CADAC remote mic preamp - most parts ever used?

[cuelist]

Just found this "gem" in my ever growing schematics collection.

It's a design for CADAC by the late Barry Porter (of Trident A-range fame). This design has some interesting solutions:

- Gain switching using PhotoMOS relays (AQV251's) - these have very low on resistance indeed. Each relay is driven by a darlington pair and a 'trigger circuit' than ensures the correct make/break timing.

- Sixteen 2SB737's due to the fact that Barry was somewhat obsessed with ultra low distortion and insisted on a true differential circuit.

- Separate AC and DC feedback loops.

- The gain switching also affected the AC feedback around the op amps (see relay RL6 and RL17 for example).

 

[JohnRoberts]

That looks like a lot of 737s to parallel. The noise for one is down around .3nV rt Hz,

I've looked at ways to switch gain at mic level and never was excited enough to melt solder even with less parts than he used.

How does it sound?

JR

 

[jdbakker]

- Sixteen 2SB737's due to the fact that Barry was somewhat obsessed with ultra low distortion and insisted on a true differential circuit.

That's done for noise rather than distortion, no? Pity he had to throw away some of the '737s excellent noise performance with those emitter resistors, though.

It's interesting to see that this is one of the few examples of this topology using voltage feedback, as opposed to the current feedback in the more common Green/Cohen/Valley People/SSL9k circuit. VFB (executed this way) takes twice as many transistors as CFB, leading to 3dB more noise for the same transistor count. To achieve parity you need again twice as many transistors.

JDB.
[a CFB input with equivalent noise performance would thus have taken only four '737s, but CFB might have exposed nasty nonlinearities in those OptoMOS beasties]

 

[cuelist]

That's done for noise rather than distortion, no? Pity he had to throw away some of the '737s excellent noise performance with those emitter resistors, though.

As far as I remember, Barry told me that he used this approach because the current feedback alternative had much worse common mode distortion.

 

[JohnRoberts]

He is not running much current density at roughly 1/2 mA per, perhaps to mitigate the addition of noise current terms at input.

I looked at an approach using multiple LTPs in parallel with different fixed gains, that could be turned on/off by alternately steering the bias current to desired active pairs., but I became discouraged by the high parts count. So I never got to wrestle with the other obvious problems.

JR

 

[abbey road d enfer]

- Sixteen 2SB737's due to the fact that Barry was somewhat obsessed with ultra low distortion and insisted on a true differential circuit.

That's done for noise rather than distortion, no? Pity he had to throw away some of the '737s excellent noise performance with those emitter resistors, though.

Maybe for VLF distortion due to temperature varying with instant current. I remember some designers were concerned with this aspect of things in the 80's. Some designed moving-coil pre-pres with medium power transistors just for this reason (and the low rbb')

 

[Samuel Groner]

Thanks for the schematic, anything more from him (I have the EQ thing) is appreciated.

Quote from his (now off-line) homepage:

One particular microphone amplifier which I designed is quite interesting. The requirement was for a stage which could be placed remotely from the control room, while being capable of having its gain changed. Not too difficult, except that the gain change had to be totally silent, so that if necessary it could be carried out during a performance without the slightest hint of a click occurring.

The initial choice of gain switching device was a small relay, but in spite of trying many different types, they all introduced a click. Another problem was caused by a large electrolytic capacitor which was necessary to isolate the gain switching. No matter which type was used, the leakage allowed a few microvolts of DC to reach the gain switching element, which also contributed to the click level.

The problem was solved by a two-pronged attack. Firstly, the large capacitor had to go, yet the gain switching had to take place where there was no DC present. This was achieved by using separate AC and DC feedback loops, so that the AC path could be isolated by relatively small, non-polarised electrolytic capacitors, which have a much lower leakage.

Because the DC feedback path has a resistance of some 50k ohms, it was necessary to use a long tailed pair for the input stage, but this configuration introduces more noise that the more usual single ended approach. As a consequence, it was found that each transistor shown above needed to be four devices in parallel in order to achieve the aimed for 1dB noise figure.

The next problem was to find a way of switching the gain setting resistors with total silence. Again, relays were tried, but it was then discovered that even with the smallest types, the movement of the contacts in the magnetic field of the coil was generating a few microvolts of noise, so had to be rejected. In the end, small solid state relays with an “On” resistance of less than 1ohm were used. It was possible to ramp the switching action so they changed state over a few milliseconds, and these proved to be totally silent in operation. Problems solved after many long days at the bench...

Samuel

 

[abbey road d enfer]

Thanks for this historical explanation, Samuel.

 

[JohnRoberts]

Thanks... It's always good to know the back story..

JR

 

[dmills]

If that is what I think it is then the thing to realise is that they were putting 16 of them in the M16 remote mic preamp (256 2SB737 per box).

Regards, Dan.

 

[cuelist]

Yes, most probably.