AMS -1
October 10, 1969
W. Reynolds, R. Tucker, R. Stillman, J. Brdiges

Mass Spectrometers in a Time Shared Computer Environment

W. E. Reynolds, R. B. Tucker, R. A. Stillman, and J. C. Bridges
Instrumentation Research Laboratory
Genetics Department, Stanford University School of Medicine
Stanford, California

ABSTRACT

The use of a time shared computer system for mass spectrometer data acquisition
and/or instrument control is described. The computers involved are an IBM 360/50 wits an
interconnected IBM 1800 and a classic LINC. The use described is in a time shared mode,
where numerous data processing and instrumentation users are simultaneously served. The
mass spectrometers connected are a mix of a high resolution double focusing, a medium
resolution single focusing, and a quadrupole instrument in diverse areas, ranging to
1500 feet from the central computer. These and other instruments are served without any
prior scheduling or the particular knowledge of other users. Programming is done on
remote terminals in a high level language. The experiences in these modes enables
comments to be made concerning the relative merits and useful roles, deficiencies, and
benefits computer time sharing can offer. This work was sponsored in part by National
Aeronautics and Space Administration Grant NGR-05-020-004 (Genetics Department), NIH
Grant AM 04275-07-S1l (Chemistry Department), and NIH Grant 5 PO? FROO311-02 (Medical
School-Computation Center, Advanced Computer for Medical Research).

In 1965 a group of Stanford researchers and computer science planners formed a
committee to establish the basic design of a time-shared computer system that would not
only provide the time shared terminal access as had been pioneered by MIT and others, but
also would include comparable time shared data channels with connections directly into the
laboratory. Early large computers and batch processing had separated the user from the
computer. Time shared computers later restored the user to direct contact with the
computer. Somewhat in the same manner it is hoped to free the laboratory instrumentation
from the constraints of the small computer, or from the restrictions and delays of
submitting recorded data to a batch system. The time shared large computer should be able
to give acceptable real time service to the instrument in the laboratory.

In Figure 1 the Stanford system, ACME, is shown composed of a 360/50 with 2,065,000
bytes of core with a satellite IBM 1800. The latter functions only as low speed
instrument data channel. I have included the Genetics Department's classic LINC computer
in the general mass spectrometer systems. It can operate in the system or stand alone.
There are three distinct instrumentation data paths possible from a laboratory to the
central computer:

1. <A high speed unit, up to 25 K bytes/second, called the 270X-270Y. The

270Y is remote in the laboratory. The 270K is local to the computer
and has access through the multiplexer.

2. Low speed, up to 1 K samples or words per second, via the 1800 inputs.

3. Last, if used, the a-to-d converter of the LINC.

There are output channels via each path that have analogous attributea. On the left side
of Figure 1 are the principal mass spectrometers of this system,

1. Double focus, High Resolution, AEI MS-9.

2. Single focus, Integer Resolution, ATLAS CH 4.

3. Quadrupole, Integer Resolution, Finnigan 1015.

4, Time-Of-Flight, Integer Resolution, Bendix Mod. 12.

A typical laboratory instrumentation center in our concept would contain a data
terminal, a keyboard for programming and control, and a graphical output capability.
This is functionally arranged in Figure 2.

The computer complex and domain is below the dotted line. Physically the computer

connections are in the laboratory, but design, maintenance and specifications of the
terminal lie with the computer administration. The special adaptors and additions

A-1
*zrgandwoo 0¢/09€ pereys—suzy
ay. BuzAoTdwa waists gajowoajqoeds ssew oy} UT SqFuN eyy Fo wezzetp ¥

T aan3sty

SOON Su Trl
AG IdSIO
p LY

 

 

 

 

"GO HINWVASIY INGWNYLSNI °° 1d30 SOTLIN39

 

 

2300W XIONIS
LH9T14~- JO-SWI1
NOIiNIOS3Y YIISINT

 

 

 

 

 

 

A¥OLEYOSE
H3evISIe

NOT LE LNSWNYLSNT
SNQHd blvd

 

 

WNL I43dS

 

SIJINOLUMOSYT ASOIOTEOXS* *1d30 SIT13N39

319

 

 

T/1d WNYLISdS

 

 

[ Suasn |

  

 

 

 

 

  

 

 

  

 

 

 

      

 

 

 

 

 

 

 

       

 

 

 

 

 

 

 

 

 

 

 

 

 

  

 

 

 

  

 

 

 

 

 

 

 

 

 

 

 

 

 

 
  

 

 

 

 

 

 

 

 

 

ALISHYSAINN GYOsNULsS

 

NOTIYLNIWNY ISNT Gal .lb ON’ ddl dWONlLOddS SSUW

 

37 waaay aa]
all @88T_Y3H10 41 aT Hoda | 43 ST@T_NUOINNI 4
aBaT 2% wOLdede || NoTinqOs3a HagSINT
)

189 o “S801 AMLSHOWIISdS SSHUS "dd AMISTIHGHD

AGLS mH
as Y313WOMHITO
7GNNWHO Husa YBLNdWOD INIT a ue INIYTO
33d 07 = —| JTLANSEW

YALNGWOD BBeT SOA

can fae]

S3W4 ISIC JNOWSH2 on STOy MYOTIS
GT
nail 1334 @OST O01 aN | *| notin tossed” HaSSINT
Q3YvHS SAWIL Mines .
7WHINGI 6 Jub SSONULSTO

BML TANNGHD Fe
eyed Csr 483) et ary
OTH XOLe SUE IT) a 3 sn304° FONTS
_— L_"| NOTLN1OS3a H393NT

Q@S/@SE SONGOTTTAUINT NOIL93173S 335IA30 J3d5¥1 3h

Sod Wr sly WNNUW HLTA 7H
4812 ONY O 34dS914 3s 6-SW IS

a

 

$n303_378n00
NOITiN710S34 HSIH

 

 

“TP LNAWNYLSNT
AMS -1
Pag

Ls)

aD
wa

needed to connect the mass spectrometer, are identifiable and indeed built, as a separate
special purpose interface. Design, maintenance and even computer interchangeability are
enabled or simplified by this arrangement.

Last, but not least in requirements or expense, is a system to make this work and be
useful. The realization is best identified by the computer program written to effect
operation. The general use concept is implied in the program. We have tended to identify
these systems by their program names. In Figure 1 and Figure 2 these system names appear
in the oval boxes that tie together the operational paths.

Our programming is all in PL/1. Programs may and must be written at any of the
remote keyboards. Since there is an interactive compiler, programs may be modified,
altered and/or controlled right at the using site at any time, even concurrently with
operational use. Figure 3 shows an elementary program that would cause 100 a-to-d
conversions from a laboratory instrument and then plot these values as a graph on the
laboratory plotter.

The eight systems shown in Figure 1 connecting mass spectrometers of the time
shared net, are of three classes. The top two, BIGSPEC and METASPEC both are quite
special to the MS-9. BIGSPEC is actually a collection of three major programs for the
data acquisition and processing of high resolution runs. Element mapping has not yet
been included; typed output gives fractional mass position of the found peaks. Provision
is being made to send this data directly to other computers at Stanford engaged in the
artificial intelligence project and their organic structure elucidation programs.
METASPEC is a little more than a slide rule to aid in metastable identification.

LILATLAS and CHEMONE are data logging programs and integer peak identification.
This type of program has been well reported in che literature. With only minor changes
to accommodate the type of scan, we will use the same program on the magnetic focus, the
electrostatic quadrupole and the time-of-flight instruments.

A TIME SHARED COMPUTER INSTRUMENTED LABORATORY

 

MASS SPECTROMETER
OR_OTHER INSTRUMENTS
WITH ELECTRONIC OUTPUT

 

 

 

 

 

 

 

 

 

 

 

 

 

 

[ere PURPOSE INTERFACE | BIGTTAL
ayy
| | / [LABORATORY
|] TNSTRUNENTS
ri COMPUTER
SYSTEN
tt
ef
x ANALOG DATA DIGITAL ORTA
aon LF
AND/OR AND/OR A
SUTPUT ourPuT fo
COMPUTER DATA TIE PANEL

 

 

 

 

L_¢ A SYSTEM (PROGRAM) )———t

76 THE TINE SHARED COMPUTER |

 

 

Figure 2

The SPECTRUM series are computer control and data acquisition. SPECTRUM does not
use the features of the time shared system, only the LINC. The System was described at
last year's E~14 in Philadelphia. The SPECTRUM logic has been rewritten into PL/1 and
operates as SPECTRUM PL/1 in the time shared mode. It presently features completely
automatic calibration and a mass range from 10 to 500. The other two versions of
SPECTRUM will be nearly identical, but differ in the hardware path used. Interchange-
ability of hardware modes is made possible with the standardization of the CALL READ
statement shown in the PL/1 example.
AMS -1
Page

Provision is made to type or plot results in the laboratory. Figure 4 shows two
of the standard plot formats we employ for integer resolution data.

An information presentation problem has arisen. Profuse data acquisition and
impatience for results have led us to devise interactive data presentation methods. In
the SPECTRUM systems the users typically take 10 or 20 spectra of a solid sample, and
200 or more of a GLC run. The results are in the computer and instantly available. But
typing or plotting takes time and our users have become impatient with delays of more
than 10 or 15 minutes.

We are now working out ways of giving data abstracts to enable the user to "home in
on the spectra or data he wants. This interactive information extraction runs like this:

a. The user asks for a certain small set of data based upon his estimate of what
ia the pertinence data. If his estimate 1g vague, presumably he will ask for
a condensed set or simply indicates the set or sets.

b. The program responds with a quick presentation.

ce. The user may use this computer output to improve his estimate of what he
required. Return to a.

Thia may take the form:

a. Ask for the 8 highest peaks in all 10 spectra taken during a solid sample
run,

b. Computer types the 80 items.

ce. The user sees that the higher masses did not show up in the sets of 8 high
peaks. But he sees that peaks from spectrum 7 onwards are saturated.

d. Reasks for the 8 highest peaks from mass 200 onwards, in spectra 1 to 6.
e. Computer responds with 48 items.

f. User notes that a good fragmentation was indicated in spectrum 4.

g-. User asks for a full plot of spectrum 4,

The central time shared computer allows shared files. We are building a file
system with standard formats. Each mass spectrometer user's data may be made available
to all,

1.000 PLOTSOME:Procedure;
10.000 /* This is a PL/1 program to take 1000 «/;
11,000 /* sample points and then plot them, a/;
12.000 declare LINE2H (1000);
13.000 CALL READ(18&, LINE2H);
14.000 /* 18 addresses the laboratory Instrumentation «/3

15.000 /* terminal. */3
16.009 nO t=l1 to 1000;

17,000 CALL PLOT(76, I, LINE2H(1),2);

18,000 END;

19,000 /* 76 addresses a specific digital plotter f/3
20.000 /* tn the laboratory, a/;

21.000 END PLOTSOMF;

Figure 3

A sample program in PL/1 that would accept laboratory data and plot it
on the laboratory plotter.
tg

Oo

99

a OB
I

IT BR

The uses of this are just emerging. Perhaps the first practical use will be that
of sending samples to a different mass spectrometer laboratory. The results may be filed
by the mass spectrometer operator and the requestor may get his results directly on his
own laboratory terminals and process it in his own programs. All the described
operations used the computer in a time shared mode and concurrent with other users and
other instrumentation. No advance notice is required to operate, other than to request
a "line" from the operator. The operator does reserve the right to deny service if the
usage is too high. This normally happens only in the teletypewriter ports or when there
is some abnormal operating condition of the computer.

Observations based upon our experience with time shared instrumentation:

A time shared large computer does greatly aid in program development. The ability
to write instrumentation systems software in a high level language and with on-Line
compiler does enhance program development. Also since much system improvement will be
largely to the software, such improvements may be introduced easier than if machine code
or assembly languages had to be used.

It should be noted that much more support and computer capacity are needed for
system development than for operational uses. Certainly program changes and
documentation is easier with high level languages and hence less costly by an order of
magnitude. The large system is peculiarly suitable and capable in this initial phase.
After the algorithims are tried and proved usable and experience has been gained in the
operational use, the justification of the large computer does decrease. Operations
could be carried on largely or solely by a small computer.

Another useful aspect of powerful time shared compilers is the availability of common
functions, exponentials, log, trigometric and the like. Also the ease of floating point
and double precision does make life more enjoyable. No interaction between user and the
system is enhanced. If nothing else, the output formatting ease of high level languages
greatly aid programming for user interaction.

Ie had been our experience that small programs may be written in a tenth of the time
in the time~shared mode as compared to small computer assembly languages. However the
advantage diminished as the systems get larger and the structure of the high level
language program gets complex.

Some weaknesses that have shown up in our system:

1. The lack of overlay capability or object code stored programs or procedures
can cause uneconomic or cumbersome operation. Total systems of instrument setup, data
acquisition, data processing and/or filing, and data presentation can run to very large
proportions. Our systems run 5 to 25 thousand words for just instructions and
auxiliary working areas.

There appears to be a fundamental conflict between the incremental compiler used in
very responsive time shared programming systems and precompiled program or program
segments. In our system, programs are filed on disk as manuscript source code and
compiled upon entry into object code. Thus a program or procedure must be recompiled
each time it is brought from storage. This forces us to large programs held in core
for hours even though the use is intermittent. The penalty to avoid the delays of
recompiling is very uneconomical use of core. Our experience clearly indicates the need
to swap programs or program segments in the order of seconds.

2. Fast disk or drum data filing should be provided. Most sophisticated time
shared systems have very protective file systems that are well calculated never to
make an error. But the price paid due to redundancy, flexibility, and cross checks
is slow filing. Instrumentation data usually can pay a price of bit errors of the
order of one in 10? to 1042 if speed is gained. And there is the real time need to file
data rapidly to avoid large core storage. Again the price paid for slow filing is the
large amount of core needed to buffer real time data input.

3.. For instrument-computer interaction, fast channel turn around is desirable,
This ia the ability to do sequential operations of input, then output or vice versa.
This is not necessarily allowed with complex computer operating systems. Such systems
tend to be oriented to fast data streams in one direction as to or from disk or tape
units. If one wished to accept a datum point, and then react with an output, then
repeat, it may be found that it takes a big computer many milliseconds to change from
input to output and back. Similar difficulties may be found if complex orders of data
input/output are attempted.
AMS -]

 

 

 

 

Page 6
see a
oo Ly
oe 16
9
t é
=
a u
z o
2
6 &
~ 2 s *
sa tee ise 200
1 - HONANONE, PLOTTED fer. uk. Lee
DATA TAKEN ON 112 NAT 1869 MAGS /CHAREE neve sad venei Ty
ING TRUNENT» FINNIGAN a erm
5 - NONANONE
OATA TAKEN ON 11 MAY 1069 3

TNS TRUNENT © FINNIGAN

“3 PLOTTED rey. 11. 1868
BYAMFORO UNIVERSITY
0a. TRL

Pigure 4

4. At this state of the art large computer systems are prone to "crashes". The
complex program that makes a large computer a time shared facility will be faced with
conditions or malfunctions it cannot handle. At such times the system becomes inopera-
tive and remains so until operator intervention restores normal operation. Often data
or program extras made sometime prior to the "crash" are lost. Also the user's
programs normally have to be recompiled. Even a few moments of master computer down
time can cost the user hours of his and his laboratory's time.

Despite the limitations described, which largely are the price paid for a
developmental system, we feel that the mass spectrometer complex we have is now
reasonably current with the state of the art in mass spectrometer instrumentation. But
this is not the point of merit. The fact that the system is capable of much development
is the most exciting feature. The basic instrumentation approach, the data channels
provided, the versatile access to the computer from the laboratory, the general purpose
computer service in real time, and high degree of programmability provided, all give us
a vehicle especially suitable for continuing the enhancement of mass spectrometer-
computer-instrumentation.
Dist:

AMS -1
Page 7

INTERACTIVE INFORMATION PRESENTATION. EDITING, AND FILING

 

What data and what range
do you want presented?

i

Computer output of
abstracted or more
complete data.

 

 

 

Return to
other tasks,

 

 
  
    
 
 
   
 
 

  
 

Rephrase
request.

 

 

 

 

 

 

no Is this that which

you destre?

  

Do you wish to
file the results:

 

Please type in
COMMENTS (Noten),

 

 

 

 

Completely formatted typed
or plotted output available

 

 

 

 

Computer only from files.
Files

Figure 5

REFERENCES ~

1. W. J. Sanders, G. Breitbard, G. Wiederhold, et al,, “An Advanced Computer for
Medical Research", Fall Joint Computer Conference Proceedings, ACM, page 497, 1967.

2. J. F. Corbato, et al., "The Compatible Time-Sharing System", MIT Press, 1963.

3. R. J. Spinard, "Automation in the laboratory", Science 158 (3797), 55, 1967.

4. R. A. Hites and K. Biemann, "Computer recording and processing of low resolution
mass spectra", International Mass Spectrometry Conference, Berlin, Sept. 1967.

5. W. E. Reynolds, J. Bridges and T. Coburn, "A computer operated mass spectrometer”,
Genetics Dept., Instrum. Research Lab. Technical Report IRL 1062.

6. C. A. McDowell, Ed., Mass Spectrometry, McGraw-Hill, New York, 1963.

7. +B. Halpern, J. W. Westley, E. C. Levinthal, and J. Lederberg, "The Pasteur Probe:
An Assay for Molecular Asymmetry", Life Sciences and Space Research IV, M. Florkin
and A, Dollifus, Eds., p. 239-249, North-Holland, Amsterdam, 1967.

8. B. Halpern, J. W. Westley, I. von Wredenhagen, and J. Lederberg, “Optical Resolution
of DL amino acids by gas chromatography and masa spectrometry", Biochem. Biophys.
Res. Comm., 20, 710, 1965.

9. W. E. Reynolds, J. C. Bridges, R. B. Tucker and T. B, Coburn, “Computer control of
mass analyzers'', Conference Proceedings of Sixteenth Annual Conference on Mass
Spectrometry, ASTM Committee £-14, Pittsburgh, Pa., 1968.

Staff/All