Feedback in mic buffers

Been studying. What do you think about the use of feedback in mic buffers? Have you done any listening to mics with and without feedback? I understand a big part of Klaus’ U87 mods is removing feedback, along with selected FET and bias, replaced transformer and caps, etc.

Feedback:

U47 - no
C12 - no
M49 - yes
ELA M251 - no
KM54 - yes
U67 - yes, and the most unique I’ve seen in a mic. Tertiary winding on the OT.
U87 - yes

I have looked at every mic schematic designed by Oliver Archut - no feedback on any of them.

I plan to record acoustic guitar tomorrow night with two of my “poor man’s KM54’s”  (MV691 bodies with M94 capsule with custom tube buffer). One with feedback, one without. Should be interesting.

Curious to hear thoughts. Thanks!

Depends on the reason NFB is used. The signal level in a mic buffer is rather small and since tube distortion is level dependent, there is not much distortion to start with so it may well not be necessary to use NFB to reduce it. Noise is probably the biggest issue in a tube mic buffer and NFB does nohing to help this so another reason not to use it. However, transformers do produce distortion at low levels so the tertiary winding is a common way of linearising a transformer so perhaps that is why NFB is used in that case.

Cheers

Ian

Removing feedback in U87? Send me U87 i will send you back TLM103?  Wow ;D

Oliver circuits, especially mods are usual simple plate follower, fixed bias and 1G for high impedance section.
Only thing which i like in Oliver circuits is that he had awareness of input topology importance, unfortunately these 1G resistors weren't best option for it.
M49, U67, U87 feedbacks were implemented for equalisation of response.
Feedback in KM84 (KMi) is flat for example.
You can also find that kind of feedback in some chinese tube mikes.
Anyway i experimented a lot with feedback loops inside microphones and there are pluses and minuses
I can speculate that Neumann used feedback in mics for repeatability of production. With feedback there's a lot easier to get exactly same output level and frequency response.
With a feedback and fixed bias for example, you can use much lower values of cathode cap than without feedback.
In some tube circuits i was able to use 2-5uF instead 22uF-100uF, so there's option to use film capacitor instead electrolytic.
Also biasing point is different, and you can find optiomal point (sweet spot) with lower cathode resistance - higher output level. Output capacitor also can be lower than without feedback. LPF with cap from plate to ground - you can use much higher values (easier to set more occurate curve).
Feedback circuits are more linear (if there is no eq applied of course).

AusTex64 said:

I understand a big part of Klaus’ U87 mods is removing feedback, along with selected FET and bias, replaced transformer and caps, etc.

Click to expand...

Klaus Heyne has a very poor understanding of how feedback works. He views NFB like the arrow in the Zenon of Elea paradox, something that never reaches its target. As Ian pointed out, NFB is not always necessary. Most tube headamps do not need NFB for linearity or headroom; when NFB is used, it's for another reason; in the U67, it is used for linearizing the capsule's frequency response.
BJT's need NFB for correct linearity (low THD), FET's not so much.
A voltage-follower (cathode-follower in vacuum-state, source-follower in FET's or emitter-follower in BJT's) uses near-100% NFB.

I'm not aware of any SS design by Oliver Archut, but a couple of his tube schematics flying around show cathode degeneration (no cathode cap), which is a form of NFB.

abbey road d enfer, do you understand how the feedback in the M49 works?

My understanding is that the NFB is injected into a LPF with a breakpoint of about 3Hz (R3 5 Mohm and C3 0.01 uF).
But, since the feedback is negative, the filter will act as a HPF, cutting of a bit of the low end.

The reason that I think it works that way is because of the AC701 “Electronic tube” from Phaedrus Audio (I bought one and tested it). That “tube” has a very steep phase shift at around 10 Hz (more than 180 degrees) and it turns the NFB into a PFB below 10 Hz. The result is oscillation that goes out of control after a few seconds.

The remedy, according to Phaedrus Audio, is to remove the feedback link.
I’ve tried it, and indeed one has to remove the feedback link to stop the oscillation.

This behaviour, and the solution, was first pointed out by RuudNL, if I remember correctly.
Any thoughts on that?

stelin said:

abbey road d enfer, do you understand how the feedback in the M49 works?

My understanding is that the NFB is injected into a LPF with a breakpoint of about 3Hz (R3 5 Mohm and C3 0.01 uF).
But, since the feedback is negative, the filter will act as a HPF, cutting of a bit of the low end.

Click to expand...

You mean the M49c, surely. The earlier ones did not have this. Yes, I fully understand how it works. And you're correct in your analysis.

The reason that I think it works that way is because of the AC701 “Electronic tube” from Phaedrus Audio (I bought one and tested it). That “tube” has a very steep phase shift at around 10 Hz (more than 180 degrees) and it turns the NFB into a PFB below 10 Hz. The result is oscillation that goes out of control after a few seconds.

Click to expand...

I don't really understand your comment. The "reason it works that way" is because the designers designed it that way, period; this is a well proven concept.
The incongruous behaviour of the "AC701 Electronic Tube" that you describe is what makes the concept not working properly. It simply does not obey the stability criteria for stability of a closed-loop system.

The remedy, according to Phaedrus Audio, is to remove the feedback link.

Click to expand...

That is indeed a remedy to the oscillation problem; however, Neumann designers have included this NFB for a reason - it was supposed to be an enhancement. By disconnecting the FB loop, this advantage is lost. I agree it turns an unusable mic into a usable one, but something is probably lost along the way.

Thank you, abbey road d enfer,

I wrote "The reason that I think it works that way", not "The reason it works that way". 

I had to think about the feedback and how it works (yes, in the M49c) because of the oscillation problem.

An interesting point is that I got an email from Phaedrus Audio where they said that the feedback loop in the M49c was there to <booster> the low end of “bad AC701 tubes”.  That made me wonder.

I have since put the AC701 “electronic tubes” aside and are now using the 6S6B-V tubes instead. No phase shift and very low noise, low microphonics (both much better than the 5840) and they are very cheap. Thanks for that to, RuudNL!

And I can now keep the feedback as it was design. But, I still want to understand how the circuitry works.

stelin said:

An interesting point is that I got an email from Phaedrus Audio where they said that the feedback loop in the M49c was there to <booster> the low end of “bad AC701 tubes”.  That made me wonder.

Click to expand...

That is a tad surprizing, as I think these guys are quite knowledgeable; maybe an intern or the janitor answered the mail that day...or simply the mktg dept 

M49B also had the same feedback. Anyway cut-off point is pretty high. Rather it was designed to reduce proximity effect, years later Neumann did this also in U87.  It's big part of M49 sound. If someone doesn't like it, then i would suggest to change capacitor value to move cut-off point lower, but keep feedback to get similar level and other artifacts
If phadreus can't work with the feedback, then it isn't 100% replacement.

To summarize your responses, NFB in mics is almost always about affecting frequency response in one way or another. Which makes total sense.

AusTex64 said:

To summarize your responses, NFB in mics is almost always about affecting frequency response in one way or another. Which makes total sense.

Click to expand...

Yeap
U47Fet also use a feedback and also to modulate frequency response

Abbey, you wrote <I'm not aware of any SS design by Oliver Archut, but a couple of his tube schematics flying around show cathode degeneration (no cathode cap), which is a form of NFB.>

I found this schematic that appears to depict what you’re describing: http://www.tab-funkenwerk.com/id68.html

Will you please explain how the resistor only cathode bias is a form of NFB? Afraid I am outside my pay grade. Thank you.

AusTex64 said:

Will you please explain how the resistor only cathode bias is a form of NFB?

Click to expand...

Wow! You make me put on my teacher's hat, an exercise I have not practiced for decades. 
OK, I'll try to make it simple. Suppose a positive voltage increase on the grid, as a consequence, the tube current increases, which decreases the anode voltage, but increases the cathode voltage. As a result, the Vgk does not increase as much as the input voltage. That is a form of NFB, which samples the output current and uses it to decrease the overall gain. Since it tends to regulate the output current (not the voltage), it increases the output impedance. The input current is unchanged (in linear conditions, it is null), so is the input impedance.
Because the output Z is increased, it needs to be used only with non-reactive loads for linear frequency response; this is the main drawback. The advantages are simplicity and improved THD.
You see: no math.

AusTex64 said:

Will you please explain how the resistor only cathode bias is a form of NFB? Afraid I am outside my pay grade. Thank you.

Click to expand...

The signal current in the plate resistor also flows through the cathode resistor. This develops  a voltage across the cathode resistor. The voltage between the the grid and 0V is the grid/cathode voltage (which is what the tube amplifies) plus this signal voltage across the cathode.

so Vin = Vgk + VcathodeR

So the input to the tube is:

Vgk = Vin - VcathodeR hence the negative feedback.

Cheers

Ian

@Abby and Ian

Nice explanation but it doesn't work on linked schematics that way.  There is no NFB or it is very low.

moamps said:

@Abby and Ian

Nice explanation but it doesn't work on linked schematics that way.  There is no NFB or it is very low.

Click to expand...

Amittedly, 29 ohms in the cathode does not provide significant NFB.

Thanks guys for for patiently explaining that.

old thread but a goodie.

ln76d said:

Yeap
U47Fet also use a feedback and also to modulate frequency response

Click to expand...

I have been investigating simple fet mic head amps and would appreciate some discussion about the fet47 circuit especially; as I understand it the response should be flat
why so much silicon?
how is the frequency reponse modulated?

https://cdn.groupbuilder.com/groupdiy/u/39511/58d1402a02abc.pdf

shabtek said:

I have been investigating simple fet mic head amps and would appreciate some discussion about the fet47 circuit especially; as I understand it the response should be flat

Click to expand...

NFB there is of the voltage-to-current type, via R2/C2 and R3/C3.

why so much silicon?

Click to expand...

T7: supercap filter for B+ rail
T5: constant current source for output stage
T6: output
T3: supercap filter for drain of FET
T4: constant voltage source for T2
T2: common-base stage constituting cascode with T1
T1: input FET
One could speculate that Neumann engineers had a contest with AKG regarding who had the most transistors in their mics. 

how is the frequency reponse modulated?

Click to expand...

As far as I can see, it is not. The response is flat down to about 40Hz, where there is a 6dB/octave roll-off.
Here NFB is used in an attempt to reduce headamp distortion. This arrangement is called a charge amp. Actually, it has an inherent flaw, which is a source of distortion, but since it appears at very high SPL, and produces mainly 2nd harmonic, many appreciate it. It has also the advantage of making the circuit somewhat less sensitive to the effects of dirt and humidity on the diaphragm.

thank you