The Moog Modular designs were made during a time when the new silicon transistor was relatively expensive. So, these circuit designs show an economy of function. There are other insights though, looking again now, compared with the first looks I did in 2001. Some of these insights may only be EE techniques of the early 1960s, but others may be Robert Moog design genius. The Moog Modular 900 series designs do not seem typical of 1960s engineering, based on circuit encyclopedias and other reference information like transistor design books from the 1960s. And, some of these designs are not borne out either from internet search 50 years later. There has likely been information loss since the 1960s, the Internet for sure is not omnipotent as in information source, esp. for data preceeding it. So, kudos to Dr. Moog appear to be in order. What can be learned from these techniques?
1. When I made the first voltage adder in 2001, it now appears that I was actually using a Moog “output” type of circuit, and not the actual “adder” type circuit. This bears more scrutiny, and for both type of circuits.
2. The differential output circuits are powerful, and need to be looked again. The output stages are analogous to the LM3900 line driver I developed this year, but those circuits lack the differential benefit of the Moog circuits. These "current-mode" stages though do work quite well, offering excellent bandwidth, programmable signal swing, load-driving ability, and low-distortion. Also, these circuits were stable operating open loop, and Moog's VCA was dependant on this type of circuit. The differential gain aspect seem well-worthy also.
3. Moog circuits took advantage of resistor voltage dividers to obtain levels for baselines, as the levels between differential stages are set for operation. Kind of clever, and seems to work. Take note of that.
4. Then, the voltage adder has some hidden surprises. One is that it had both current or voltage out options, something I did not see. It does use feedback for gain stability, and it provides a virtual ground. It should be capable of very accurate addition. More particularly, the output fed to an antilog transistor in adjoining circuits.
5. A particular thing to note was how a diode offset the base of the antilog transistor with a range trimmer, so as to enable the adder circuit to submit either voltage or current to a transistor emitter which is very close to 0 volts. Moog didn't use matched pairs in every circuit, nor any temperature compensation until it now appears the Moog 921 VCO. This temperature compensation was not a PTAT type, but rather one of running the antilog transistors at a constant temperature, via the Fairchild µA 726 device. Temperature compensation is far less critical for VCAs ... which may rely more on balance trims to control offsets so as to enable use for control voltages.
6. The +12V/-6V power supply approach used by the 1960s Moog Modular circuit designs also seems interesting. Probably lowers power requirements, and provides sufficient headroom for at least audio signal ranges. There is some likelihood as well that a maximum voltage range of 18V matches better the breakdown limits of some silicon transistors of the day. In other words, a good margin of safety. The 2N3392 had a 25V limit, and the limit for the 2N4058 was -30V. So, a ±12V supply would be a little bit close to these breakdown limits.
7. The current ranges used for a lot for the transistors in mid-1960s Moog designs seems kind of high. However, the transistors listed above have a pretty good low-current β of ≈200. So, these seem like high-end transistors of the day. With later-generation silicon transistors having higher β at lower currents, the same circuit toplogies should be possible with higher resistor values for the same general voltages otherwise.
8. The Moog VCA and VCF definitely relied on transconductance, but not like the Gilbert precision multiplier, which came perhaps 5 years later. But the Gilbert gain cell was always in an integrated circuit context and Moog's mid-1960s circuits used discrete transistors (even if some were matched). In particular, the Moog VCFs used transonductance. Need to look into that approach from the discrete BJT angle.
Some very interesting schematics are now available from the Moog Foundation:
Comments
Post a Comment