We recently saw how certain audio power amplifiers can be used as oscillators. This Design Idea shows how those same parts can be used for simple amplitude modulation, which is trickier than it might seem.
Wow the engineering world with your unique design: Design Ideas Submission Guide
The relevant device is the TDA7052A, which we explored in some detail while making it oscillate. It has a so-called logarithmic gain-control input, the gain in dBs being roughly proportional to the voltage on that pin over a limited range.
However, we may want a reasonably linear response, which would mean undoing some of the chip designers’ careful work.
First question: why—what’s the application?
The original purpose of this circuit was to amplitude-modulate the power output stage of an infrasonic microphone. That gadget generated both the sub-10‑Hz baseband signal and an audio tone whose pitch varied linearly with it, allowing one to hear at least a proxy for the infrasonics. The idea was to keep the volume low during relatively inactive periods and only increase it during the peaks, whether those were positive or negative, so that frequency and amplitude modulation would work hand in hand.
The two basic options are to use the device’s inherent “log” law (more like antilog), so that the perceived loudness was modulated, or to feed the control pin with a logarithmically-squashed signal—the inverse of the gain-control curve—to linearize the modulation. The former is simpler but sounded rather aggressive; the latter, more complicated but smoother, so we’ll concentrate on that. The gain-control curve from the datasheet, overlaid with real-life measurements, is shown in Figure 1. Because we need gain to drive the speaker, we can only use the upper, more bendy, part of the curve, with around 26 dB of gain variation available.
Figure 1 The TDA7052A’s control voltage versus its gain, with the theoretical curve and practical readings.
For accurate linear performance, an LM13700 OTA configured as an amplitude modulator worked excellently, but needed a separate power output stage and at least ±6-V supplies rather than the single, split 5-V rail used for the rest of the circuitry. An OTA’s accuracy and even precision are not needed here; we just want the result to sound right, and can cut some corners. (The LM13700’s datasheet is full of interesting applications.)
Next question: how?
At the heart of this DI is an interesting form of full-wave rectifier. We’ll look at it in detail, and then pull it to pieces.
If we take a paralleled pair of current sources, one inverting and the other not, we can derive a current proportional to the absolute value of the input: see Figure 2.
Figure 2 A pair of current sources can make a novel full-wave rectifier.
The upper, inverting, section sources current towards ground when the input is positive (with respect to the half-rail point), and the lower, non-inverting part does so for negative half-cycles. R1 sets the transconductance for both stages. Thus, the output current is a function of the absolute value of the input voltage. It’s shown as driving R4 to produce a voltage with respect to 0 V, which sounds more useful than it really is.
Conventional full-wave rectifiers usually have a voltage output, stored on a capacitor, and representing the peak levels. This circuit can’t do that: connecting a capacitor across R4 merely averages the signal. To extract the peaks, another stage would be needed: pointless. By the way, the original thoughts for this stage were standard precision rectifiers with incorporated or added current sources, but they proved to be more complicated while performing no better—except for inputs below ~5 mV, where they had less “crossover distortion.”
The maximum output voltage swing is limited by the ratios of R4 to R2 (or R3). Excessive positive inputs will tend to saturate Q1, so VOUT can approach Vs/2. (The transistor’s emitter is servoed to Vs/2.) With R4 = R2 = R3, negative swings saturate Q2, but the ratio of R3 and R4 means that VOUT can only approach Vs/4. Q1 and Q2 respond differently to overloads, with Q2’s circuit folding back much sooner. If R2, R3, and R4 are all equal, the maximum unclipped voltage swing across R4 is just less than a quarter of the supply rail voltage.
Increasing R1 and making R4 much greater than R2 or R3 allows a greater swing for those negative inputs, but at the expense of increased offset errors. Adding an extra gain stage would give those same problems while needing more parts.
Applying the current source to the power amp
Conclusion: This circuit is great for sourcing a current to ground, but if you need a linear voltage output, it’s less useful. We don’t want linearity but something close to a logarithmic response, or the inverse of the power amp’s control voltage. Feeding the current through a network containing a diode can do just that, and the resulting circuit is shown in Figure 3.
Figure 3 Schematic of a power amplifier that is amplitude-modulated using the dual current source.
The current source is just as described above. With R1 = 100k, the output peaks at 23 µA for ±2.5 V inputs. That current feeds the network R4/R5/D3, which suitably squashes the signal, ready for buffering into A2’s Vcon input. The gain characteristic is now much more linear, as the waveforms in Figure 4 indicate. The TDA7052A’s Vcon pin normally either sinks or sources current, but emitter follower Q3 overrides that as well as buffering the output from the network.
Figure 4 Some waveforms from Figure 3, showing its operation.
To show the operation more cleanly, the plots were made using a 10-Hz tri-wave to modulate a 700-Hz sine wave. (The target application would have an infrasonic signal—from, say, 300 mHz to 10 Hz—modulating a pitch-linear audio tone ranging from about 250 to 1000 Hz depending on the signal’s absolute level.)
Some further notes on the circuitry
The values for R4/R5/D3 were optimized by a process of heuristic iteration, which is fancy-speak for lots of fiddling with trimmers until things looked right on the ’scope. These worked for me with the devices to hand. Others gave similar results; the absolute values are less important than the overall combination.
R7 and R8 may seem puzzling: there’s nothing like them on the PA’s datasheet. I found that applying a little bias to the audio input pin helps minimize the chip’s internal offsets, which otherwise cause some (distorted) feedthrough from the control voltage to the outputs. With a modulating input but no audio present, trim R7 for minimum signal at the output(s). The difference is barely audible, but it shows up clearly on a ’scope as traces that are badly slewed.
The audio feed needs to come from a volume-control pot. While it might seem more obvious to incorporate gain control in the network driving A2.4—after all, that’s the primary function of that pin—that proved over-complicated, and introduced yet more temperature effects.
Temperature effects! The current source is (largely) free of them, but D3, Q3, and A2 aren’t, and I have made no attempt to compensate for their contributions. The practical solution is to make R6 variable: a large, user-friendly knob labeled “Effect”, thus turning the problem into A Feature.
A2’s Vcon pin sinks/sources some (temperature-dependent) current, so varying R6 allows reasonable, if manual, temperature compensation. Because its setting affects both the gain and the part of the gain curve that we are using, the effective baseline is shifted, allowing more or less of the audio corresponding to low-level modulating signals to pass through. Figure 5 shows its effect on the output at around 20°C.
Figure 5 Varying R6 helps compensate for temperature problems and allows different audible effects.
Don’t confuse this circuit with a “proper” amplitude modulator! But for taking an audio signal, modulating it reasonably linearly, and driving the result directly into a speaker, it works well. The actual result can be seen in Figure 6, which shows both the detected infrasonic signal resulting from a gusty day and the audio output, whose frequency changes are invisible with the timebase used, but whose amplitude can be seen to track the modulating signal quite nicely.
Figure 6 A real-life infrasonic signal with the resulting audio modulated in both frequency (too fast to show up here) and amplitude.
—Nick Cornford built his first crystal set at 10, and since then has designed professional audio equipment, many datacomm products, and technical security kit. He has at last retired. Mostly. Sort of.
Related Content
- Power amplifiers that oscillate— Part 1: A simple start.
- Power amplifiers that oscillate—deliberately. Part 2: A crafty conclusion.
- Revealing the infrasonic underworld cheaply, Part 1
- Revealing the infrasonic underworld cheaply, Part 2
- Ultra-low distortion oscillator, part 1: how not to do it.
- Ultra-low distortion oscillator, part 2: the real deal
The post Why modulate a power amplifier?—and how to do it appeared first on EDN.
