Digital Control Computers In Analytical Chemistry

Digital Control Computers by Jack W. Frazer Lawrence Radiation Laboratory University of California Livermore, California

PRESENT UPSURGE in the use THE of digital control computers is

significantly changing the operational procedures in scientific laboratories. This is nowhere more prevalent than in analytical chemistry. Many of the leading manufacturers of scientific instruments and manufacturers of digital control computers have developed, or are in the process of developing, computer controlled systems for the laboratory. These systems generally contain a 12 to 24-bit digital control computer and interfacing hardware to perform data acquisition tasks and certain control functions for one or more analytical instruments. In some cases complete systems can be purchased including the computer, the software, and the analytical instrument (s). In addition, many scientists have developed individual systems tailored to their own particular needs. We will discuss in some detail a few of the many commercial and non-commercial computerized systems and the relative merits of several different approaches to digital computer automation. The tremendous increase in scientific and technological knowl26 A

·

ANALYTICAL CHEMISTRY

edge has resulted in a great increase in the quantity of data and the complexity of experimentation. The computer is an appropriate tool to be coupled to the scientific environment for further increasing scientific capabilities and productivity. With the advent of relatively inexpensive control computers, it has become very advantageous to have them perform our more routine laboratory functions, but this is only a first step. The real value in the use of control computers in the laboratory will be the addition of new dimensions to experimentation. For those not familiar with digital control computers, a brief description of their operational capabilities is in order. These computers are similar to their forerunners inasmuch as they contain an instruction set that permits the scientists to write programs that perform mathematical computations and that transfer information to and from such input/output (I/O) devices as teletypes, magnetic tapes, paper punches, card readers, etc. In addition, the control computer contains hardware facilities and the corresponding programming

capability to allow the scientist to easily communicate with the outside world (the chemistry apparatus) by means of relays, stepping motors, switches, analog-to-digital converters, digital-to-analog converters, counters, etc. With these and other similar control and information-transformation devices the scientist can perform the control and data-acquisition functions in his laboratory. The computer can thus be hardware interfaced and programmed to perform many routine control functions such as sample introduction for a gas chromatograph, generation of special electrical functions to "drive" instruments, measuring and adjusting pressures and temperatures of a system, opening and closing valves, adjusting a magnetic field, etc. The performance of such control functions and the necessary data-acquisition, together with the computational power of the computer, result in an increased through-put per analytical instrument. This versatility also gives the chemist a capability of performing new and different experiments not possible without the computer.

REPORT FOR ANALYTICAL CHEMISTS

In Analytical Chemistry The actual automation of instruments has followed three general courses. Computers were first used as data-acquisition devices. Second, they were used for data-acquisition and on-line processing of the data acquired. This allowed the scientist to immediately evaluate the results of the experiment. He could then make any necessary adjustments of experimental parameters before continuing the experiment. Finally, computers are being simultaneously used for data-acquisition, on-line computations, and control of the instrument. Advantages of Automation

If sufficient care is used in programming, the digital computer along with its associated interface, permits greater accuracy and versatility in signal analysis. Although accuracy is obviously limited by the instrument and subsequent digitization of the instrument signals, operations performed on the digitized data can be accomplished so that no inherent instrument accuracy is lost. This is not generally true in analog systems. These digital techniques

readily lend themselves to mathematical treatment whereby spectral data are smoothed and reproduced to the inherent capabilities of the instrument, i.e., the quality of the analysis is not limited by the accuracy of the output device (e.g., a strip chart recorder). Techniques for signal-to-noise enhancement can be applied to increase the apparent sensitivity of the instrument. Also where applicable, various techniques from information theory can be used on-line to improve data accuracy and control functions. Accuracy is also improved, inasmuch as human error is removed from the performance of the repetitive-routine operations. When the proper algorithm is correctly programmed for the computer, all functions and data-acquisition are performed exactly the same every time they are required. The accuracy and precision will always be maximum for that system. The computer will not transpose numbers or "forget" as a scientist often does. Another benefit of computer automation is the opening of new experimental possibilities. At present the chief emphasis is still

on data-acquisition by various means; time-sequencing of instruments, real-time sharing of the computer by several instruments or via a computer per laboratory instrument. However, there is a very marked increase of interest and activity in the control aspect of computers. The control functions required are different from instrument to instrument, and in many cases are still not fully understood or recognized. In addition, new and better interfaces and transducers will be required as we learn more of the sophisticated science of control. The ability to perform dataacquisition and fully control the instrument or experiment, together with on-line computations, allows the experimenter to "close-theloop." That is, the direction of an experiment or analytical measurement can be changed at any time based on all the information obtained up to that instant. The science of closed-loop experiments will be a major area of investigation in the near future. I t should be noted that closed-loop experiments will be of two general classes: one where the algorithm is completely defined and proVOL. 40, NO. 8, JULY 1968

·

27 A


grammed in such a manner that the scientist can interact with the system only at the beginning, or at certain discrete times during the experiment. The second class of closed-loop experiments will allow the scientist to interact with the algorithm at any time. Using his knowledge, judgment, and intuition, he will be able to interact with the software program and change the direction of the experiment essentially at will. Some experiments of this type have already been performed. It is obvious from the above discussion that a properly automated laboratory will take much of the drudgery out of analytical chemistry. In addition, the analytical chemist should have more free time to spend on research and development projects. It is my firm belief that automation of the scientific laboratory will not decrease the number of scientists required but rather accelerate the rate of scientific advancement. Methods of Automation

Automation of the scientific laboratory by means of digital computers can take one of three general directions or any combination thereof. Data can be accumulated by any of several methods and at some later time, data reduction and computations can be performed by a central computer. This technique is sometimes referred to as "batch processing." Such a system by itself severely limits the versat i l e and dimensions of laboratory automation. However, it is highly desirable as support to an "on-line" control and data acquisition system. Batch processing is of greatest value when applied to lengthy mathematical calculations required for many theoretical and experimental problems. These include the inversion of large matrices, regression fits of mass spectrometer data, solution of Fourier transforms, etc. The batch processing technique, when it constitutes an integral part of an on-line control and data acquisition system has real merit. Several companies use a variation of this approach in certain of their systems. 28 A

·


The experimental scientist needs, in addition to computations, the control capabilities of the computer and the versatility it offers in the data acquisition domain. These latter desirable features can best be obtained by either of two methods: one computer per instrument or experiment, or a somewhat larger computer operating in a timeshared environment. The direction any given laboratory pursues should be determined by the tasks to be performed, and the temperament of the scientists. In no case should the more dynamic scientist be unnecessarily restricted by the computer system chosen. Time-shared systems can most easily be implemented where there are a number of relatively routine control and data acquisition functions to be performed. Notably they have been applied to such tasks as the automation of gas chromatography. Several timeshared systems and their relative merits, are described later in this paper. These time-shared systems often operate in a sequential mode. That is, the computer performs the data acquisition functions by sequencing from one unit to the next in an orderly predetermined priority. This is very reasonable when like systems are being controlled. However, when systems having entirely different time response requirements are timeshared, sequential operation can lead to improper control and a loss of valuable output information. Under such circumstances it is desirable to provide a software executive monitor that allows all instruments to run asynchronously. This software package will be expensive. Time-shared systems have certain advantages. First, the computer can often be more efficiently used when servicing many instruments. Instead of waiting for an "event" to occur on a single instrument, the computer can service another instrument which is ready. Second, for a given amount of money per laboratory instrument, more computer and I/O capability can often be obtained by means of a time-shared system. Also, the faster computer used for time-sharing usually has a larger word size

and instruction set, which can be advantageous. The alternative to time-sharing is the smaller stand-alone computer assigned to one task at a time. This approach has several inherent features that recommend it for many research applications. First, it allows the scientist complete freedom of action within the computer's limitations. He does not need to worry about any of the vulgarities of the executive monitor. This means several things: one, timing errors can usually be minimized; two, he need not worry about any interactions with other programs; three, his programs can be written in any format compatible to his machine and scientific problem. In addition, with a computer of his own, he never finds it necessary to wait because other experiments have the machine loaded. A last but certainly not insignificant feature is the lack of "crosstalk." In newly created time-shared systems, noise produced by one instrument or associated hardware interface will often result in changing the control function or affecting the data of another dissimilar instrument. When making the final decision concerning the approach to be taken, three very crucial factors can be overlooked. First, if a timeshared system is to be effectively incorporated into a laboratory, the day-to-day decisions must be made by a knowledgable, "strongwilled" scientist, not by a committee. Second, the cost of automation is a rapidly changing phenomenon. The creation of sophisticated software is an expensive proposition. However, the cost of hardware components is rapidly decreasing. Therefore, examine carefully the possible hardware/software tradeoffs and their ability to answer to your needs. Third, the system chosen should provide a maximum of freedom for the imaginative scientist. Any system that severely limits the creative, hardworking scientist is to be avoided. Software

The synthesis of application programs (software) to perform the experiment can be approached in


After 3,000 Hours of Use, This Water Bath Shaker Might Need a Drop of Oil.

The AO Water Bath Shaker was built to give trouble-free service. And it has in hundreds of laboratories. Sturdy, leak-proof construction features a large volume reservoir made of cast aluminum and coated with Teflon. A ball-bearing eccentric drive and nylon rollers for rack support assure a smooth, quiet shaking action. You can use this versatile workhorse for a wide variety of reactions requiring either agitation and/or incubation. Temperature range is 25°C to 120°C, with temperature controlled to within ±0.5°C. Choose any shaking speed from 50 to 200 cycles per second. Stroke length is adjustable at three different settings. And there's a wide variety of trays available to hold v i r t u a l l y any flask, beaker or test tube. It isn't often you find such a dependable, well-proven instrument. For literature and ordering information, contact your distributor or write to us.

AO INSTRUMENT COMPANY Analytical Instrument Division 200 South Canard Blvd. · Richmond, California, U.S.A. 94804 A Division of American Optical Corporation

Circle No. 13 on Readers' Service Card 30 A

·


several ways. Consider first only the language used, available core memory, and timing problems involved. The language used can be anything from machine language to a higher level language such as Fortran IV. As a general rule, the higher level languages tend to use more core memory per given function. In addition the resulting program is more slowly executed and timing requirements are more difficult to understand. Therefore, programming should be done in machine language when timing problems are prevalent, and/or core memory capacity is marginal as compared to the task to be performed. In general the higher level languages are not applicable for realtime control and data acquisition problems. However, should the experiment not require the full capacity or capability of the computer, it is often reasonable to write the computations and data reduction programs in a higher-level language, if this saves programming time. Under such circumstances, and where auxiliary memory such as a drum or disk file is available, the more bulky program written in higher-level languages can be swapped between core memory and the auxiliary memory as needed. This present dilemma will gradually change as higher level languages, which answer to the control and data-acquisition needs, are developed. As mentioned above, real-time control and data acquisition can generally be programmed only in machine language. This is not a very great disadvantage. Machine language programming is no more difficult to learn than some of the higher-level languages. In general, its use in the laboratory environment allows you to minimize; timing errors, computation time and total number of computer words required per given task. There are scientists, such as myself, who have never programmed in anything but machine languages. A further important consideration concerning control programming for laboratory experiments is that of personnel. On the surface

it would seem that professional programmers would be ideally suited to this task. This is usually not the case. Professional programmers are in general trained in use of higher-level language programming for mathematical computations. Others who do work with the machine language are involved in writing compilers, assemblers, and timesharing control systems dedicated to computational or bookkeeping functions. It is usually very difficult for these people to write control and data-acquisition programs for laboratory experiments unless a complete flow chart is furnished. It should be remembered that most of these people have never been involved in laboratory experimental operations. Spectrometers, vacuum systems, distillation columns, chromatographs, etc., and associated interface hardware are not equipment with which they are intimately familiar. Why not then have the experimental scientist furnish a complete and comprehensive flow chart so that professional programmers can then easily write the required software? This can obviously be done, but has several pitfalls. First, the synthesis of a complete flow chart is generally a large fraction of the work involved in developing the program. Inevitably some large or small function of the control, data acquisition, or feed-back aspect will be omitted or ill-defined. Also, with the passing of time experiments and equipment change and as a result software changes are required. The implementation of software changes is difficult when the experimenter does not understand the software and the programmer does not understand the experiment and hardware. The production of suitable software is generally the most difficult and time-consuming aspect of laboratory automation. Therefore, it should be approached in a manner that most nearly assures effective utilization of the scientist's time. Experience has generally shown that for scientific control applications it is better for the motivated scientist to write his own programs. As high-level real-time languages are developed, it is possible that


the person who is only interested in computer programming will assume more of the responsibility for control programming. The creation of appropriate software is very expensive: possibly more expensive than all of the associated hardware. For this reason alone it is often appealing to have a small computer (such as Digital Equipment Corporation's P D P 8/1, or Varian's 620-1, etc.) connected to a single instrument. Programming is then limited to the exact needs of the single instrument or experiment. The need for a monitor to provide time-sharing capabilities in a real-time domain is eliminated. Also eliminated are the hardware and software interaction problems occurring between different users. For the time-shared systems there is needed a monitor (software program) to administer the use of the computer ; herein programming problems balloon. It would be handy if each applications program were self-contained, and did not need to converse with other pro-

grams, but this is usually incompatible with the amount of core memory available and with the time-sharing concept. I/O conversion, disk storage, and magnetic tape handling are examples of large programs that many users will need to use in common. These programs with their associated buffer areas may take up a large percentage of core memory, and it would therefore be impossible to allow each client to duplicate this amount of code. Besides, the clients must take turns at the common I/O equipment and this in itself implies a monitor program which administrates the use of equipment, queues, and buffer areas. There is one important question to be asked concerning monitors for small time-shared systems: What is the optimum way to structure a monitor to allow execution of all required functions "on-time" and also allow maximum usage of the computer? Under real-time conditions it is only necessary that each user feel as if he has the entire use of the computer. That is,

Figure 1.

the individual user should receive service at a rate compatible with his requirements. In many laboratories this is not a major difficulty. The problems arise when, for instance, several large programs must each have, in a relatively short period of time, a turn in a small core memory. Similar problems can arise from a need to service, on a time-shared basis, several instruments requiring high speed data-acquisition rates and simultaneously requiring data-reduction and storage of the information acquired. These are only two examples of the types of control and computational combinations that complicate time-shared implementation. The monitor should be kept as simple as is consistent with the performance requirements. Everybody will approach the problem in a different way. In the Analytical Chemistry Section at the Lawrence Radiation Laboratory we have set the following ground rules. The only priority tasks are those that must be taken care of immediately

Diagram of on-line NMR spectrometry system VOL. 40, NO. 8, JULY 1908

·

31 A


Recording Electrobalance®

Registrierende Elektrowaage

Electrobalance Enregistreuse

leaift^ff a3ue|eqoj)38|i Suipjoaaa

• • • • •

in order to clear a computer inter rupt. H a r d w a r e priority interrupts are not used. All applications ser vices are placed in a job queue with no priority. T h e only exception to this rule is experiments which have very fast data rates and must be serviced immediately after a clock interrupt. All systems are run asynchronously and serviced as re quired. Timing is provided by a 1kHz clock located in the com puter. The application programs actually perform their own I / O to the de\'ice interfaces for control and data acquisition. Only for the higher data rates does the monitor actually perform data-acquisition tasks. This monitor not only con trols the m a n y time-shared experi ments, but has facilities for up to fourteen teletypes located through out the laboratory. I t is desirable to keep to a mini mum the information transferred back and forth between the monitor and the various applications and I / O programs without making the monitor to general purpose. To illustrate this point a few of the simple questions t h a t must be an swered in the construction of a

monitor arc listed below : • How does the monitor know which program to direct a message

to? • How m a n y characters will the monitor accept before deciding to convert the incoming teletype code into a) character, b) octal data, c) decimal data, and how does it know which of these conversions to per form ? • Where is the d a t a stored so the client can retrieve it as needed? • How is the client told t h a t data are ready? • When and where docs the client regain control of the computer in order to process d a t a ? • How does the client tell the monitor t h a t data is expected? • W h a t happens if someone leans on a keyboard of one of the tele types and extraneous data arc input? I t is readily apparent t h a t m a n y questions must be answered in order to provide a time-shared monitor t h a t answers the needs of all clients. One programmer should be in charge of the monitor. This "keeper of the monitor" must provide simple and concise instructions to all other

Thermogravimetric Analysis Adsorption Studies Vacuum Studies Magnetic Susceptibility Force & Torque

Phone or Write for details

CAHN

D ιν ιsια Μ

VENTRON

ΙNSΤRUMΕNΤS CORP

7500 Jefferson St.. Paramount

Calif

90723

213' 634 1935 27 Essex Rd . Dartford. Kent. U.K.. Ph. 21540

Figure 2. Automated NMR spectrometer Circle N D . 97 on Readers Service Card

32 A

·


REPORT

programmers defining the methods used to cause the monitor to perform the desired functions.

all new & all automatic...

Applications

The application of digital control computers to laboratory automation is proceeding along many paths. This can best be illustrated by describing some of the recent commercial and noncommercial applications. Those described below are not intended to represent a comprehensive list. Rather, they illustrate several different approaches and are systems that are most familiar to the author. Varian Associates has under development several automation systems. One system pursues the concept of one small computer per instrument. Figure 1 is a block diagram of the engineering prototype system as interfaced to an NMR spectrometer and shown in March 1968 at the Pittsburgh Conference on Analytical Chemistry and Applied Spectroscopy. The central processor used for this system is the 620/i computer. It is a 16-bit machine having a memory reference cycle time of 1.8 microseconds and 4K words of core memory. Although many options (such as hardware multiply-divide, magnetic disk, priority interrupts, etc.) are available, none are required for control and data acquisition of the NMR spectrometer. For ease of operation and conservation of core memory, there has been considerable trade-off between hardware and software. Figure 2 is a photo of this system, consisting of an A-60 NMR spectrometer, teletype, and the 620/i computer plus the control console. The control console allows the scientist to select one at a time, many different operations and enter the desired parameters. As a simple example, he may want signalto-noise enhancement by means of multiple scans. He first selects the operation "multi-scan average." Next for "parameter" he may want to enter 15 for the "number of scans," etc. The many operations and corresponding parameters necessary for NMR analysis are selected and defined via this control

the Shimadzu AQV-50 digital spectrophotometer ideal for multiple analyses a n d column monitoring This versatile unit, w h i c h measures in the w i d e range from far u.v. to near infrared (183 to 1200 m y ) , now offers these new, exclusive automatic features: A u t o m a t i c zero and 100% setting w h e n standard sample is put in the optical path. A u t o m a t i c digital display with data presented in % T or a b s o r b a n c e . This data can further be fed into a printer, recorder, or a tape puncher t h r o u g h the use of optional output driver. A u t o m a t i c sample changer for up to 50 samples. Photometric A c c u r a c y : ± 0 . 2 % T Reproducibility: ± 0 . 1 % T standard deviation

DIGITAL READOUT Accessories available for reflectance studies, fluorescence, etc. FREE BOOKLET describes the accessories, operating features and comparative advantages of the versatile AQV-50. For your copy, write:

COSMOS SCIENTIFIC CO. 921 Bergen Ave., Jersey City, N.J. 07306 · (201) 925-1101 Stocking S Servicing Dealers: Ace Scientific Supply Co. Preiser Scientific Co. Charleston, West Virginia Linden, New Jersey American Instrument Co., Inc. Physicians & Hospital Supply Co. Silver Spring, Maryland Minneapolis, Minnesota Circle No. 34 on Readers' Service Card

VOL. 40, NO. 8, JULY 1968

·

33 A


console. The computer then con trols the NMR spectrometer, per forms the requested data acquisi tion tasks, drives the recorder to output the raw data (if requested) and last outputs the processed spec tra after performing the required computations. A Teletype could have been used exclusively for input of operations and parameters; however, this would have required a large section of core memory and comparatively speaking would have been much slower and awkward to use. As de signed, the monitor has merely to check the switch and if the switch is on, the program branches to an appropriate subroutine. This sys tem is actually easier to use than the analytical instrument alone. Control is provided for several functions. First, the computer (via software and the hardware inter face) sets the current in the Y gra dient and curvature shim coils in such a way as to optimize the height of a selected reference peak. This function can be performed be fore every run or every Ν runs as required. The program accepts starting parameter instructions from the control console. These in clude the scan start point in Hz, the range (sweep width) in Hz, and

sweep time in seconds. During the scan 1024 data points are stored in core memory. When requested, the raw data are simultaneously output under computer control on an x-y recorder. The software for the NMR spec trometer includes provisions for scanning, programmed (variable speed) scan, base line drift correc tion, integrations, digital signal smoothing, resolution enhancement, peak detection, and spin simula tions. In general, this system's ap proach integrates data processing with instrument operation to allow the user to optimize spectrometer tuning and to process the spectrum signal so the user can obtain per tinent information immediately. Another Varian approach to au tomation is a time-sequencing sys tem in which like instruments are interfaced to a 620/i. At present, an engineering prototype is inter faced to 10 gas chromatographs. For this particular application, the computer performs only data acqui sition and data processing tasks. Every 10 msec the computer se quentially steps to the next gas chromatograph and "reads" the de tector signal. The resulting chromatogram is processed and the per tinent information is extracted and

Figure 3 . 34 A

·


Cycle of events a l o n g one c h a n n e l

output to the teletype by the appli cation program. Digital Equipment Corporation has a somewhat similar system. It uses a PDP-8/I computer (12-bit word, 1.5 microsecond memory ref erence time) that can be interfaced to 20 or more gas chromatographs for data acquisition and data pro cessing. Again, this is a time-se quencing system. The maximum data rate capability of the system is 240 points per second. Sampling rates for individual chromatographs can be set at 3.75, 7.5, 15, 30, or 60 points per second. Hardware op tions include four relay drivers for each gas chromatograph. These can be used for control functions. This larger system requires 8K words of core memory, fast mul tiply/divide hardware, and a 32K disk file as well as the associated interface. Both systems provide reasonably comprehensive software for data reduction. There are many control comput ers being used in biomedical and clinical laboratories. As an ex ample, Stanford University uses PDP-8's in the study of animal be havior. As in most areas the com puter is still used largely for data acquisition. The AutoChemist de veloped by AGA Corp., Lidingô,

REPORT FOR ANALYTICAL CHEMICAL

Sweden, is a fine clinical example of data logging by means of a control computer. The AutoChemist was developed for large-scale blood analyses. I t has a capacity of approximately 135 samples per hour. In turn, up to 24 determinations can be made per hour per sample. This is approximately 3000 determinations per hour (when provisions are made for blanks and standards) and is equivalent to the work load of at least 100 qualified technicians using conventional methods. A simplified processing flow schematic is shown in Figure 3. The mechanical trains, reagent and washing stations, photometer measuring heads, etc., are shown in Figure 4. Mechanical control of this unit is via an electrical-mechanical lineage. Any drive train can be stopped within one tenth of a second, thus allowing precise positioning for the addition of reagents, etc. The computer's only task is to data log from the 24 photometers, then process and output the data in a form acceptable to the medical profession. The system has a guaranteed precision of < 1.0% relative standard deviation per analytical channel. Dr. Charles Sederholm of IBM

Corporation has taken a different approach. He believes that in general, a proper time-shared computer system offers most experimentalists more capability per instrument per unit cost. He has accordingly built a time-shared system. I t is designed around IBM's 1800 control computer and I / O equipment. They have taken the viewpoint that "most experiments or instruments in the laboratories to be automated have a reasonably low duty cycle; have reasonably large systems requirements as far as I / O devices, peripheral storage devices, and computational capabilities are concerned, and have turn-around time requirements on the order of tens of milliseconds." Around this concept they have designed some special interfacing capabilities and written an extensive software monitor for their time-shared system. The heart of this system is a digital multiplexer having thirty-two subchannels. Most instruments are interfaced through one or more of the subchannels. The multiplexer is initialized by the applications/ monitor program. Thereafter, data are transmitted between the instrument or I / O device and the core of the computer in a demand/response

AN ADVANCED OSMOMETER ISFITTOBETRIED

JUST ASK

us

We'll prove it. Advanced Osmometers are the most reliable Instruments you can buy to determine the osmolality of solutions with greatest accuracy. Eight models are available — each designed for varying laboratory requirements. Read about them. Then try one in your own laboratory and see for yourself. Send the coupon today for a free brochure and/or a trial.

ADVANCED INSTRUMENTS, INC. 44 Kenneth Street / 617 DEcatur 2-8200 Newton Highlands, Massachusetts, 02161

Mail t o : Advanced Instruments, Inc., 44 Kenneth Street, Newton Highlands, Massachusetts, 02161 •

Please send the Free brochure.

Osmometer

•

Please arrange an Osmometer demonstration in my laboratory.

NAME POSITION ORGANIZATION ADDRESS CITY TELEPHONE

Figure 4.

STATE TAP

Inside view of the conveyor system-central chemical processor Circle No. 11 on Readers' Service Card

VOL. 40, NO. 8, JULY 1968

·

35 A

new ultra superior purity lets you see the past more clearly

Only Ultra Carbon offers four levels of purity to provide conditions needed for standard to critical evaluation. When your analysis requirements call for a low metallic blank level and samples are restricted to mgms.— whether it be for archaeological samples or aircraft engine c o n t a m i n a n t s Ultra Carbon's new Ultra Superior Purity graphite can provide the sensitivity and precision needed. A n d for less critical evaluations, Ultra Carbon supplies three other purity levels: Ultra " F " , the famed purity level that has led the industry for 25 years; SpectroTech, an intermediate purity level; and Standard, for routine metallurgical analysis. Write for more detailed descriptions of any of these products. Ultra Carbon Corporation, P.O. Box 747, Bay City, Michigan 48706. Ultra " F " , Specfro-Tech and Ultra are Registered Trade Marks of Ultra Carbon Corporation,

ULTRA CARBON

See ACS Laboratory Guide for All Products/Sales Office Circle No. 79 on Readers' Service Card 36 A

·



mode. That is, the data are transferred when the device says that new data are needed, rather than when the computer decides that it is time to transmit data. This demand/response data transfer facility adds flexibility in that data can be acquired at equal increments of the independent variable(s) of the instrument as well as at equal increments of time. When required, each instrument has its own hardware clock for timing purposes. Associated with each subchannel of the multiplexer is an address register and word count register. The multiplexer "cycle-steals" for the transmission of information to and from core memory. Three memory-reference cycles (eight microseconds total time) are required per word transmitted. One cycle increments the core memory address being accessed, one transmits data, and the third decrements the word count register. The address register together with the word count register define the core locations used for data acquisition or storage of the control function data. Double buffers are used so that data can more easily be transferred between the core memory and the special purpose multiplexer, and between core memory and the disk file. The programming language available is a macro assembly language. Many high-level macros simulating Fortran statements are available. In addition, many general purpose, low-level macros are available for operating the device interfaces. The monitor provides the necessary isolation for the individual applications programs. This allows each applications program to be written without regard to any other applications program. Variable core is dynamically assigned to applications programs as needed in pages of 512 words. When an applications program terminates execution, its pages of core are returned to the system. All of the software processing associated with running applications programs out of noncontiguous sections of core is performed during the loading process, rather than at execution time.

Loading of a program (from the disk) is considered a background, non-real-time process, whereas execution of a loaded program is realtime. All real-time programs presently in execution are given five millisecond time slicers. With a maximum of five real-time programs concurrently in execution, and a maximum system overhead of 100% associated with answering hardware interrupts and starting I/O operations, this assures each real-time user's program a turnaround time of 50 milliseconds. Non-real-time application programs may be running in a middleground mode at a lower priority than the real-time programs. Middleground programs receive 100 millisecond time slices when no real-time programs require computer facilities. Housekeeping for the I/O devices and loading of programs are considered background operations. Remember that all control and data-acquisition functions can be controlled by the individual devices and performed through the multiplexer via cyclestealing. If the above assumptions are correct and if each program in execution requires a five millisecond time slice every 50 milliseconds, this system is capable of handling five instruments (NMR spectrometer, mass spectrometer, etc.) simultaneously. Should each of the individual instruments require less than 5% of the capability of the 1800 computer, it may be possible to simultaneously have more than five instruments in operation. Such a system requires that portions of the monitor be re-written. This is a sizeable system. As demonstrated at the March 1968 Pittsburgh Conference on Analytical Chemistry and Applied Spectroscopy, it consisted of IBM 1800 computer with 32K words of core memory, a 512K sixteen-bit data word disk, a multiplexer with 32 sub channels, line printer, x-y plotter, card reader, teletype, and four analytical instruments. These instruments were a Varian A60 NMR spectrometer, M66 mass spectrometer, and two Varian Aerograph gas chromatographs. The monitor occupied 13K words of core memory.

CELL RECONDITIONING SERVICE CELL RECONDITIONING SERVICE

REPORT This system, as shown at the con ference, was developed jointly by IBM Corporation and Varian Asso ciates. Varian also provides the interfaces and analytical instru ments. Dr. Sederholm's approach has been one of trying to provide the user with a system that he, the user, can use for the solution of control and data-acquisition problems. He has provided one interesting ap proach. In the Analytical Chemistry Sec tion at the Lawrence Radiation Laboratory, we have taken still a different approach. We have a PDP-7 computer with 8K words of core memory operating in a real time time-shared mode. This com puter is presently interfaced to five analytical instruments and the sixth instrument is being interfaced. All six instruments are different. The computer performs the re quired data acquisition and control functions. The first object of this system was the automation of many rou tine determinations and analyses. This phase of the effort could be reasonably well defined and the boundary conditions established. Since high speed data rates (e.g., above approximately 1kHz for any one instrument) were not required, all timing is performed from a 1kHz clock located in the com puter. This clock interrupts the program every 1 millisecond, or in teger value thereof, in order to al low the monitor to interrogate the software clocks of each applications program and thus determine the need-for-service. All systems are run asynchronously and serviced as required. The individual applica tions programs perform their own data acquisitions functions except when higher data rates (greater than 300 Hz) are required. When data from a single instrument is needed more often than approxi mately every 3 milliseconds, the monitor performs the data acquisi tion at the time of the interrupt, thus minimizing timing errors. Time-slicing as such is not used. All control, data acquisitions, and I / O functions are performed as re quired and all remaining time is given to computations. Waiting for

CELL RECONDITIONING

Ul

m

υ

iω

m Ο ζ ζ

ο Ρ 5 ζ ο ο U

κ _ι -ι

Id

υ

? ô πι

ATTENTION INFRA-RED

Ο

m r r

S Ο Ζ ο

SPECTROSCOPISTS HERE'S HOW M O T

H

δ ζ ζ

T0 D 0 IT

η (Λ

YOURSELF

π

111

υ

> κ

U (Α

π

? Ο

m ο η ι39 m ο Ο Ζ σ

INTERNATIONAL CRYSTAL LABORATORIES

104. The pressure transducer and gas buret are contained in an oven regulated to better than 0.01 °C. A glass frit is inserted between the Toepler pump and the buret to provide an accurately reproducible gas buret that can be set automatically. On the last gas transfer step (pumping), the mercury in the

Toepler pump is forced up against the glass frit, thus providing a very reproducible buret volume. The Toepler pump is also controlled by the computer. In summary, this system is completely automated. As we gather more information and further refine our computer programs, we expect to routinely obtain determinations having an accuracy of ±0.03 per cent. The vacuum fusion apparatus is equally well controlled. The small mass spectrometer that is attached to the system is under full computer control as are the valves, sample dropper, and MKS pressure transducer. We have attempted, insofar as is practical, to completely control all apparatuses interfaced to the PDP-7 computer. We anticipate eventually having eight to ten different analytical instruments simultaneously operating asynchronously in real-time. All programs including the monitor will reside in the 8K words of core memory. The disk will be used for bulk storage of mass spectrometer data, hardware debugging programs, assemblers, etc. For this task all boundaries can be defined. Therefore, fixed point arithmetic is used for all computations. The usual approach is to use floating-point routines ; however, these are very large programs that are slowly executed. Where limits can be defined, all arithmetic should be performed in fixed point and scaled to the specific application. It is readily apparent from the foregoing that there are many approaches to analytical automation. This is a new and exploding field of experimentation. In all likelihood several "best" methods will be developed to answer to most of the automation needs in the scientific laboratories. The number of control computers is rapidly on the increase. It is going to be mandatory that we as scientists make effective use of these tools.

TODAY ITS FREE RINSING

Sud'n

for BRUSH and ULTRA-SONIC Washing of all laboratory glassware, missile and outer space components. . .used wherever the ULTIMATE in cleaning is ESSENTIAL —where any thing less would be too EXPENSIVE!

CLEANS completely— RINSES freely— without film or residue even in COLD WATER! SPECIFY SUD'N See your laboratory supplier or mail coupon for sample.

SEP-KO CHEMICALS, INC. 3 9 0 0 Jackson SI., N.E. Minneapolis, Minnesota O K send f r e e test p a c k a g e o f I I SUD'N f o r hand washing • LAB-GLO Mechanical cleaning

Conclusions

We have tried to illustrate several different approaches to automation in the analytical chemistry laboratory. In conclusion, two

Name Address City

State Circle No. 74 on Readers' Service Card

V O L 4 0 , N O . 8 , JULY 1 9 6 8

·

39

A

flyingspot


photometer

densitometer or scintillation counter

The new range of EMI photomultipliers with "SUPER" S-ll photocathodes will enhance your project performance. High quantum efficiency, (23/24%) high gain at relatively low overall voltage, and low dark current at the rated over all sensitivity are typical of these types. They maintain the EMI standard of excellent gain stability and linearity. The narrow spread in characteristics makes these types ideal for sys tems or for multiple installations. The table below gives the typical values for the signifi cant parameters. Anode Dark Amps/ volts/ Current Lumen Overall Nanoamps 50 1150 2 50 1250 5

Type No. 9656R 9708R

Dia. 2" 3" 3.5"

9531R

200

1300

25

4"

9732R

50

1250

10

5"

9709R

50

1350

15

Note that the anode dark current is given for the overall voltage at the specified overall sensitivity. The maximum overall sensitivity is 10 times the values given above. Each Tube is individually cali brated and data is supplied with the tube.

points should be re-emphasized. First, automation is a systems prob lem. T h a t is, it includes not only the problems associated with the computer and hardware interfaces but also the software and the ana lytical instrument (s). For m a n y applications special transducers will need to be developed before the au tomation can be successfully com pleted. I n other instances, the in strument itself will require exten sive modifications. As an example, it is usually more effective to elimi nate noise a t the "source" t h a n to a t t e m p t to improve the apparent signal by means of extensive m a t h e matical manipulation. Automation should be ap proached in a manner t h a t most ef fectively solves your problem. Our approach as stated above has been to use no priority interrupts. I n stead, all timing is performed via the internal 1kHz clock. How ever, in the near future we will be interfacing a high resolution mass spectrometer to our P D P - 7 timeshared system. At t h a t time we will provide an appropriate inter face whereby we can cycle-steal for high speed data acquisition. Since this will be the only instrument re

quiring service more often t h a n every msec, it and only it, will be allowed to cycle-steal. I n short, we have attempted to solve our prob lems by treating each instrument separately, not by forcing all in struments to operate in the same mode. For our laboratory this has been the most effective solution. T h e other point to be re-empha sized is the software requirements. The best hardware computer system is useless if good software is not available. Software cost will gen erally be as great or greater t h a n the hardware costs. As much imag ination and creativity is required in the development of adequate soft ware as in any other p a r t of the au tomated system. The systems described above in no way represent all the approaches to automation. M a n y scientists and manufacturers are developing similar a n d / o r different automation concepts. The systems described were used only to illustrate different methods of automation. T h e same generalization applies to the com puters used in these systems. M a n y other different computers can and are performing similar tasks in other systems.

JACK W. FRAZER is in charge of ana lytical chemistry operations at Law rence Radiation Laboratory, Livermore, California. He was born in 1924 in Forest Grove, Oregon, and served as a pilot with the Army Air Force during World War II. He re ceived a B.S. in chemistry from Hardin-Simmons University in 1948, and was a chemist with Los Alamos Scientific Laboratory 1948—53. Mr. Frazer has been at Lawrence Radiation Laboratory since 1953. His principal interests are computer automation in analytical chemistry and development of analytical methods involving gas analyses and inorganic syntheses.

Send for our new 64 page catalog giving data and technical information on the complete range of EMI photomultipliers.

W

C

hittaker O

R

P

O

R

A

T

I

O

N

G E N C O M DIVISION βΟ Express St., Plainview, L.I., N.V. 516-433-5900 Circle No. 145 on Readers' Service Card 40 A

·


Circle No. 2 on Readers' Service Card

>-

Digital Control Computers In Analytical Chemistry

Recommend Documents