Computational ircuits
When someone mentions the word “computer,” a digital device is what usually comes to mind. Digital circuits represent numerical quantities in binary format: patterns of 1’s and 0’s represented by a multitude of transistor circuits operating in saturated or cutoff states. However, analog circuitry may also be used to represent numerical quantities and perform mathematical calculations, by using variable voltage signals instead of discrete on/off states.
Here is a simple example of binary (digital) representation versus analog representation of the number “twenty-five:”
Digital circuits are very different from circuits built on analog principles. Digital computational circuits can be incredibly complex, and calculations must often be performed in sequential “steps” to obtain a final answer, much as a human being would perform arithmetical calculations in steps with pencil and paper. Analog computational circuits, on the other hand, are quite simple in comparison, and perform their calculations in continuous, real-time fashion. There is a disadvantage to using analog circuitry to represent numbers, though: imprecision. The digital circuit shown above is representing the number twenty-five, precisely. The analog circuit shown above may or may not be exactly calibrated to 25.000 volts, but is subject to “drift” and error.
In applications where precision is not critical, analog computational circuits are very practical and elegant. Shown here are a few op-amp circuits for performing analog computation:
Each of these circuits may be used in modular fashion to create a circuit capable of multiple calculations. For instance, suppose that we needed to subtract a certain fraction of one variable from another variable. By combining a divide-by-constant circuit with a subtractor circuit, we could obtain the required function:
Devices called analog computers used to be common in universities and engineering shops, where dozens of op-amp circuits could be “patched” together with removable jumper wires to model mathematical statements, usually for the purpose of simulating some physical process whose underlying equations were known. Digital computers have made analog computers all but obsolete, but analog computational circuitry cannot be beaten by digital in terms of sheer elegance and economy of necessary components.
Analog computational circuitry excels at performing the calculus operations integration and differentiation with respect to time, by using capacitors in an op-amp feedback loop. To fully understand these circuits’ operation and applications, though, we must first grasp the meaning of these fundamental calculus concepts. Fortunately, the application of op-amp circuits to real-world problems involving calculus serves as an excellent means to teach basic calculus. In the words of John I. Smith, taken from his outstanding textbook, Modern Operational Circuit Design:
Differentiation is the first operation typically learned by beginning calculus students. Simply put, differentiation is determining the instantaneous rate-of-change of one variable as it relates to another. In analo differentiator circuits, the independent variable is time, and so the rates of change we’re dealing with are rates of change for an electronic signal (voltage or current) with respect to time.
Suppose we were to measure the position of a car, traveling in a direct path (no turns), from its starting point. Let us call this measurement, x. If the car moves at a rate such that its distance from “start” increases steadily over time, its position will plot on a graph as a linear function (straight line):
If we were to calculate the derivative of the car’s position with respect to time (that is, determine the rate-of-change of the car’s position with respect to time), we would arrive at a quantity representing the car’s velocity. The differentiation function is represented by the fractional notation d/d, so when differentiating position (x) with respect to time (t), we denote the result (the derivative) as dx/dt:
For a linear graph of x over time, the derivate of position (dx/dt), otherwise and more commonly known as velocity, will be a flat line, unchanging in value. The derivative of a mathematical function may be graphically understood as its slope when plotted on a graph, and here we can see that the position (x) graph has a constant slope, which means that its derivative (dx/dt) must be constant over time.
Now, suppose the distance traveled by the car increased exponentially over time: that is, it began its travel in slow movements, but covered more additional distance with each passing period in time. We would then see that the derivative of position (dx/dt), otherwise known as velocity (v), would not be constant over time, but would increase:
The height of points on the velocity graph correspond to the rates-of-change, or slope, of points at corresponding times on the position graph:
What does this have to do with analog electronic circuits? Well, if we were to have an analog voltage signal represent the car’s position (think of a huge potentiometer whose wiper was attached to the car, generating a voltage proportional to the car’s position), we could connect a differentiator circuit to this signal and have the circuit continuously calculate the car’s velocity, displaying the result via a voltmeter connected to the differentiator circuit’s output:
Recall from the last chapter that a differentiator circuit outputs a voltage proportional to the input voltage’s rate-of-change over time (d/dt). Thus, if the input voltage is changing over time at a constant rate, the output voltage will be at a constant value. If the car moves in such a way that its elapsed distance over time builds up at a steady rate, then that means the car is traveling at a constant velocity, and the differentiator circuit will output a constant voltage proportional to that velocity. If the car’s elapsed distance over time changes in a non-steady manner, the differentiator circuit’s output will likewise be non-steady, but always at a level representative of the input’s rate-of-change over time.
Note that the voltmeter registering velocity (at the output of the differentiator circuit) is connected in “reverse” polarity to the output of the op-amp. This is because the differentiator circuit shown is inverting: outputting a negative voltage for a positive input voltage rate-of-change. If we wish to have the voltmeter register a positive value for velocity, it will have to be connected to the op-amp as shown. As impractical as it may be to connect a giant potentiometer to a moving object such as an automobile, the concept should be clear: by electronically performing the calculus function of differentiation on a signal representing position, we obtain a sig al representing velocity.
Beginning calculus students learn symbolic techniques for differentiation. However, this requires that the equation describing the original graph be known. For example, calculus students learn how to take a function such as y = 3x and find its derivative with respect to x (d/dx), 3, simply by manipulating the equation. We may verify the accuracy of this manipulation by comparing the graphs of the two functions:
Nonlinear functions such as y = 3x2 may also be differentiated by symbolic means. In this case, the derivative of y = 3x2 with respect to x is 6x:
In real life, though, we often cannot describe the behavior of any physical event by a simple equation like y = 3x, and so symbolic differentiation of the type learned by calculus students may be impossible to apply to a physical measurement. If someone wished to determine the derivative of our hypothetical car’s position (dx/dt = velocity) by symbolic means, they would first have to obtain an equation describing the car’s position over time, based on position measurements taken from a real experiment — a nearly impossible task unless the car is operated under carefully controlled conditions leading to a very simple position graph. However, an analog differentiator circuit, by exploiting the behavior of a capacitor with respect to voltage, current, and time i = C(dv/dt), naturally differentiates any real signal in relation to time, and would be able to output a signal corresponding to instantaneous velocity (dx/dt) at any moment. By logging the car’s position signal along with the differentiator’s output signal using a chart recorder or other data acquisition device, both graphs would naturally present themselves for inspection and analysis.
We may take the principle of differentiation one step further by applying it to the velocity signal using another differentiator circuit. In other words, use it to calculate the rate-of-change of velocity, which we know is the rate-of-change of position. What practical measure would we arrive at if we did this? Think of this in terms of the units we use to measure position and velocity. If we were to measure the car’s position from its starting point in miles, then we would probably express its velocity in units of miles per hour (dx/dt). If we were to differentiate the velocity (measured in miles per hour) with respect to time, we would end up with a unit of miles per hour per hour. Introductory physics classes teach students about the behavior of falling objects, measuring position in meters, velocity in meters per second, and change in velocity over time in meters per second, per second. This final measure is called acceleration: the rate of change of velocity over time:
The expression d2x/dt2 is called the second derivative of position (x) with regard to time (t). If we were to connect a second differentiator circuit to the output of the first, the last voltmeter would register acceleration:
Deriving velocity from position, and acceleration from velocity, we see the principle of differentiation very clearly illustrated. These are not the only physical measurements related to each other in this way, but they are, perhaps, the most common. Another example of calculus in action is the relationship between liquid flow (q) and liquid volume (v) accumulated in a vessel over time:
A “Level Transmitter” device mounted on a water storage tank provides a signal directly proportional to water level in the tank, which — if the tank is of constant cross-sectional area throughout its height — directly equates water volume stored. If we wer to take this volume signal and differentiate it with respect to time (dv/dt), we would obtain a signal proportional to the water flow rate through the pipe carrying water to the tank. A differentiator circuit connected in such a way as to receive this volume signal would produce an output signal proportional to flow, possibly substituting for a flow-measurement device (“Flow Transmitter”) installed in the pipe.
Returning to the car experiment, suppose that our hypothetical car were equipped with a tachogenerator on one of the wheels, producing a voltage signal directly proportional to velocity. We could differentiate the signal to obtain acceleration with one circuit, like this:
By its very nature, the tachogenerator differentiates the car’s position with respect to time, generating a voltage proportional to how rapidly the wheel’s angular position changes over time. This provides us with a raw signal already representative of velocity, with only a single step of differentiation needed to obtain an acceleration signal. A tachogenerator measuring velocity, of course, is a far more practical example of automobile instrumentation than a giant potentiometer measuring its physical position, but what we gain in practicality we lose in position measurement. No matter how many times we differentiate, we can never infer the car’s position from a velocity signal. If the process of differentiation brought us from position to velocity to acceleration, then somehow we need to perform the “reverse” process of differentiation to go from velocity to position. Such a mathematical process does exist, and it is called integration. The “integrator” circuit may be used to perform this function of integration with respect to time:
Recall from the last chapter that an integrator circuit outputs a voltage whose rate-of-change over time is proportional to the input voltage’s magnitude. Thus, given a constant input voltage, the output voltage will change at a constant rate. If the car travels at a constant velocity (constant voltage input to the integrator circuit from the tachogenerator), then its distance traveled will increase steadily as time progresses, and the integrator will output a steadily changing voltage proportional to that distance. If the car’s velocity is not constant, then neither will the rate-of-change over time be of the integrator circuit’s output, but the output voltage will faithfully represent the amount of distance traveled by the car at any given point in time.
The symbol for integration looks something like a very narrow, cursive letter “S” (∫). The equation utilizing this symbol (∫v dt = x) tells us that we are integrating velocity (v) with respect to time (dt), and obtaining position (x) as a result.
So, we may express three measures of the car’s motion (position, velocity, and acceleration) in terms of velocity (v) just as easily as we could in terms of position (x):
If we had an accelerometer attached to the car, generating a signal proportional to the rate of acceleration or deceleration, we could (hypothetically) obtain a velocity signal with one step of integration, and a position signal with a second step of integration:
Thus, all three measures of the car’s motion (position, velocity, and acceleration) may be expressed in terms of acceleration:
As you might have suspected, the process of integration may be illustrated in, and applied to, other physical systems as well. Take for example the water storage tank and flow example shown earlier. If flow rate is the derivative of tank volume with respect to time (q = dv/dt), then we could also say that volume is the integral of flow rate with respect to time: