Schoeps/Alice circuit white noise ... why did this fix work?

[midwayfair]

I hate it when this happens: I fix something but can't figure out why it was wrong int he first place and the fix doesn't make much sense.

I built a couple schoeps circuits on the same layout. I matched the transistors, resistors, everything that needed to be matched. One had some white noise and the other was perfectly silent. The signal output was the same on both.

[EDIT: All the changes I made: I added a 47R+22nF LPF to the output (pretty common, I think); I used 2K7 instead of 2K4 for the source and drain resistors; there is now a 220nF from the base of Q2 to the +12V regulated; and there is a switch to put a 10K from the base of Q2 to the base of Q3 (but the white noise was identical in both switch positions). Also, the transistors I left in are 2N4403, which are low-noise types.]

I tracked the noise down to the base of the PNP that follows the drain output.

I rechecked all my components, replaced all of the transistors (I tried pairs of multiple transistor types -- incl. 3906, 4403, and BC328/338), and even used a socket to audition transistors that weren't matched.

In the end, the only thing I could find that would kill the while noise was putting a cap from the base of that transistor to ground or the power rail, i.e., creating a low-pass filter in only one band. I used 220nF -- it was the smallest I could go to get the second mic down to comparable noise levels.

This ought to unbalance the output, which should give me MORE noise I would think. It should also screw up the frequency contentbut I ran it through my DAW and a frequency analyzer with a banjo and compared with the other mic. I get approximately the same content as far up as 17KHz, beyond which the banjo isn't really registering.

The only thing I can think of is that somehow I ended up with more gain on the upper band, and the filter on that band brought the total gain back down to identical levels.

EDIT: a little more experimentation ... I was able to get the other mic a little quieter doing the same thing ... specifically, though, it's the cap from the base of the upper transistor to the +12V power rail that's the absolute quietest operation -- not the collector (which is the "upside down ground") and not circuit/chassis ground. Connecting the lower (source-output) base makes the noise worse. So maybe something about this version of the circuit is NOT balanced and requires a correction.?

 

[ln76d]

Did you use proper orientation for PNP transistors? You are using different types and some probably have different pinout than others.  Did you cleaned PCB after soldering? Cold solder joints? Zener Wattage (higher than 0,5W?)?
Best what you could is to apply signal to the input and track whole audio path.

 

[midwayfair]

Did you use proper orientation for PNP transistors? You are using different types and some probably have different pinout than others.  Did you cleaned PCB after soldering? Cold solder joints? Zener Wattage (higher than 0,5W?)?
Best what you could is to apply signal to the input and track whole audio path.

Definite yes on the orientation. I always check the datasheet for the exact part I bought and verified the pinout on a breadboard. (They're pretty much all Fairchild anyway.) I even tried intentionally putting them in backwards at one point just in case. The only transistors with a difference pinout of those that I mentioned are the BC types (which are flipped 180degrees).

Zener is a half watt and the cap in parallel with it is a good Nichicon low ESR.

I rechecked all the solder joints.

It's perfboard, so there's only so much cleaning that can be done, but the second half of the circuit (where the noise is coming from) is a low-impedance situation, so I would be astounded if any stray flux could introduce white noise. The 1G and the gate are elevated off the perfboard and the 1G and gate leg got an alcohol scrub afterward.  The wires are as short as I could make them.

I will say that the white noise is a consistent problem with me with the simplified versions of the Schoeps circuit. This was my fifth attempt and though it's better than the first time, I can't find anything that I'm doing wrong -- e.g. the 4403 ought to be one of the quietest transistors I can use here but it made no discernable difference. I can't imagine that people are finding the level of noise that I'm experiencing from the stock circuit acceptable on all but the loudest sources. It's certainly much higher than the version of the circuit that's in the e.g. MXL990 with the oscillator circuit.

With the cap added, there's slightly more hiss than I get from the Aurycle mic I built recently, which is also a slightly worse S:N than my modded MXL990. In other words, it's usable now, though not the first thing I'd reach for (not that I was looking for that -- these were just fun/backups).

p.s. See my notes I added below the schematic -- I don't think any of those parts are contributing to the noise based on where it's being introduced, however, but one of them might.