Application of a Microcomputer-Based System to Control and Monitor ...

APPLIED

AND

ENVIRONMENTAL MICROBIOLOGY, Feb. 1984,

0099-2240/84/020239-06$02.00/0 Copyright © 1984, American Society for Microbiology

p.

239-244

Vol. 47, No. 2

Application of a Microcomputer-Based System Monitor Bacterial Growth

to Control and

JEFFREY A. TITUS,t GREGORY W. LULI,t MICHAEL L. DEKLEVA, AND WILLIAM R. STROHL* Department of Microbiology, The Ohio State University, Columbus, Ohio 43210

Received 12 September 1983/Accepted 7 November 1983

A modular microcomputer-based system was developed to control and monitor various modes of bacterial growth. The control system was composed of an Apple II Plus microcomputer with 64-kilobyte random-access memory; a Cyborg ISAAC model 91A multichannel analog-to-digital and digital-to-analog converter; paired MRR-1 pH, P02, and foam control units; and in-house-designed relay, servo control, and turbidimetry systems. To demonstrate the flexibility of the system, we grew bacteria under various computer-controlled and monitored modes of growth, including batch, turbidostat, and chemostat systems. The Apple-ISAAC system was programmed in Labsoft BASIC (extended Applesoft) with an average control program using ca. 6 to 8 kilobytes of memory and up to 30 kilobytes for datum arrays. This modular microcomputer-based control system was easily coupled to laboratory scale fermentors for a variety of fermentations.

A general philosophy for datum acquisition and analysis from computer-controlled fermentation systems was developed ca. 10 years ago (10). Since then, the use of computers for datum acquisition and control of fermentation processes has increased rapidly, particularly in industry (2, 5). Recently, several research (7, 9, 11, 12) and review (2, 3, 5, 13) articles have appeared which describe various computerassisted fermentation control systems. Most of these systems have involved minicomputers such as the Digital Electronics Corp. PDP 11 series (2, 3, 5, 11) or combined microand minicomputer hierarchical systems (5, 12). These are excellent for industry or laboratories with large budgets but offer little to most university facilities with limited space, budgets, and electrical engineering expertise. The use of microcomputers to control fermentation processes was first suggested at the Dijon Conference in 1973, although many doubts about their applicability were expressed (6). Hampel (3) stated in 1979 that computer control of fermentations in biology laboratories is very limited because biologists lack expertise in advanced programming skills and electrical engineering. Also in 1979, Jefferies et al. (6) developed a single-circuit board microcomputer for control of a fermentor, exemplifying the potential simplicity possible for microcomputer fermentation control. Since that time, commercial microcomputers have become reasonably priced, increasingly popular, and more useful in microbiological laboratories. Moreover, many microbiologists are now mastering BASIC computer language, a higher-level language which is relatively easy to learn. These factors, coupled with the recent increase in availability of less expensive analog-to-digital (A/D) and D/A converters on the market, should make it possible for more microbiologists to utilize microcomputers to control various processes, including fermentation. Hatch (5) recently stated that "the use of microcomputers for control of instruments and experiments is now receiving intense scrutiny." The use of microcomputers in fermentation processes has been usually limited to combined micro- and minicomputer hierarchical systems (5,

12). Recently, an Apple microcomputer was used to control an enzyme electrode system via a D/A-A/D converter system (7). Thus, the use of microcomputers for independent fermentation control seemed to be the next logical step. We describe here a relatively simple and inexpensive microcomputer-coupled fermentation system which has the flexibility and power to simultaneously acquire data from and control different types of fermentation processes. MATERIALS AND METHODS Bacteria and media. Vibrio natriegens ATCC 14048 and Bacillus thuringiensis DM-1 were used in this study. V. natriegens was grown at 37°C in a modification of M9 medium (8) containing 1 or 10 mM glucose as the carbon source. The medium was modified by the addition of 1.5% NaCl and filter-sterilized vitamins as follows (in micrograms liter-1): biotin, 130; folic acid, 130; pyridoxine, 665; thiamine, 330; p-aminobenzoic acid, 330; riboflavin, 330; nicotinic acid, 330; pantothenic acid, 330; vitamin B12, 330; and lipoic acid, 130. B. thuringiensis was grown at 32°C in tryptose-phosphate broth consisting of the following (in grams liter- ): tryptose (Difco Laboratories, Detroit, Mich.), 20; glucose, 2; NaCl, 5; and Na2HPO4, 2.5. Fermentation hardware systems. Figure 1 shows the hardware setup for the microcomputer-assisted fermentation system. The central system consisted of an Apple II Plus microcomputer (64 K random-access memory) with a monitor (cathode ray tube) and dual disk drives and a Cyborg ISAAC model 91A multichannel A/D and D/A converter. ISAAC contains 16 channels of A/D input, 4 channels of D/A output, 16 bits of binary input-output (digital I/O), and four isolated Schmitt trigger switches. Four channels of D/A output were added to expansion slot no. 7 of ISAAC, yielding eight total output channels. This gave the AppleISAAC system the capability to control, with feedback, eight analog systems, plus up to six binary (on-off) systems. The ISAAC system also contained 5-V (transistor-transistor logic level) and 12-V outputs which could be used as power sources for certain functions or devices. Because ISAAC put out a latched 5-V signal on binary demand, a relay box was constructed containing eight separate optically isolated 5-V relays which opened or closed 115-V alternating current line voltage. The relay box could be used for any control feature

* Corresponding author. t Present address: Bristol Laboratories, Syracuse, NY 13221. t Present address: Battelle Columbus Laboratories, Columbus, OH 43210. 239

240

APPL. ENVIRON. MICROBIOL.

TITUS ET AL. A

C

D

FIG. 1. Diagram of the microcomputer control hardware (not drawn to scale). The Braun MRR-1 fermentation control system included a P02 amplifier and controller (A), a foam controller (B), a pH amplifier and controller (C), and digital and strip chart displays (D). The Apple II Plus microcomputer, cathode ray tube (CRT) monitor, disk drives 1 and 2, and ISAAC D/A-A/D converter were connected to the fermentation system via a junction box (E). Binary outputs (Ea) were used to activate 115-V alternating current circuits via the relay control box (F) to operate pumping systems (G) for the turbidostat. Analog outputs (Eb) were used to override various functions of the MRR-1 control unit (top of figure) and operate the servo controller (H) of the agitation potentiometer. Various feedback parameters, such as the IC temperature probe (1), the PO2 probe (J), the agitation feedback DC motor (K), the pH probe (L), and the CdS turbidimeter (M), were connected to the analog input channels of the junction box (Ec). These were sent to ISAAC for conversion to digital signals. A peristaltic pump (N) provided the turbidimeter with a closed culture loop for continuous measurement of turbidity.

requiring the turning on and off of pumps, switches, or valves. Paired MRR-1 fermentation control units (B. Braun, Inc., Burlingame, Calif.), each containing a strip chart recorder and PO2, pH, and foam controls, were connected to the Apple-ISAAC system via the analog inputs. The MRR-1 pH and P02 control modules were overridden by the computer D/A output system according to the instruction manual for the MRR-1 fermentation control units. To obtain the desired control sensitivity in the P02 module, the inputs and outputs of the P02 system were inverted to 0 to +5 V for use with ISAAC (R. Bailey, B. Braun Instruments, Inc., personal communication). Ingold autoclavable dissolved oxygen potentiometric electrodes mounted in each fermentation vessel were connected to the MRR-1 P02 controller to measure dissolved oxygen. Nonautoclavable Cple-Palmer sealed epoxy body pH electrodes (0.6 by 22 cm) were sterilized by overnight immersion in Chlorox solution (Chlorox Co., Oakland, Calif.). Before they were mounted in the fermentors, the electrodes were aseptically washed with sterile water. The pH electrodes were then mounted in 20-ml syringe barrels that were fitted with rubber stoppers into the top plates of the fermentor vessels (Fig. 1). Holes were cut through the syringes and their plunger seals to accommodate the pH probes and hold them tightly in place. This was found to adequately seal the assembly and fermentors against contamination. The pH and oxygen control systems of the MRR-1 fermentation control units shared a common ground

with the ISAAC A/D converter. For calibration of the pH and P02 controller systems of the computer program with the MRR-1 fermentation control unit, values above and below the desired set point values were put into the computer before the fermentation run to set up standard curves. The program contained algorithms to initiate and save the standard calibration curves for each parameter and to compute obtained input values with them. A Harvard model 1203 peristaltic pump (Harvard Apparatus Co., South Natick, Mass.) was used for medium inlets to the fermentors. Because the Harvard pumps are highly accurate and linearly calibrated, no direct feedback was required on the pumping system. The pump was connected to the relay box in which 5-V optically isolated relays, triggered by the 5-V binary outputs of ISAAC, were housed. The pump speed could be set at a given dilution rate, or if semicontinuous additions were desired, the changes could be made by altering the frequency of the pumping action with the binary output via the relay box. The number of times that the pump was turned on was recorded and saved in a special datum file by a counter function written into the Labsoft BASIC program. The primary control for dissolved oxygen, as monitored by the computer via the MRR-1 P02 amplifier and control modules, was by the adjustment of agitation speeds between 150 and 350 rpm. The agitation rate was controlled by servo controllers (pulse width comparator motors used for radio control models and available at local hobby stores; 0 to 1200

VOL. 47, 1984

turning radius over 2.5 to S V) attached to the agitation control potentiometer. The analog- to-pulse width conversion for the servo control was based on the circuit described by Wolfe (14). To provide feedback of the agitation rate, a direct current (DC) motor (Radio Shack no. 273-208) was mechanically attached to the drive shaft of each fermentor. These DC motors generated an analog signal of ca. 0 to + 1 V for 0 to 500 rpm which was put into the Apple-ISAAC system. The agitation feedback DC voltage to the computer was calibrated against several mechanically determined agitation rates. Likewise, the servo controllers on the agitation potentiometers were calibrated to between 50 and 500 rpm so that a certain voltage (analog) output from the computer was reliably translated into the proper degree of turn for the desired agitation rate. The secondary P02 control (although never required for the runs shown here) was the airflow rate. This was regulated by a servo-controlled needle valve in the MRR-1 P02 module on command from the computer. Fermentor temperatures were monitored by a temperature-sensing integrated circuit (IC) probe (LM334; Radio Shack no. 276-1734) which took a 12-V output voltage from ISAAC and a series of two resistors, a 220-f adjustable resistor which connected back to the IC, and a 10,000-fl resistor connected to a ground according to instructions provided with the temperature sensor. The IC was coated with a silicone glue and was placed in an external temperature well. The signal from the IC was connected to an A/D input channel of ISAAC and precalibrated against known temperatures of between 20 and 45°C. Turbidity was measured either by a continuous flow cell used with a Spectronic 21 spectrophotometer (Bausch & Lomb, Inc., Rochester, N.Y.) or by a special laboratoryconstructed turbidirieter. The turbidimeter worked by continuously pumping culturd- through closed loops which ran through a dark box containing a 6-V light bulb and two isolated CdS photoresistors (Radio Shack no. 276-116). A glass tube (inside diameter, 10 mm) connected to the fermentor and pump via standd hosing was positioned between the light and the CdS detector. A 5-V signal powered the CdS detector which essdntially acted as a variable resistor. As the cultures increased in density (thus diminishing the amount of light the CdS detector received), the resistance increased and the voltage decreased. Thus, the voltage input to the computer was inversely related to the optical density. Standard curves also have been constructed to relate analog output from the CdS detectors to optical density. Software system. The programming language used was Labsoft, an extension of Applesoft BASIC, which was supplied with ISAAC. The Labsoft software, containing unique commands differentiated from Applesoft by a preceding anipersand, was loaded into the Apple-ISAAC system before the actual fermentation program. The fermentation program, developed in our laboratory, is outlined by the flow diagram in Fig. 2. Analog inputs from the various sensing devices were fed into ISAAC via a junction box and were converted to digital values that were read by the AppleISAAC system. The Apple-ISAAC system was programmed to collect these values, log them in sequential datum files, and make decisions based upon them. Data were acquired and saved by a series of averaging loops. The Apple-ISAAC system took ca. 200 readings per s for each input sensor, averaged them, and checked the average against the pre-set point limits. If corrective measures were required, the proper D/A outputs were activated and the parameter was corrected to pre-set point limits. The entire loop took ca. 10 s, and those averages were constantly shown and updated

MICROCOMPUTER-CONTROLLED FERMENTATIONS

241

after each 10-s loop on the cathode ray tube monitor. After 10 passes through the averaging loop (ca. 1.5-min total), the 10 values were averaged, stored in a datum array, and considered as a datum point. Each datum point thus represented the average of ca. 2,000 readings. The datum points for each sensor were graphed on an NEC PC-8023A-C dot matrix printer. The Apple-ISAAC system was programmed to accumulate 50 datum points for each of 10 functions and then save them (ca. once every 76 min) on a datum file disk. It took the disk operating system ca. 2 min to save the 500 datum points in 10 separate files, during which time the fermentors were not under computer control. This was found to be the best compromise which allowed both a minimum of lost control time and a maximum efficiency for datum collection. The datum averaging and saving system described above was designed to accommodate both short fermentation runs (i.e., 10 days) by a variable which changed the datum point times according to the length of the run. The values shown were for a 48-h run. The data were later retrieved from the datum file disk with an in-house-modified Scientific Plotter (Interactive Microware, Inc., State College, Pa.) plotting program. Thus, the program was written so that the data could be viewed in three ways. (i) They were continuously displayed and updated every 10 s on the cathode ray tube monitor for an instantaneous examination of the fermentation progress. (ii) They were printed every few minutes on a hard copy to follow the progress of, and the changes in, the fermentation runs. (iii) They were stored on a standard 5.25-in. (ca. 13-cm) floppy disk for future retrieval and analysis. Fermentation runs. For the experiments shown in Fig. 3, computer-controlled versus uncontrolled batch growth of B. thuringiensis was carried out in paired 14-liter (10-liter working volume) New Brunswick fermentors. These experiments tested the computer monitoring of temperature maintenance at 32°C, P02 control by alteration of the agitation rate, pH control in both directions, foam control by the MRR-1 fermentation control units, and continuous monitoring of turbidity by a flow-through cell hooked to a Spectronic 21 spectrophotometer set at 550 nm. The chemostat versus turbidostat experiments shown in Fig. 4 were run with paired 1-liter (500-ml working volume) glass jar fermentors maintained at 37°C by a circulation water bath. Agitation of the fermentors was carried out by stirring each sample with 1.5-in. (ca. 4-cm) stir bars, and the airflow rate was computer controlled via the MRR-1 P02 controller. The Vibrio cultures were grown in monitored but uncontrolled batch culture to the mid-log phase. Then flow rate control (chemostat) versus turbidity feedback control (turbidostat) was initiated. The chemostat culture contained 1 mM glucose for carbon limitation, and the turbidostat culture contained 10 mM glucose for excess carbon. RESULTS AND DISCUSSION Figure 3 shows results obtained from the modular microcomputer system for computer-monitored and -controlled versus computer-monitored (only) batch growth of B. thuringiensis in paired 10-liter fermentors. These runs were typical of several B. thuringiensis growth experiments performed with this system in which the results were very reproducible. The optical densities of B. thuringiensis in the computercontrolled and uncontrolled vessels, as monitored by a laboratory-constructed flow-through cell with the Spectronic 21 spectrophotometer, were identical and reached final densities of ca. 0.43 absorbance unit after 21 h of fermentation. In the computer-controlled vessel, the servo control on

242

TITUS ET AL.

APPL. ENVIRON. MICROBIOL.

FIG. 2. Flow diagram of the fermentation control program. Control set points were entered by floppy disks or keyboard. The initial control set points were saved on the file disk and used throughout the fermentation runs. During each command loop the data were stored in files according to the total elapsed time. A series of decision control loops then were accessed to control each parameter within the set point values. Each sensor was read and fed back to the computer via ISAAC for simple decisions. Changes made to the cultures (i.e., acid and base additions) were incremented by a program-internal increment counter. Cascade control is exemplified by the computer operation of the servos via analog-to-pulse width (A-PW) converters to turn servos for agitation control.

the agitation potentiometer increased the agitation from an initial set point of 150 rpm to a final rate of ca. 325 rpm to maintain P02 above 50% of saturation. Because the agitation rate did not exceed the preset limit of 350 rpm, the airflow rate was not increased. Thus, the cascade control of dissolved oxygen by agitation was sufficient to maintain the P02 above the set point. Other experiments have shown that the secondary p02-controlling mechanism, airflow, was activated once the maximum preset agitation rate was achieved. The dissolved oxygen of the uncontrolled vessel declined rapidly to 0% of saturation after ca. 8 h of fermentation. The agitation of the uncontrolled vessel was not altered but was monitored at a constant 200 rpm via the DC voltagegenerating feedback motor. The pH of the uncontrolled vessel declined to 5.9, whereas the pH in the computercontrolled vessel was properly maintained between 7.0 and 7.4 by base and acid additions, respectively. This demonstrated the ability of the on-line system to control and alter dissolved oxygen and pH while collecting, saving, printing out, and displaying the data. The temperature of both vessels was measured and plotted by the computer at a constant 320C via the temperature-sensitive IC but was controlled externally by the built-in heat exchanger of the 14-liter New Brunswick fermentation units. Although not shown in this study, the temperature could be controlled for temperature shift experiments by attaching the servo control system described herein

to the potentiometer of the heat exchanger system. V. natriegens was grown in monitored but uncontrolled defined medium batch cultures for ca. 5 h (Fig. 4). The computer control was then initiated (first arrow, Fig. 4), and the vibrios were subsequently grown in chemostat versus turbidostat cultures. The cultures responded to the set point commands given by the computer, and the chemostat reached the steady state after 10 to 12 h. The dilution rate of the chemostat was set at D = 0.48 h-'; thus, it took the vibrios about five to six residence times to reach the steady state. The flow rate of the turbidostat was calculated to be 350 ml/h, so Dturb = 0.7 h-1. The turbidostat did not reach the steady state until ca. 25 h after the control was started (second arrow, Fig. 4). This may have been due to the changes in dissolved oxygen which eventually stabilized. The dissolved oxygen lower limits for the chemostat and turbidostat were set at 80% of saturation, and the control was by MRR-1-regulated airflow. Enough growth occurred in the turbidostat to force control of the dissolved oxygen by the MRR-1 airflow system (Fig. 4). Growth in the glucoselimited chemostat, however, was at a low enough level that the dissolved oxygen remained above 95% of saturation. The pH of both the chemostat and the turbidostat was maintained between the set points of 6.9 and 7.0. These experiments demonstrated that the simple and inexpensive turbidimetry system was sensitive enough to feedback control the turbi-

VOL. 47, 1984

MICROCOMPUTER-CONTROLLED FERMENTATIONS

.1 ~ ~

~

~

~

~

~

0

TIME (hours) FIG. 3. Batch growth of B. thuringiensis in tryptose-phosphate medium in fermentors with a 10-liter working volume. In one vessel the following parameters were only monitored at preset values: dissolved oxygen (DO-U), agitation (RPM-U), and pH (pH-U). The second vessel was computer controlled at 50% dissolved oxygen (DO-C) by increasing the agitation rate based on PO2 (RPM-C). The pH (pH-C) was controlled via acid or base additions. Temperature (T) and culture turbidities (OD, optical density) were monitored (and were identical) for both vessels.

dostat while simultaneously providing an on-line analysis of the steady state in both the chemostat and turbidostat systems. The comparatively slow growth of the vibrio could

a

pH 7

a

I-5. EZ

.

3.

+II 00

v

"'-

so DT

24

30

36

42

TIME (hours)

FIG. 4. V. natriegens was grown in modified M9 medium in fermentors with a 500-ml working volume under batch conditions. Computer control for chemostat ( ) versus turbidostat (------) conditions was started after ca. 5 h of growth (small arrow). The pH for both systems was controlled between 6.9 and 7.0. Culture densities were monitored with CdS photoresistors and are represented as arbitrary resistance units. The chemostat culture (OD-C) reached the steady state after ca. 15 h, whereas the turbidostat culture (OD-T) fluctuated with unstable dissolved oxygen conditions (DO-T) for ca. 20 h before entering steady-state conditions (large

arrow).

probably be attributed to the defined medium in which it was grown. The power of the ISAAC-Apple system is that ISAAC uses 12-bit D/A and A/D converters which give 4,096 digital steps over a 0- to 5-V analog signal. The analog parameters thus have a resolution of ca. 82 digital steps per every 100 mV. Particularly when 0- to 1-V steps are used, such as with the DC agitation feedback measurement, the resolution offered by the 12-bit converter is even more important. There are several inexpensive 8-bit converters available, but they give a resolution equal to only 256 steps over either a 0to 1- or 0- to 5-V analog signal. These converters would not offer sufficient resolution to control fermentation processes such as the types shown here. The microcomputer fermentation control package described here obviously does not compete with the DEC PDP11 series or any other mini- or combined micro- and minicomputer hierarchical fermentation control system. Instead, this system seems ideal for laboratories that are not able to afford or that do not require the power of a mini- or main frame computer. We have shown microcomputer control of the basic functions needed to initiate process control on fer entations. The datum storage disks easily have enough roon to save twice the maximum amount of data shown here. The Apple-ISAAC system is also capable of handling several more input-output variables than used here. Such possible functions could include substrate or product level feedback via enzyme electrodes (1) or exit gas analyses (4, 5, 9, 11). These would allow for a more sophisticated level of control, i.e., material balances and respiratory quotient-type control from on-line gas and substrate concentration measurements

(13).

Most of the control loops employed here were simple algorithms in which the actual datum values were regularly checked and updated against programmed set points. However, one cascade loop, the control of dissolved oxygen via the agitation rate (primary control) and then the airflow rate (secondary control), was included. This demonstrates the potential for using the Apple-ISAAC system in several different types of fermentations. Moreover, the system is flexible enough that the algorithms and control loops could be used to control and monitor laboratory processes other than fermentations. The controlling modes used in this work were of a relatively simple nature and demonstrated our first steps toward microcomputer control of fermentation processes. We are now engaged in adding more high level-type control processes, i.e., the controlled addition of medium components based on calculated values such as the respiratory quotient or biomass balance as mentioned above. Our system would then meet criteria suggested by several investigators for a "complete" computer-controlled fermentation

laboratory (4, 5, 13). One potential disadvantage of the Apple-ISAAC control package is that programming is done with a BASIC interpret-

I

16

243

er which results in slow program execution. Fermentation

processes, however, require long time periods, and the speed of BASIC execution has never been a factor in any of our fermentation runs. The lack of problems with this is exemplified by the number of time-consuming averaging loops placed in the datum analysis and storage portion of the program. An advantage of the BASIC language is that many students and researchers already have acquired BASIC programming skills or can learn them without difficulty. The Apple-ISAAC fermentation control package described here offers a relatively inexpensive, reasonable, and realistic alternative to small laboratories who wish to com-

f

244

TITUS ET AL.

puter control fermentation processes. Although the system has certain limitations, particularly in datum storage, typical fermentations such as those shown here offer no difficulties. We have also used this system to computer control longer and more complex antibiotic fermentations by certain streptomycetes, demonstrating once again its versatility at the laboratory scale level. Furthermore, we have recently learned that other investigators are using the Apple-ISAAC system for computer control of certain other fermentation systems, such as the monitoring and set point control of gas additions for anaerobic growth of Desulfovibrio sp. (D. J. Cork and F. K. Konan, Dev. Ind. Microbiol., in press). ACKNOWLEDGMENTS Development of this system has been supported in part by several sources, including The Ohio State University Departments of Microbiology and Biochemistry and the College of Biological Sciences and a grant (PCM-8204778) from the National Research Foundation to W. R.S. We thank J. I. Frea for his suggestions about the CdS sensors, L. M. Khoury for assistance with the figures, and S. Schlasner for advice in preparation of this manuscript. We especially thank R. M. Pfister for his encouragement, support, and thoughtful discussions which led to the development of this project and this manuscript. The fermentation software programs, schematics of the alterations made to existing equipment, and schematics of our in-house-built relay systems are available upon request from W.R.S. LITERATURE CITED 1. Barker, A. S., and P. J. Somers. 1978. Enzyme electrodes and enzyme based sensors, p. 120-151. In A. Wiseman (ed.), Topics in enzyme and fermentation biotechnology, vol. 2. Halsted Press, Chichester, England.

APPL. ENVIRON. MICROBIOL. 2. Dobry, D. D., and J. L. Jost. 1977. Computer applications to fermentation operations, p. 95-114. In D. Perlman (ed.), Annual reports on fermentation processes, vol. 1. Academic Press, Inc., New York. 3. Hampel, W. A. 1979. Application of microcomputers in the study of microbial processes. Adv. Biochem. Eng. 13:1-33. 4. Harmes, C. S., III. 1972. Design criteria of a fully computerized fermentation system. Dev. Ind. Microbiol. 13:146-156. 5. Hatch, R. T. 1982. Computer applications for analysis and control of fermentation. Annu. Rep. Ferm. Proc. 5:291-311. 6. Jefferies, R. P., III, S. S. Klein, and J. Drakeford. 1979. Singleboard microcomputer for fermentation control. Biotechnol. Bioeng. Symp. 9:231-239. 7. Kernevez, J. P., L. M. Konate, and J. L. Romette. 1983. Determination of substrate concentrations by a computerized enzyme electrode. Biotechnol. Bioeng. 25:845-855. 8. Maniatis, T., E. F. Fritsch, and J. Sambrook. 1982. Molecular cloning, a laboratory manual. Cold Spring Harbor Laboratory, Cold Spring Harbor, N.Y. 9. Mohler, R. D., P. J. Hennigan, H. C. Lim, G. T. Tsao, and W. A. Weigand. 1979. Development of a computerized fermentation system having complete feedback capabilities for use in a research environment. Biotechnol. Bioeng. Symp. 9:257-268. 10. Nyiri, L. K. 1972. A philosophy of data acquisition, analysis, and computer control of fermentation processes. Dev. Ind. Microbiol. 13:136-145. 11. Park, S. H., K. T. Hong, J. H. Lee, and J. C. Bae. 1983. On-line estimation of cell growth for glutamic acid fermentation system. Eur. J. Appl. Microbiol. Biotechnol. 17:168-172. 12. Rolf, M. J., P. J. Hennigan, R. D. Mohler, W. A. Weigand, and H. C. Lim. 1982. Development of a direct digital-controlled fermentor using a microminicomputer hierarchical system. Biotechnol. Bioeng. 24:1191-1210. 13. Weigand, W. A. 1978. Computer applications to fermentation processes. Annu. Rep. Ferm. Proc. 2:43-72. 14. Wolfe, G. W. 1982. Computer peripherals that you can build. TAB Books, Blue Ridge Summit, Pa.