INSTRUMENTATION A discussion of servomechanisms emphasizes the relatively new concept that a power device can replace the operator, detect errors, and apply corrective forces
by Ralph /"~\N TWO occasions at least we have referred to servomechanisms ^-^ and their increasingly important role to modern analytical instruments. Two months ago [ANAL. CHEM., 20, 21A (1948)] we described equipment available from Servomechanisms, Inc. As a result, many inquiries have come in, asking in effect, " W h a t is a servomechanism and what are its uses?" Again we repeat Hazen's definition, "A servomechanism is a power amplifying device in which the amplifying element driving the output is activated by the difference between the input and the output." This simple statement describes a philosophy of measurement and control which has already wrought fundamental changes. Despite this achievement, it is contrary to much of the thinking acquired in our early contacts with the theory and practice of measurement. Let us consider the following problem: I t is required t o measure and record a D.C. potential of the order of a few millivolts with a precision of one p a r t in Ά thousand and to de tect fairly rapid changes in t h a t potential. The potential in ques tion may arise from a thermocouple, phototube, or electrode sys tem or any other device which h'as translated a chemical or physical phenomenon into an equivalent e.m.f. Stage I. A classical approach would involve connecting the source to a slightly underdamped short-period galvanometer and recording the motion of its mirror photographically. The sources of error in this technique include zero drift, magnetic impurities, and nonlinearity in galvanometer response. These can be de termined and due allowance made for their magnitude. Vibration can be eliminated by shock-mounting the galvanometer. Ther mal parasitics and pickup from stray fields can be eliminated by lagging with insulation and by electrical shielding. At this stage it may be found t h a t rapid changes in e.m.f. are not followed faith fully, i.e., they are superimposed upon the dynamics of the moving coil of the galvanometer. Stage I I . T h e decision may be reached to select a gal vanometer with a shorter period and lower sensitivity and then to correct the latter deficiency by preliminary amplification. A stable amplifier is built and before long the amplifier requires con stant voltage regulators, batteries, shielding,'and even tempera ture and humidity control. By this time, one has acquired an impressive array of apparatus, and if useful measurements have been made, a publication is in order. Stage III. If the instrumental equivalent of this array of ap paratus is to be designed, the approach will be radically different. A well known solution involves the following: The primary D.C. signal is chopped or switched synchronously by a rotating or vibrating commutator and applied to the primary of a small transformer. The output is fed to an A.C. amplifier of very high gain (> 10,000 X ) . A final power amplifier feeds this signal to one leg of a two-phase motor; the other leg is excited in quadrature from the A.C. power fine. This motor will run at a speed more or less proportional to the magnitude of the primary signal and in a direction governed solely by its polarity. If the polarity reverses, the motor will rotate in the opposite direction. If one now can forget his classical instincts to measure the motor speed, and calibrate the amplifiers and the converter element, one is now in a position to take the next step and build a servomechanism.
M.
Müller
This is the important step. Through suitable reduction gears, the motor is coupled to the slide wire of a voltage divider or "potentiometer." T h e proper potential from a steady source is maintained across the voltage divider and as the motor drives the slider over its surface a fraction of the total slide-wire voltage is selected and this is applied in series-opposition to the voltage under measurement. As a consequence, the converter element and its associated amplifiers are constantly excited by a signal which is equal to the difference between the e.m.f. under measure ment and the compensating e.m.f. In other words, the corrective mechanism operates from an "error signal" only. The linkage which drives the compensating potentiometer also actuates a re cording pen or bold indicating pointer or both. As a rule, numer ous automatic devices periodically compare the potentiometer voltage with a standard cell and the same servosystem readjusts the working current when necessary. This is essentially the principle of modern recording potentiometers such as the Brown Electronik and the Leeds & Northrup Speedomax G. T h e im portant point to note is the extent to which the characteristics of most components in this chain of operations became relatively unimportant. For example, the amplifier gain is very high b u t need not be constant. A reduction in its gain reduces the sensi tivity but not the accuracy, because the gain can be raised to the point where it begins to approach the noise level of the system and in all but a few cases, this will be at least one or two orders of magnitude smaller than the signal. The equipment will probably work just as satisfactorily in an airplane as in a laboratory. Almost any phenomenon can be measured in a related manner. It becomes necessary to seek a means of measuring the phenom enon and at once another or identical means for nullifying the effect. The intermediate step requires enough amplification to supply power to the nullifying device in proper phase and ampli tude. Under those conditions the feedback is an accurate measure of the primary variable. One might infer t h a t a servomechanism does not necessarily involve moving parts. Indeed an electronic amplifier in which a definite fraction of the output is fed back degoneratively into t h e input is an exact equivalent of this principle. Historically, servomechanisms antedated the inverse feed-back amplifier, but today generally include the feed-back principle, and the theory of both subjects is almost indistinguishably bound by identical concepts. The servomechanism principle is relatively simple when ap plied to instruments, although the possible applications are almost without limit. The subject becomes most complicated, however, if a servosystem involves high inertia elements or if sharp transi ents occur. For example, in the preceding example the restoring motor and the pen which it, drives might be expected to follow potential fluctuations of a few hundred microvolts per second faithfully, but transients occurring within a matter of milli seconds would escape undetected. Similarly, if the servomotor were required to drive not only the pen and compensating slidewire but also a large gate valve, it could do so only if the measured system were changing very slowly. When servotheory is to be applied to systems involving high
23 A
24 A
ANALYTICAL
CHEMISTRY
inertia loads and severe transients it is high time to consult an expert. ^Some important literature sources are : MacColl, L. Α., "Fundamental Theory of Servomechanisms," New York, D . Van Nostrand Co., 1945. Bode, H. W., "Network Analysis and Feedback Amplifier De sign," New York, D . Van Nostrand Co., 1945. Batcher, R. R., and Moulic, W., "Electronic Control Hand book," New York, Caldwell-Clements, 1946. Smith, E. S., "Automatic Control Engineering," New York, McGraw-Hill Book Co., 1944. Greenwood, Ι. Α., Holdam, J. V., and MacRae, D., "Electronic Instruments," New York, McGraw-Hill Book Co., 1948.
Klett•••• Photometers Klett-Summerson Photoelectric
Glass Cell Coloremeter • No. 900-3
•
The Klett Fluorimeter •
No. 2070
Designed for the rapid and accurate d e t e r m i n a t i o n of thiamin, riboflavin, and other substances which fluoresce in solution. The sensitivity and sta bility are such that it has been found particu larly useful in determining very small amounts of these substances. •
KLETT SCIENTIFIC PRODUCTS ELECTROPHORESIS APPARATUS BIO-COLORIMETERS GLASS ABSORPTION CELLS COLORIMETER NEPHELOMETERS GLASS STANDARDS KLETT REAGENTS
•
Klett
Manufacturing Co.
1 7 9 E A S T 87TH STREET, NEW YORK, Ν. Υ.
The mathematical treatment of transient response is still an· active field of research. Practical designs have not lagged too greatly behind the theoretical advances and the late war provided many examples of high power servos controlled by relatively fee ble primary impulses. These included fire control, gun-laying, and numerous position stabilizing systems. A great convenience is the extensive number of components available for translating linear or angular displacements into equivalent electrical quantities or simple mathematical functions thereof. These include synchros, rotatable transformers, differ ential synchros, etc. Several companies and numerous consult ants can design servosystems to meet almost any requirement. The velocity of sound through liquids and gases has been measured time and again since Kundt's early work and usually with so much attention to elementary definitions that few new approaches have arisen. Conventionally, one toots into one end of a tube and measures the time for the sound wave to reach the far end. Alternatively, one locates nodes by a movable receiver and by these and other modifications can get a reading relating to the composition of the medium. To play safe, one calibrates the entire system empirically. 'General Electric feeds an impulse into one end of a resonator, picks it up at the far end, amplifies slightly, and feeds it back at the input. The system soon resonates at a period governed by the speed of sound in that medium. A frequency meter in the amplifier circuit provides the information. This is the equivalent of the old trick of holding a telephone receiver in front of the transmitter mouthpiece. A simple balance can be loaded progressively and the beam de flection can be measured photoelectrically. To follow the progressive change in weight as a function of time, would require a constant light source, stable phototube and amplifier, and uniform deflec tion sensitivity of the balance. The servomechanism principle would pass the amplified photocurrent through a solenoid in the core of which a soft iron counterweight was hanging, suspended from the opposite pan of the balance. The solenoid current would be recorded in this case. The sole drawback of this im proved technique lies in the very careful design required to pro duce a solenoid in which the pull on the suspended core is strictly proportional to the current, even over a limited range. The self-balancing potentiometer of Gilbert and as used in the General Electric Autopot admits the e.m.f. to a mirror gal vanometer and projects a beam of light on a twin phototube. The amplified photocurrent may be used as an accurate measure, of the input voltage provided this current is used to generate an fii drop, through a resistor, which is applied back to the gal vanometer in series-opposition to the unknown e.m.f. This principle has been improved to the extent of removing the residual disadvantages of a delicate galvanometer. In place of the latter a deflection vane moves in the plane of two oscillator coils. For an oscillator frequency of 30 megacycles or more, a vane motion of a few thousandths of an inch is sufficient to stop oscilla tion and produce large changes in the D.C. plate current. These are fed back to the vane deflector coil for cancellation. I t has been found advantageous to use a separate compensator coil in order to isolate output from input. Even temperature com pensation is achieved by temperature sensitive shunting net works. This system is immune to severe shock or vibration. Null methods have long enjoyed their place in precise measure ments. What is relatively new is the general concept that a power device can detect errors, evaluate sign and magnitude, and apply corrective forces much more rapidly than an operator.
