U67 de-emphasis network

[OPR]

The fed back signal is in the mV range and has to be proportional to the "input" signal capsule provides. Otherwise it would become dominant and the whole feedback thing would fall appart.

This actually supports what I'm saying. .e If the LF signal is attenuated by way of phase cancelling the positive signal with the negative. Then to bring LF plosive back to 0v it would need to be proportional to the input.

This is the reason i don't agree with OPR, because that level of feedback is miniscule in order to combat the attraction force of polarization voltage.

Well I don't agree with that statement either lol! It's not what I'm saying.

What I am saying NFB in the U67 are AC voltages(Audio)and are in opposition to the input voltage detected by the capsule from the real world. They function by phase cancellation at chosen frequencies.

All you have to understand is that if a microphone capsule is fed an AC signal it will move the diaphragm in the direction dictated by the signal. If a duplicate signal of equal voltage but of opposite phase is fed to the capsule then it will cease to move. much like a speaker when preforming a null test in a DAW. ie import a track,duplcate it to another track, set at equal volumes flip the Phase. viola! silence...

I don't think it's that controversial.

Gotta get back to work now! Great conversation.

 

[kingkorg]

Also another thing. Here's a THD measurement of an actual u87ai. Real one, from Neumann. The injected signal level is something you would expect in a real world vocal situation. THD at 1K is at 0.063%. But the transformer is crapping out despite the feedback, and at 25hz THD is 3.3%! My point is, at this level, at these THD percentages an EQ would do the same job. Miniscule reductions of THD are negligible. And the transformer you have in your mic will still saturate at those low frequencies, and give you ''warmth''. Depending on what transformer you use of course.


Here we have a graph where the signal gives 1% THD at 1K. We see higher THD drop at HF which is about 0.5%, follows the same trend at LF, but the transformer craps out again and gives 17% at 25hz. Plosives can easily give this SPL level.



And then the paradox! Here's the same measurement, but including 2nd harmonic (red). Even though THD is decreasing overall, especially at HF, 2nd harmonic is increasing at HF despite the feedback.



Sadly, i don't have a real u67 to measure. Measuring clones would be pointless as it's a complex system made out of specific selected components for a reason. Also worth to mention, u87 is still very different, and has a different feedback network. Mind you, this is under extreme SPL conditions. It's complicated...

 

[Murdock]

Under extreme conditions yes, not under normal operation. Feedback definitely DOES reduce THD at those frequencies, but we are arguing about if feedback level is enough to physically push the diaphragm away from the backplate, and i say absolutely not.

We are talking about miniscule differences that happen below 1% THD under normal operation, and TBH i don't care about those. Hence my opinion that using an EQ in post is giving the same results. But you have a good point that if we approach ''breakup'' region feedback will have larger impact. But how many people are using this type of mic in that manner? IDK...

Also philosophical approach... The paper came out in the period when least possible THD was desired. Today we use THD as a tool, for nostalgia... I don't think anyone today uses tube mics because they want less THD. IDK, it's complicated.

Soliloqueen brought up an interesting thing up in a private conversation we had. What happens when clipped LF signal at the plate gets fed back to the capsule when in-phase signal coming from the capsule isn't clipped? I have no idea.

I think I'm bit lost what the discussion is about... I thought you're saying, the feedback in U67 is pointless and EQ after could do the same just more simple.
Wasn't it you who measured output of a K67 type capsule. It can be really high, right?
Would be interesting to measure a real U67 or 1:1 clone and compare with let's say a U87 under high low frequency pressure levels.

I don't know what you mean when you talk about the tube "closing." I'm guessing you're talking about the tube amp saturating/clipping.

I would think this circuit would be limited in its ability to keep e.g. a plosive from saturating the circuit, because the LF has to get through the circuit forwards before it can be fed back phase-rotated to cancel itself out. Wouldn't the feedback be too late to keep the rising edge from clipping?

Edit: I posted this before seeing kingkorg's post above it. I don't know if we're talking about the same thing.

I just quoted the patent paper. In another paper they call it " overloads and the well known "blocking" effects" (caused by exaggerated excursions of the capsules diaphragm due to wind or mechanical shock interference).

Here is another description from a Gotham Paper and an excerpt where the feedback is explained.

 

[kingkorg]

I thought you're saying, the feedback in U67 is pointless and EQ after could do the same just more simple.

Exactly, under normal operating conditions.

On the other hand if we get into non-linear region, extreme conditions where capsule collapses, or close to it, i have no idea what will happen. But for sure, feedback wouldn't help against capsule collapse IMHO.

I wouldn't even use that mic in such extreme case.

If you are into that, i'd say you have to go with the original. 99.9% of replicas are not accurate enough to be considered as appropriate substitute with or without feedback. The mesh plays a huge role in plosive dispersion, capsule tension, exact transformer performance, tube gain, feedback accuracy.

Most replicas are just hodgepodge of parts assembled more or less accurately using nothing more than the schematic. Most people don't even check for FR of the feedback network. In this case i'd rather use a solidstate mic with an EQ and get closer to the desired result than what people nowadays consider to be a "clone".

 

[GeorgeToledo]

You can’t eq away THD.

 

[GeorgeToledo]

Yes that's correct the k67 was specked for 60v and compliments the 67 with it's tube circuit and negative feedback tuning. The fet circuit has less headroom and a more strident sound, relatively speaking, so the lower 48v polarization voltage added some extra headroom for the fet as well as a more laid back sound from the capsule, when compared to the later AI. Regarding capsules and polarization voltages, working back from the maximum voltage generally allows the diaphragm to be more compliant which sounds ,subjectively,more pleasing but at the expense of output. No such thing as a free lunch...

Another difference that I felt is worth circling back around to touch on, is to remember that there is parasitic capacitance from the rear capsule diaphragm with the U87 circuit but not U87A.

 

[kingkorg]

You can’t eq away THD.

You are either not reading what i'm posting, or you have strong convictions that are difficult to change even with posted measurements.

The improvement in THD achieved by using NFB is negligible. Even in extreme conditions it is just 0.5% (@20K) of improvement in u87. And even then THD at LF is dominated by THD caused by the transformer. Even at lower levels. Neumann chose to use NFB instead of simple filter is because why add THD (by using simple filter) when you can reduce THD by using NFB.

To anyone reading this, If you are a kind of person that hears 0,5% THD diference at 20Khz, while at the same time 2nd harmonic rises at those very frequencies, i apploud you. In that case you have to buy the real deal because there's no clone that will ever fulfill all the expectations. That is also if you record at above 130db which is what we are talking about here. And yes, you are probably a bat.

 

[OPR]

I just did a test with my 1:1 67 Clone feeding a test LF signal into the plate and using a second microphone I recorded the signal from the capsule.
So......drum roll Please!....................... It's definitely moving the diaphragm and therefore influencing it's behavior.

Once S2 is switched out( NFB disconnected) I get silence at the capsule.
That's the end of this particular story for me and proves that post EQ and this particular NFB trick are not the same.

 

[Paul W]

I just did a test with my 1:1 67 Clone feeding a test LF signal into the plate and using a second microphone I recorded the signal from the capsule.
So......drum roll Please!....................... It's definitely moving the diaphragm and therefore influencing it's behavior.

But how much? And how does the resulting output differ from just increasing the electrical cancellation?

I don't think any of us doubted that the capsule would act as a speaker when driven with an audio signal, and that there would be some additional cancellation effect by opposing the motion of the diaphragm.

That leaves open the question of how much the diaphragm's motion is actually inhibited by the feedback, vs. the straightforward electrical cancellation.

And what is the difference in waveform, vs. what you'd get if the capsule magically didn't act as a speaker at all, and you just used more electrical cancellation to get the desired high pass filtering effect?

I could imagine that the waveform is different partly because the capsule functioning as a speaker doesn't have flat frequency response in the bass, so low frequencies wouldn't be cancelled out as consistently by the motion-opposing speaker-like effect as they would be by the straightforward electrical effect.

 

[OPR]

But how much? And how does the resulting output differ from just increasing the electrical cancellation?

what do you mean by electrical cancellation?

That leaves open the question of how much the diaphragm's motion is actually inhibited by the feedback, vs. the straightforward electrical cancellation.

again I'm not understanding "straightforward electrical cancellation?

And what is the difference in waveform, vs. what you'd get if the capsule magically didn't act as a speaker at all, and you just used more electrical cancellation to get the desired high pass filtering effect?

I could imagine that the waveform is different partly because the capsule functioning as a speaker doesn't have flat frequency response in the bass, so low frequencies wouldn't be cancelled out as consistenly by the motion-opposing speaker-like effect as they would be by the straightforward electrical effect.

The purple line is with S2 open (no LF Negative Feeback
Brown is with NFB
You can see the effects well beyond the intended low frequencies.
If you've ever used a 67 or accurate clone you'd notice that they almost sound compressed compared to other mics of the era
47. 251 C12, even against a later U87 everything sounds very processed. This is likely an effect of the NFB limiting diaphragm
excursion while feeding back the LF.

Here a more expanded view.

 

[Paul W]

what do you mean by electrical cancellation?

I mean that even if the capsule didn't act like a speaker at all, feeding the 180-degree delayed low-frequency output back to the capsule connections would work just fine as a high-pass filter, because it's summing that signal with the capsule's output at the input of the amplifier.

As I understand it, that is the main mode of operation of the LF feedback---it's feeding the out-of-phase LF output back into the input of the amplifier, cancelling the (undelayed) LF part of the signal coming direct from the capsule.

For that aspect of operation, the capsule connections are just a summing bus for the in-phase capsule signal and the out-of-phase LF being fed back from the output.

If the capsule is a very inefficient speaker, the mechanical effect you've been talking about will be smaller... the feedback voltage may oppose the LF motion of the diaphragm, but not much. The capsule will vibrate almost as much as it would without the feedback, and the electrical summing effect will dominate.

If the capsule is an efficient speaker, the low-frequency vibrations of the diaphragm will be opposed more, and that may be a significant fraction of the high-pass filtering effect.

Even in that case, it's not clear what difference it makes in practice. If the effect on diaphragm motion is small, you can just use more feedback so that the electrical cancellation of LF at the input of the amplifier does almost the entire job. (Assuming that you're not right at the point of banging the diaphragm against the backplates.)

 

[OPR]

I mean that even if the capsule didn't act like a speaker at all, feeding the 180-degree delayed low-frequency output back to the capsule connections would work just fine as a high-pass filter, because it's summing that signal with the capsule's output at the input of the amplifier.

As I understand it, that is the main mode of operation of the LF feedback---it's feeding the out-of-phase LF output back into the input of the amplifier, cancelling the (undelayed) LF part of the signal coming direct from the capsule.

For that aspect of operation, the capsule connections are just a summing bus for the in-phase capsule and the out-of-phase LF being fed back from the output.

If the capsule is a very inefficient speaker, the mechanical effect you've been talking about will be smaller... the feedback voltage may oppose the LF motion of the diaphragm, but not much. The capsule will vibrate almost as much as it would without the feedback, and the electrical summing effect will dominate.

If the capsule is an efficient speaker, the low-frequency vibrations of the diaphragm will be opposed more, and that may be a significant fraction of the high-pass filtering effect.

Even in that case, it's not clear what difference it makes in practice. If the effect on diaphragm motion is small, you can just use more feedback so that the electrical cancellation of LF at the input of the amplifier does almost the entire job. (Assuming that you're not right at the point of banging the diaphragm against the backplates.)

I dont think you can have one without the other.
The point is a capsule is not just a capacitor, its a capsule therefore is designed to be sensitive to vibrations and AC signals.
It is what is.. The effect is unavoidable it's the nature of the best. If Neumann wanted no effect on the capsule it would have introduced the feedback at different stage of the of the circuit.

 

[Paul W]

I dont think you can have one without the other.

I don't either. I still think it's an important question how big the different effects are.

If Neumann wanted no effect on the capsule it would have introduced the feedback a different stage of the of the circuit.

I guess so.

It's still a good question just how much difference it actually makes to do it that way, and exactly what kind of difference you actually get.

I wouldn't assume it's something mic modelling can't approximate satisfactorily. I wouldn't even assume it's a difference worth modelling, beyond getting the overall frequency response right.

 

[OPR]

It's still a good question just how much difference it actually makes to do it that way, and exactly what kind of difference you actually get.

I wouldn't assume it's something mic modelling can't approximate satisfactorily. I wouldn't even assume it's a difference worth modelling, beyond getting the overall frequency response right.

On the contrary... we can easily copy a frequency response using software like Fabfilter pro q or whatever you prefer. it may look right on a FR graph (which is great for marketing B.S and makes the crowd say yay!) but the real difficulty and is in emulating the transient behavior ,resolution and pattern, which is unique to each transducer.

 

[Matt Nolan]

Chiming in here with some thoughts.

Surely a condenser mic diaphragm is as good a "speaker" as it is a microphone - not in terms of radiation efficiency and so on, but mV (or delta capacitance, but let's not get too hung up on that) per displacement and displacement per mV. They have to be almost exactly the same, no? Especially given the minuscule mass. That's just physics.

A condenser capsule is not perfect. The larger the displacement, the further away from linear response it will get. Though I admit, no clue as to how large this effect is. So I can see a good argument that transient behaviour will be affected by the feedback to the capsule like this. There are parallels with servo control going on here.

Thinking in terms of diaphragms hitting backplates is a red herring, I feel. Wrong scale.

Perhaps A/B tests can be done, looking at impulse responses using injected signals with a real capsule versus the least microphonic well-matched dummy capacitance in its place?

 

[ccaudle]

looking at impulse responses using injected signals

Do you have any reason to think a capsule is not minimum phase? If a capsule is minimum phase, then the transient response can be directly derived from the frequency response.