MagnetoSound
Well-known member
Matador said:
No argument from me, I would do exactly the same - although I think Rossi's argument has illustrated just why bypassing the drain resistor is the best approach to the mod, if you decide to do it.
negative Feedback in a Schoeps circuit?
- Thread starter maye4
- Start date
Help Support GroupDIY Audio Forum:
No argument from me, I would do exactly the same - although I think Rossi's argument has illustrated just why bypassing the drain resistor is the best approach to the mod, if you decide to do it.
RuudNL
Well-known member
Rossi said:
Yes, I do agree with you that the 22 nF at the output is a heavy load for HF audio.
If the only reason is to avoid (V)HF entering the circuit, it would be possible to reduce the value to 1000 pF or so. In that case the load for HF audio would be less.
A couple of small inductors instead of the output resistors might also help.
I am not sure if you can speak of 'distortion' when you put a capacitor between the drain and the source. The transient response will suffer of course, but that is the case with every integrator. (Low pass filter)
Do you increase distortion if you turn the treble control on your amplifier to the left?
I will do some distortion measurements on a dummy circuit and see what happens.
It might not be the most correct way to match a K67 style capsule, but it is the audible result that counts.
Will be continued...
RuudNL said:
It's been a while since I last worked on Schoeps-style circuits. I think I tried the source-drain cap method once and didn't like it. I may be wrong, but my feeling is that it is not much different from adding caps from drain and source to ground. Basically you have the two two signal legs fighting each other at higher frequencies. So I'm talking about nonlinear distortion.
Anyway, I'm curious to see your findings & measurements.
It's in the Yahoo MicBuilders Group. You have to join.gary o said:
I won't go into this but there is absolutely no need for the voltage on both output pins to be the same. Only the impedance on both legs needs to be matched. Once you realise this, you can do a lot with mods to simple circuits like the Schoeps which BTW is NEVER strictly balanced.
Is there a disadvantage? Yes. Dip Ing. Wuttke uses a very clever feature of FETs in his very subtle circuit to reduce distortion by careful trimming. But hardly anyone takes the trouble to do this. Scott Wurcer talks about this in his Linear Audio series.
But if you do stuff like I do in "ChinaMod+U87hybrid", the THD is still better than a KM84, or other "single ended" solutions, some much more complex than this.
Also, there is a noise penalty using the Neumann feedback circuit.
RuudNL
Well-known member
Just a thought...
It may be that the amplitude on the drain of the 'classical Schoeps' circuit is too low to use it as a feedback signal (superimposed on the polarisation voltage) as in the U87 compensation circuit.
But...if we add a low-current opamp to amplify the drain signal, so we have a feedback signal with a higher amplitude?
Maybe in this case the U87 feedback circuit could work (better).
I will have to do some test to see if the drain signal 'as it is now' can be used for the frequency compensation.
(Because: 100% feedback would mean no amplification at all!)
Maybe the best thing is to use a 'real world' circuit, so leave the capsule in place during the measurements.
Then inject the test frequencies through a small (<= 1/10 C-capsule) capacitor and see what happens!
In listening tests I noticed that the HF boost in most China capsules/microphones could be compensated by reducing 2 dB at 10 KHz. and 5 dB at 15 KHz.
It would be nice if we could improve the (no too bad) China microphones, with a relative simple add-on circuit.
Interesting enough to invest some time in this project!
And YES, I fully agree that you can NOT change a bad microphone into a good one by just modifying the frequency response. But what we can do, is make a reasonable sounding microphone sound better!
By the way: what a great forum this is! Lots of people with knowledge and experience who are all willing to share their expertise, without competition!
It may be that the amplitude on the drain of the 'classical Schoeps' circuit is too low to use it as a feedback signal (superimposed on the polarisation voltage) as in the U87 compensation circuit.
But...if we add a low-current opamp to amplify the drain signal, so we have a feedback signal with a higher amplitude?
Maybe in this case the U87 feedback circuit could work (better).
I will have to do some test to see if the drain signal 'as it is now' can be used for the frequency compensation.
(Because: 100% feedback would mean no amplification at all!)
Maybe the best thing is to use a 'real world' circuit, so leave the capsule in place during the measurements.
Then inject the test frequencies through a small (<= 1/10 C-capsule) capacitor and see what happens!
In listening tests I noticed that the HF boost in most China capsules/microphones could be compensated by reducing 2 dB at 10 KHz. and 5 dB at 15 KHz.
It would be nice if we could improve the (no too bad) China microphones, with a relative simple add-on circuit.
Interesting enough to invest some time in this project!
And YES, I fully agree that you can NOT change a bad microphone into a good one by just modifying the frequency response. But what we can do, is make a reasonable sounding microphone sound better!
By the way: what a great forum this is! Lots of people with knowledge and experience who are all willing to share their expertise, without competition!
There are several myths about the very subtle Schoeps circuit. It works very well but not for the reasons people believe.Ishi said:
Why do you need balanced voltages?
There is no need for balanced voltages. Look at Douglas Self & Rod Elliot sites for simple explanations.
What you need is balanced impedances on both legs.
But the Schoeps circuit is NEVER perfectly balanced. The reason why it works so well compared to "properly balanced" circuits is that it is also very low impedance; about 25R on a good implementation. That means the slight inherent imbalance is very small compared to the imbalances of other circuits at high impedance.
Also, it is very low noise compared to the Neumann feedback circuit.
You are right that same impedance in both legs is what is required. Many mics these days are single ended with impedance balanced output.
I don't know why you say the Schoeps circuit is never perfectly balanced. When you implement it correctly, it is. The phase splitter produces unity gain on both legs, and the emitter followers on the output are identical, thus producing same signal level and the same output impedance. So where's the unbalance? Note that in the original schematic the PNP transistors are even matched. In China-implementations they aren't, but you're not likely to use an MXL603 in large orchestra recordings with 100m cable runs.
A KM84 has about the same self-noise as a Schoeps mic. A U87A (okay, large diaphragm, but with the ominous frequency dependent NFB circuit) is 12 dB-A in cardioid mode. That's not as low as you can go with newer circuits, but lower than a Schoeps and lower than you need in practical applications. Actively driving two legs has a small noise penalty, too. Which is why a lot of newer mic circuits (e.g. KM184) work single ended, transformerless, impedance balanced. Even so, the KM184 is just one dB quieter than a KM84, because the capsule really is the limiting factor. You do get a lot more headroom (max SPL) with a transformerless circuit, though.
While we haven't resolved the (nonlinear) distortion question of your approach, I see another potential problem. Your bypass cap on the drain resistor changes the frequency response in the drain leg only, but not in the source follower leg. In the preamp's mic input both signals are summed, and now the phase difference, created by EQing one leg only, may become audible in an unpleasant way. I say may - it's something to listen for.
I don't know why you say the Schoeps circuit is never perfectly balanced. When you implement it correctly, it is. The phase splitter produces unity gain on both legs, and the emitter followers on the output are identical, thus producing same signal level and the same output impedance. So where's the unbalance? Note that in the original schematic the PNP transistors are even matched. In China-implementations they aren't, but you're not likely to use an MXL603 in large orchestra recordings with 100m cable runs.
A KM84 has about the same self-noise as a Schoeps mic. A U87A (okay, large diaphragm, but with the ominous frequency dependent NFB circuit) is 12 dB-A in cardioid mode. That's not as low as you can go with newer circuits, but lower than a Schoeps and lower than you need in practical applications. Actively driving two legs has a small noise penalty, too. Which is why a lot of newer mic circuits (e.g. KM184) work single ended, transformerless, impedance balanced. Even so, the KM184 is just one dB quieter than a KM84, because the capsule really is the limiting factor. You do get a lot more headroom (max SPL) with a transformerless circuit, though.
While we haven't resolved the (nonlinear) distortion question of your approach, I see another potential problem. Your bypass cap on the drain resistor changes the frequency response in the drain leg only, but not in the source follower leg. In the preamp's mic input both signals are summed, and now the phase difference, created by EQing one leg only, may become audible in an unpleasant way. I say may - it's something to listen for.
RuudNL
Well-known member
I would highly recommend to match the PNP transistors.
In China microphones I have seen DC offsets of more than 200 mV!
I am sure a transformer coupled input won't like this...
Personally I match the transistors for Hfe and Vbe.
Also I select them for lowest noise.
In China microphones I have seen DC offsets of more than 200 mV!
I am sure a transformer coupled input won't like this...
Personally I match the transistors for Hfe and Vbe.
Also I select them for lowest noise.
I'v not fully understand it.
But i assume the impedance has something to do with how they are grounded(through what resistor and cap)?
in my emulation,i managed to match 2 signals almost perfectly-phase and frequency...
Waiting for my parts...will know the results in a few days...
But i assume the impedance has something to do with how they are grounded(through what resistor and cap)?
in my emulation,i managed to match 2 signals almost perfectly-phase and frequency...
Waiting for my parts...will know the results in a few days...
Attachments
RuudNL
Well-known member
What you don't want is DC between the two wires to the amplifier stage.
This is exectly what will happen if you use 'just two' transistors.
The DC offset may even cause core magnetisation in transformer coupled input stages.
Schoeps marked the transistors with "*", meaning the transistors should be a matched pair.
This is exectly what will happen if you use 'just two' transistors.
The DC offset may even cause core magnetisation in transformer coupled input stages.
Schoeps marked the transistors with "*", meaning the transistors should be a matched pair.
MagnetoSound said:
I think MagnetoSound is right。If the Jfet ' R is small enough compared to R2 and other resistors used in filters,then Jfet ' R is neglectable.Which means D and S is shorted.In reality,the difference of D and S is very small after appling eq.This is prooved by emulation.
I'm away from home for some time so can't explain everything but ..Rossi said:
The 2 emitter followers are fed from different Z; one from the FET source so Lo Z and the other from the drain. You may like to SPICE a Schoeps circuit and measure the output Z of the 2 legs. Make sure you use good FET & BJT models.
One reason for good performance from Dip. Ing Wuttke is that even if his circuit is out by 20%; eg 20R & 25R, this is only an imbalance of 5R.
It is the absolute imbalance which is critical for common mode rejection, not the relative (or %) imbalance. So this crap 20% imbalance in the Schoeps is equivalent to 5% imbalance of a 100R Zo circuit. See the Jensen webpage for more pontification .. especially about the 'balanced bridge' concept of imbalance.
ricardo said:
I did spice the Schoeps circuit often times. The fact that the output Z of the two FET legs is not exactly equal does not automatically create an unbalance in the output Z. Output Z is determined by the EF stage (including buildout resistors and output caps), and as long as the BJTs are matched (as well as some other components) output Z is equal for both pins.
However, in my sims (using SIMetrix), the source leg HF response is not affected by the suggested drain resistor bypass cap.
I don't doubt the excellence of the original Schoeps circuit. The point in question is, if the drain resistor bypass mod does not compromise its performance.
RuudNL
Well-known member
The output impedance of an emitter follower is:
Zout = Re / (hFE +1)
As long as the Hfe of the PNP transistors is high, the driving impedance (the impedance of the FET) does not have a big influence on the output impedance.
.
Zout = Re / (hFE +1)
As long as the Hfe of the PNP transistors is high, the driving impedance (the impedance of the FET) does not have a big influence on the output impedance.
.
