VCAs are wide open world. There are many designs and topologies to choose from, and designs are accessible for divergent applications. This one is a little different.
I know from my work on synthesizers in the 1970s that I did not like the "sound" of the CA3080 and LM13700 OTA-based VCAs. Their sound seemed sterile: "cold," and "electronic." More particularly, there was a constant tradeoff with CV feedthrough, to keep "clicks" from happening when there was 0V audio input and the VCA control voltage fed with a regular ADSR envelope.
I also have investigated LM3046 and LM394 VCAs. Most of these circuit designs I have seen used numerous trimmers to align their performance to that obtainable from the precision NPN transistors used to perform the transconductance-based multiplication required. I also have explored positive-voltage only VCAs with PNP matched pairs and the LM3900. I have also considered later-generation precision ICs like the SSM2164 VCAs, but prefer to use my own circuit designs.
Over time, I came to dislike various things in implementations for VCAs intended for use in electronic music. While arguably there is some overlap in required performance characteristics, electronic music VCAs can be quite distinguished from VCAs used for companders, compression, de-essers, noise reduction, and sound board fader automation.
For one thing, I think VCA harmonic distortion is almost irrelevant, as long as it's not blatently bad. A lot because signals in electronic music are already highly involved with harmonic generation. If the VCA contributes perhaps a 3rd harmonic of < -30dBc and less elsewhere, it's probably good enough. An audiophile grade VCA is not required here.
I also dislike AC signal paths for most applications in electronic music. It's very desireable to have inputs and outputs capable of being used either for audio or for control voltage modulation purposes. Practically doubles the utility of VCAs in a modular system. This then means very clean control of DC feedthrough is required.
With all the circuits I'd examined and/or tested previously, all of them required trimmers to achieve necessary balance, offset, limited feedthrough, etc. In particular, in circuits I breadboarded, or had made into PCBs, multi-turn trimmers were often the best solution so as to be able to "tune" the VCA performance correctly. And trims for a VCA often interacted. The lesson for me is that I really do not like VCA trimmers!
If one is already using a highly matched transistor pair (or set), could a VCA circuit topology be created which uses only precision components, such that trimmers would not be required at all?
The following VCA circuit applies a totally different engineering tradeoff: precision components set the performance, instead of using trimmers to align the performance.
The circuit design effectively preserves inherent balance in the matched pair transistors, while providing precise control voltage interoperability via the use of precision reference voltages. It really has zero trimmers, using instead precision tolerance resistors, ±1% and ±0.1%. The active semiconductors are a MAT04 matched NPN quad, and a precision OPA2227A dual op amp, as well as a pair of LM4040AIZ-2.5 voltage references. The voltage references set ±2.5V levels with a precision of ±0.1%. These references are used to command a differential 500µA reference current for the transconducting transistor pair, as well as to set a precise offset voltage for the 0-5V control voltage range.
I've been at this system design since 1981, and so over the years, I have collected a number of old stock devices. I obtained the MAT04s circa 2000, and they fit the role perfectly. They are less available today than 20 years ago, but alternatives are like THAT 300, and others. However, the LM4040AIZ and OPA2227A devices are still quite available. Other alternative choices could be made keeping the topology intact, but a quad NPN seemed preferrable to a using two LM394CHs, for temperature stability. The quad is also more compact in PCB space.
I intend to use all through-hole technology in my synthesizer, to allow repair of circuit boards for many years to come. I also find it useful to be able to see the parts, and touch them without robotic handlers or other specialized SMT equipment.
This VCA I really listened to: I set up drone patches with test and measurement equipment, and listened for hours. And I really liked the sound I heard.
I have made breadboard measurements, which I'll summarize briefly. This circuit worked so well though, I want to get it into a PCB for a module ASAP. Characterization scope fotos from a real PCB I think will reflect the obtainable results more accurately than my superstrip breadboard does.
But in general, here's how things looked:
- 100% DC signal path for audio and control voltage path.
- ±2.5V I/O, with 0-5V linear control voltage. This is my preferred setup for electronic music. The design could use other signal levels with equal aplomb.
- Gain is extremely close to 1.0 at peak control voltage, just slightly larger by a few percent.
- Wideband sinusoidal response to at least 50 KHz without attenuation loss.
- No waveform distortions for pulse, triangle, or sawtooth signals; edges are rendered with fast flat ramps. There were no deleterious overshoot or undershoot abberations. Extremely clean.
- Low noise. I cranked a monitoring amplifier up, and the VCA sounds quiet with no signal input.
- I don't have an Audio Precision measurement system, but my Tek TBS 1064 oscilloscope FFT response showed a 3rd harmonic spur for a 440 Hz fundamental to be ≈-45 dBc at full ±2.5V output. A 5th harmonic was nearly in the mud, and no other harmonics appeared to be present -- so they are also way down too.
- When the input signal is 0V, the output is also 0V, no matter what is happening with the control voltage input.