U67 de-emphasis network

[OPR]

I did the original test with pink noise and a drum beat. Makes sense what you wrote
Also this is with u87. The NFB circuit is way different.

You're up a bit late! 3:30 am in Norway..

Now when you say the audio Nulls do you mean that the residual gives you absolute silence?
nothing at all?

Can you post the wavs?

What distance from the speaker are you working at?
I have a bunch of 87's here and can repeat your test with the 67 also.

 

[kingkorg]

You're up a bit late! 3:30 am in Norway..

Now when you say the audio Nulls do you mean that the residual gives you absolute silence?
nothing at all?

Can you post the wavs?

What distance from the speaker are you working at?
I have a bunch of 87's here and can repeat your test with the 67 also.

Not absolute silence, it was close to the noise floor. I'll see if i can dig up the files. It's at 40cm, coaxial speaker.

 

[ccaudle]

devising a test

A test which would not require an acoustic test environment would be injecting a frequency sweep signal into the feedback path, and measure the differential voltage across the capsule. I think everyone would agree that any effect on capsule behavior would have to be proportional to the voltage applied to the capsule.

 

[GeorgeToledo]

Under what mechanism would varying backplate voltage not affect movement of a diaphragm?

 

[ccaudle]

Under what mechanism would varying backplate voltage not affect movement of a diaphragm?

If the entire capsule "floated" on that backplate voltage, and the diaphragm voltage increased or decreased directly with the backplate voltage then there would be no affect on the movement of the diaphragm, since it is the difference between the backplate voltage and diaphragm voltage which determines the attractive force.
From a quick look at the circuit it appears that the driving impedance of the feedback network and the impedance of the capsule at 100Hz are both much lower than the input impedance of the buffer stage, so I don't see why there should be much voltage across the capsule due to the feedback.

 

[OPR]

If the entire capsule "floated" on that backplate voltage, and the diaphragm voltage increased or decreased directly with the backplate voltage then there would be no affect on the movement of the diaphragm, since it is the difference between the backplate voltage and diaphragm voltage which determines the attractive force.
From a quick look at the circuit it appears that the driving impedance of the feedback network and the impedance of the capsule at 100Hz are both much lower than the input impedance of the buffer stage, so I don't see why there should be much voltage across the capsule due to the feedback.

I'm not totally sure I fully understand what you are saying. Are you saying that the feedback is 100% canceling with the input voltage and therefor no movement of the diaphragm caused by the NFB occurs? or am I missing something.

If that's the case? Because of the sloping nature of the LF curve caused by the NFB would it also be true that 100% cancellation would not be occurring across all the frequency of range in question? , And there for some diaphragm movement may occur? not to mention the sonic imperfections, nonlinearities, and other artifacts cause by the tube it's self as well as the NFB filters. It's certaily wouldn't be a 1:1 digital perfect copy being fed back.

 

[OneRoomStudio]

My reasoning (which I'm not sure is right) goes like this:

For signals in the middle range, between low hundreds and low thousands of Hz, an SM-57 puts out a much stronger signal than a similar-diameter condenser. At least in that middle band, then, the condenser appears to be a very inefficient transducer.

The condenser may have a pretty flat frequency response down to 25 Hz or further, but unless it has a huge bass bump, it can't be a dramatically more efficient transducer in the bass than it is in the midrange. So presumably, it's comparably inefficient in the bass to how inefficient it is in the midrange.

I like OPR's redirection to design a test, but I wanted to address this. I mentioned 25kHz, not 25Hz.

Capsules are NOT linear. Put aside the dynamic comparison for a second. We all know K67/K87 capsules have a significant rise in treble frequencies (hence the need for the "deemphasis" circuit in the first place). That "rise" is relative to mid and low frequencies. That means that the capsule is more "efficient" at higher frequencies than it is at lower frequencies - it takes less acoustic energy at high frequencies than it does at low frequencies to product the same electrical voltage.

When you are feeding signal back into a capsule that has a non-linear response, that's different than feeding signal back through a non-microphonic capacitor, which is linear (other than the RC filtration the capacitance itself introduces).

The difference between a true capacitor and and non-linear capsule is what we're discussing.

 

[soliloqueen]

I like OPR's redirection to design a test, but I wanted to address this. I mentioned 25kHz, not 25Hz.

Capsules are NOT linear. Put aside the dynamic comparison for a second. We all know K67/K87 capsules have a significant rise in treble frequencies (hence the need for the "deemphasis" circuit in the first place). That "rise" is relative to mid and low frequencies. That means that the capsule is more "efficient" at higher frequencies than it is at lower frequencies - it takes less acoustic energy at high frequencies than it does at low frequencies to product the same electrical voltage.

When you are feeding signal back into a capsule that has a non-linear response, that's different than feeding signal back through a non-microphonic capacitor, which is linear (other than the RC filtration the capacitance itself introduces).

The difference between a true capacitor and and non-linear capsule is what we're discussing.

the k87 is inefficient at 25khz tho because of the band limiting caused by the center termination i think

 

[OneRoomStudio]

the k87 is inefficient at 25khz tho because of the band limiting caused by the center termination i think

Fair. My original post about 25kHz was generically about condensers vs dynamic mics, not K87s specifically.

My actual point was the non-linearity of condenser capsules.

 

[ccaudle]

K67/K87 capsules have a significant rise in treble frequencies

That is not a non-linearity in the mathematical sense, it is just a non-flat frequency response because it is a resonant system. There are non-linearities which are noticeable at really high amplitude, but at reasonable amplitudes you are not going to get a big shift in generated distortion products from putting feedback around a capsule like you will with a transistor or tube.

 

[ccaudle]

Are you saying that the feedback is 100% canceling with the input voltage

What are you referring to as "the input voltage?"

and therefor no movement of the diaphragm caused by the NFB occurs?

Maybe you are overthinking it. Break it down into two questions:

Would you agree that for there to be movement of the diaphragm caused by a signal (or modification of the movement the diaphragm is already being subjected to by sound waves) that there must be signal voltage across the capsule, i.e measured between the backplate and diaphragm(s)?

Assuming so, then calculate or measure how much feedback signal voltage there actually is across the capsule. The circuit is the driving impedance of the feedback network, the impedance of the capsule (which is essentially just the capacitive reactance at whatever frequency you want to check), and the 400M resistor at the grid of the tube.

 

[Paul W]

I like OPR's redirection to design a test, but I wanted to address this. I mentioned 25kHz, not 25Hz.

Capsules are NOT linear.

There are very different senses of "linear" that are relevant to this discussion.

It sounds like what you're saying is that capsules' frequency response is not flat, which is a very different thing from whether it's "linear" in the sense of a "linear, time-invariant" (LTI) system, which has a lot to do with what kinds of distortion you should expect or be surprised by.

I agree that condenser capsules' frequency response is not flat. The argument I was making doesn't depend on their being absolutely flat, just that they don't have a big bass boost.

And as I understand it, K67-type capsules generally do have pretty flat frequency response across the midrange and into the bass, with a rolloff below that. They are not more efficient in the bass, or not much more, than they are in the midrange where they're typically pretty flat.

I misunderstood what you were saying about 25KHz sensitivity, but for the argument I'm making, what's going on above a few KHz is irrelevant, and what's happening at much lower frequencies like 25 Hz (not KHz) is relevant to the high-pass filtering were' talking about

The argument is just that K67 type capsule is a far less efficient transducer than an SM57 capsule across the midrange, therefore not a very efficient transducer there, and it's not a lot more efficient in the bass, so it's not an efficient transducer in the bass either.

---

Beyond that, my impression is that a good condenser capsule is generally pretty close to a linear, time-invariant system, within broad bounds, when it's not near its noise floor or very close to its maximum SPL.

One consequence of that is if you play two different signals A & B through the capsule at different times, and add the outputs later, the result is much the same as if you add the two input signals together, and play the combined A + B signal through the capsule once, and capture the output from that combined input signal.

That doesn't depend on the frequency response being flat, but it does depend on the frequency response being well-defined and consistent, for example boosting A and B by 6 dB around 10KHz separately, or boosting the combined A + B signal in the same place by the same amount.

Of course that will not be true for signals approaching the maximum SPL. The capsule might not distort significantly on A or B, if they are below that threshold, but may distort on the combined signal A + B, if adding them together puts the volume of the signal over the threshold of distortion.

 

[Matt Nolan]

to me this suggests more simply that it makes sure that excessive diaphragm excursions can't reach the preamplifier, which is true.

but couldn't that be much more simply implemented with lower value capsule biasing resistors?

Arguing against my hypothesis above: There could be reasons not to do that such as noise floor considerations, or wanting decent LF pick-up but still squashing sub frequencies. Or it could be that having the feedback circuit to linearise other things and areas of the spectrum meant that the framework was already there, so why not also add a little to it for LF control.

Practical tests with acoustic and injected inputs, real capsules and dummy capacitors still seem like a good idea to me.

 

[Paul W]

Why are (or aren't) anechoic measurements needed for making accurate CK12 and M7 capsules?

What are the limitations of doing the usual REW-type sweeps in a regular room? (And maybe half-space measurements on the ground in a parking lot for the bass?)

What does putting the mic / capsule in a tube do?

 

[kingkorg]

What are the limitations of doing the usual REW-type sweeps in a regular room? (And maybe half-space measurements on the ground in a parking lot for the bass?)

No limitations whatsoever!!!


CK12 has to be measured constantly throughout the manufacturing process in order to check all of the boxes, that's why it's so difficult to make. That is also why @OPR said the absolute capacitance measurements don't matter that much. He obviously knows what he's doing.


M7 is dependent on gluing the diaphragm, and you can't be sure the two sides are identical unless you measure them. The two sides have to be identical in order to properly work in all the patterns.

 

[Paul W]

The tube thing sounds interesting.

When thinking about how to measure the efficiency of the capsule as a speaker, I'd thought maybe you should put the capsule at the end of a tube of the same diameter, to direct the sound to a measurement mic without it radiating outward and falling off with the square of the distance. As long as you're only measuring frequencies whose wavelengths are long compared to the diameter of the tube, I'd think that'd be a fine waveguide. (Not that I really understand such things.)

But then I thought that it might have weird implications for mechanical impedance---maybe it'd be a more efficient transducer that way, and not representative of what actually happens in free air.

But for quality control measurements, maybe that doesn't really matter, and you wouldn't really even need to know how to translate between free air and in-the-tube measurements analytically... once you have a reference set of capsules that measure differently, you can put them in the tube and see how they measure there, and do the translation empirically... this is what a good capsule measures like in the tube, and that is what a bad capsule measures like in the tube, and just interpolate between a few known reference points.

So I'd guess once you've scoped out the range of behaviors in free air in a controlled environment, you could just use the tube for most QC measurements, and put the whole apparatus in a medium-sized box with a couple of layers of heavy walls and some acoustic insulation, rather than needing a separate measurement room.

For measuring the capsule as a microphone capsule rather than a speaker, of course you'd need a driver at the other end of the tube, rather than a measurement mic, but the same principle should apply.

 

[ZHOUJUN]

Measuring microphone data is not as easy as measuring speakers. You must have a reference microphone and collect data at the same distance. If the measurement environment is inconsistent, the measurement results will often deviate greatly. The biggest problem is that you don’t know how the AKG and Neumann curves are measured. To get real data, you must have an ideal free-field environment and measurement equipment. The picture shows Neumann’s modern test room in Germany.