KM54 feedback circuit

[AusTex64]

I’m trying to understand the feedback circuit in KM54. Here’s a schematic:

http://repforums.prosoundweb.com/index.php?action=dlattach;topic=36938.0;attach=2509

Specifically what does R2 do? I understand “abgliechen” means adjustment, balancing, tuning. What exactly is being tuned? Freq response? Gain? If I understand this feedback, it should reduce more HF than low. The KK54 capsule has a HF rise at 8K, and I think this feedback reduces some of that bump. I’ve been discussing this on Klaus’ forum, and a Neumann old timer posted a fascinating Neumann document that shows what the freq response of the buffer should be. It says  -1db overall, and -3db at 16K. http://repforums.prosoundweb.com/index.php?action=dlattach;topic=36938.0;attach=2754

I am trying to implement this feedback scheme on some mics I built with a cathode biased, plate follower buffer, very similar to the KM54c but using a different tube. The mics use the M94 capsule, which I’m told is very similar to KK54. They are my “poor man’s KM54’s”. But they are a bit bright. So I started studying the KM54 circuit and figured out I don’t have the feedback. I’ve added the feedback to a mic as a test, as drawn on the KM54 schematic and it works, maybe too well. Weird thing is I used a 30K trimpot for R2, but it doesn’t seem to have any effect I can measure. What controls the gain of the feedback, R1or R2?

The other difference with my mics is the backplate is charged instead of the diaphragm like KM54. But this shouldn’t make any difference, right?

Thanks for any guidance you can provide.

 

[abbey road d enfer]

Specifically what does R2 do? I understand “abgliechen” means adjustment, balancing, tuning. What exactly is being tuned? Freq response? Gain?

HF response, essentially. The circuit provides limited NFB, almost nothing below 1kHz, and  typically about 1-4 dB at 10kHz.

  If I understand this feedback, it should reduce more HF than low. 

That is correct.

a Neumann old timer posted a fascinating Neumann document that shows what the freq response of the buffer should be. It says  -1db overall, and -3db at 16K. 

That is consistent with the schemo.

  it doesn’t seem to have any effect I can measure.

The trap is that the Calibration input should be grounded. This is normally done in the PSU (via a 200r resistor IIRC) .

What controls the gain of the feedback, R1or R2?

Both, but the two critical components are C4 and R2.

The other difference with my mics is the backplate is charged instead of the diaphragm like KM54. But this shouldn’t make any difference, right?

It shouldn't.

 

[patosandia]

sorry this is slightly offtopic, but i tried the same on a royer mod and didnt worked.. what am i missing here?

 

[abbey road d enfer]

sorry this is slightly offtopic, but i tried the same on a royer mod and didnt worked..

Because the 0.1uF cap presents a too low impedance to the NFB network. You would need to replace C1 with a series RC circuit  10nF with 20kohms.

BTW 470pF seems too high.

 

[AusTex64]

Abbey, thanks very much for your reply. You wrote:

<The trap is that the Calibration input should be grounded. This is normally done in the PSU (via a 200r resistor IIRC) .>

I grounded the calibration input (C4) without the 200ohm resistor, figured that out after I posted. The HF is now quite audible, almost too much. So how do I dial it back? Add the 200 ohm resistor between the calibration input and ground, and increase the value of R2?

Thanks again for the tip on the resistor in the power supply. No info on that anywhere.

 

[abbey road d enfer]

Abbey, thanks very much for your reply. You wrote:

<The trap is that the Calibration input should be grounded. This is normally done in the PSU (via a 200r resistor IIRC) .>

I grounded the calibration input (C4) without the 200ohm resistor, figured that out after I posted. The HF is now quite audible, almost too much. So how do I dial it back? Add the 200 ohm resistor between the calibration input and ground, and increase the value of R2?

The actual value of 200r is not critical (in fact, I'm nor sure it was actually 200r, it's a vague reminiscence of my days at Barclay Studios in the 70's). Grounding direct is actually correct regarding EQ (but not if you wanted to re-enact this calibration facillity).
The academic procedure would be to actually tune the R1/C4/R2 network (ideally with a facility to measure the frequency response), but yes, you may start with decreasing R2, then you may change C4 in order to shift the turnover frequency.

the tip on the resistor in the power supply. No info on that anywhere.

Yes, it looks like no info is available on the net.

EDIT: however, looking at the schemo of NN48a, which was apparently a universal PSU capable of powering all sorts of Neumann mics (it has two Tuchel connectors, one small (KM style) and one large (U67 style), the calibration input was shorted to screen (!) in the PSU. Screen is usually tied to 0V, but it can also be disconnected!

 

[AusTex64]

Abbey, thanks for the info and schematic.

Here is an interesting comment I received from Uwe Sattler, formerly of Neumann:

<the NKMa and other compatible power supplies for the KM5x series microphones pin 6 (calibration input) of the microphone connector is tied to pin 3 (ground), rather than through a terminating resistor. It probably does not make any practical difference, as a 200Ω or even a 600Ω terminating resistor value is negligible compared to the high input impedance at that point of the amplifier. None of the standard power supplies for these microphones features a calibration input.>

Perhaps the 200 ohm resistor is used as a load during calibration?

You wrote: <The academic procedure would be to actually tune the R1/C4/R2 network (ideally with a facility to measure the frequency response), but yes, you may start with decreasing R2, then you may change C4 in order to shift the turnover frequency.>

If I wanted to decrease the amount of feedback/HF attenuation, I would INCREASE the value of R1 and R2, correct? Or do I misunderstand? BTW, I went with 150pf for C4 instead of 160pf, could not find 160pf very easily.

If I interpreted the Neumann document correctly,  the way to tune this circuit is to use a 45pf cap in place of the capsule, inject signal into C4 and measure the output loaded at 1000 ohms. I have set up this way, using my QA401 and could see the HF drop. But varying the 30K trimpot I used in place of R2 didn't seem to do anything. I am going to try a listening test next. Make adjustments and see if I can hear a change.

 

[abbey road d enfer]

< None of the standard power supplies for these microphones features a calibration input.>

If you look at the schemos of the M48, 49 &50, the 200r res is built in the mic; that's probably why I had this vague reminiscence of a resistor somewhere, used only for calibration. U67 also has a built-in res, but 68r instead of 200.

Perhaps the 200 ohm resistor is used as a load during calibration?

  It is not really there to load the generator used during calibration (although it does); it's there to ensure continuity to ground with minor effects on the frequency response. 200r is negligible compared to the 15-25k of R2.

If I wanted to decrease the amount of feedback/HF attenuation, I would INCREASE the value of R1 and R2, correct? Or do I misunderstand?

You would need to increase R1 or decrease R2; decreasing C4 also decreases NFB. It's a voltage divider where the series leg is R1+C4 and the shunt leg is R2+C3

BTW, I went with 150pf for C4 instead of 160pf, could not find 160pf very easily.

I very much doubt you could hear the difference between 150 and 160pF (less than 0.5dB).

If I interpreted the Neumann document correctly,  the way to tune this circuit is to use a 45pf cap in place of the capsule, inject signal into C4 and measure the output loaded at 1000 ohms. I have set up this way, using my QA401 and could see the HF drop. But varying the 30K trimpot I used in place of R2 didn't seem to do anything.

That is weird; are you positive you have grounded C3 during the procedure?
[/quote]

 

[AusTex64]

<That is weird; are you positive you have grounded C3 during the procedure? >

Thank you sir, may I have another?

No, I did not. I lifted the leg to ground on C4 and injected signal there. Will insert a 200 ohm resistor between C4 and ground, and inject signal at the junction of C4 and the 200 ohm resistor. Correct procedure?

 

[abbey road d enfer]

<That is weird; are you positive you have grounded C3 during the procedure? >

Thank you sir, may I have another?

No, I did not. I lifted the leg to ground on C4 and injected signal there.

How did you inject signal there? What generator?

  Will insert a 200 ohm resistor between C4 and ground, and inject signal at the junction of C4 and the 200 ohm resistor. Correct procedure?

Yes, that's what you should do. 200r is not critical, just needs to be << 15kohms.

 

[AusTex64]

<How did you inject signal there? What generator?>

The generator on the QA401.

Abbey, thank you very much for your insights. You have provided the missing links necessary to hopefully make this work in my mics. Will post back how it turns out. I need to get over my pouting about the Astros losing last night to the Dodgers....

 

[AusTex64]

Reporting back. Added 200 ohm resistor between C4 and ground. Discovered the 30K trimpot I’m using for R2 was not wired in correctly. Doh! Now we’re cooking.

Question - what do you think about leaving that trimpot there vs. replacing it with a fixed resistor? Oliver Archut used a 5K trimpot on the CS mics to dial in bias voltage exactly, but that’s not in the audio path like this. Your thoughts?

Very interesting experiment. I have two of these mics here, one without FB and one with. The FB really levels out the freq response of the buffer, as expected. The FB mic is now  more mid forward, and the one without FB sounds “scooped” or “hyped” with more bass and more zing on the hi end. The 8K boost on the M94 really comes through on the mic without FB. The mic with FB sounds more even, but not as “open” as the mic without FB. Perhaps lost a bit of headroom with the FB?

Gonna record a nice Collings acoustic next and see how they both sound.

 

[abbey road d enfer]

Question - what do you think about leaving that trimpot there vs. replacing it with a fixed resistor? Oliver Archut used a 5K trimpot on the CS mics to dial in bias voltage exactly, but that’s not in the audio path like this. Your thoughts?

It's not a matter of sonic quality, it's one of reliability; trimmers are more prone to mechanical dysfunction than fixed resistors. Unless you intend to make variable EQ a feature, or intend to swap capsules often, I would install a fixed resistor.

[/quote] Very interesting experiment. I have two of these mics here, one without FB and one with. The FB really levels out the freq response of the buffer, as expected. The FB mic is now  more mid forward, and the one without FB sounds “scooped” or “hyped” with more bass and more zing on the hi end. The 8K boost on the M94 really comes through on the mic without FB. The mic with FB sounds more even, but not as “open” as the mic without FB. Perhaps lost a bit of headroom with the FB? [/quote] certainly not headroom. NFB always help with headroom. I think it's a matter of finding an in-between setting, perhaps using a 50k trimmer. Seems like a good argument for a switchable EQ (or keeping the trimmer)...