

the process of searching the dataset for a
specific two dimensional plane of data.

a two dimensional grouping of data,
usually extracted from a multi-dimensional
dataset.

one instrument scan; i.e. one grouping of
data consisting of X,Y data pairs and the
corresponding values for the variables.

one of the four minimum or maximum values
for a variable. A value slot may become
the active area.

the carriage return key.

the PFI key on VTlOO’s, the "gold" key.
Used to start a command.

the PF2 key on VTlOO’s, next to "gold"
key. Used to obtain Help.

the PF3 key on VT100’s. -Used to start a
value.

control Z (hold the control key down and
press the Z key). Used to exit EXTRACT.

 

 

 

     

RS

"17111111711111!fillflffllwliflflllflltlfl