U87 Ai circuit: how low/high can you go? (de-emphasis capacitors)

[Khron]

... But now we (also) know the circuit gives you ~17dB of attenuation between the signal source, and wherever your measurement point is.

Later edit: If you want to see the actual (full) output, you'll want to look at the difference between the two ends of the secondary transformer. As it stands, that secondary winding is pretty much floating, so that apparent 17dB of attenuation is... unrealistic, i'm sure you'll agree. And you'll probably also want to feed the 48v through a pair of 6.8k resistors, just as your preamp would. As it is, those 2.2k resistors do nothing except present a load to the transformer secondary.

 

[OneRoomStudio]

... But now we (also) know the circuit gives you ~17dB of attenuation between the signal source, and wherever your measurement point is.

Later edit: If you want to see the actual (full) output, you'll want to look at the difference between the two ends of the secondary transformer. As it stands, that secondary winding is pretty much floating, so that apparent 17dB of attenuation is... unrealistic, i'm sure you'll agree. And you'll probably also want to feed the 48v through a pair of 6.8k resistors, just as your preamp would. As it is, those 2.2k resistors do nothing except present a load to the transformer secondary.

None of those things will have an impact of the frequency response, which is what I'm simulating. Decibels are a relative measurement anyway, not absolute, and will depend entirely on the arbitrary level of the signal recorded.

 

[OneRoomStudio]

One more addition that might be helpful for folks.

 

[micolas]

One more addition that might be helpful for folks.

View attachment 124570
View attachment 124571

How the noise of the microphone varies according to the degree of de-emphasis, respectively with the increase of the capacitor value
C6 - treble head?
Are there measurements?

 

[micolas]

@OneRoomStudios
In the mid/high frequency range with C6=0pF, does U87 behave similarly to km84 in terms of circuit linearity in frequency and noise level?

 

[OneRoomStudio]

@OneRoomStudios
In the mid/high frequency range with C6=0pF, does U87 behave similarly to km84 in terms of circuit linearity in frequency and noise level?

Noise measurements would have to be made in-circuit, and these are all simulations, so I have no idea on that front.

If c6 were removed (not shorted), the top-end would be fairly linear and limited only by the capsule and possibly the transformer.

 

[micolas]

Noise measurements would have to be made in-circuit, and these are all simulations, so I have no idea on that front.

If c6 were removed (not shorted), the top-end would be fairly linear and limited only by the capsule and possibly the transformer.

I see
Maybe someone did some measurements though. I would be curious about the use of capsules that do not require de-emphasis in the U87 circuit. Experiment
(C6=0pF means removed
C6=0 ohm means shorted)
Indeed transformers (7:1 and 10:1) can be very different in terms of all parameters.
Thanks!

 

[RuudNL]

an increase to 100uF will get you back about ¼ dB at 40Hz.

Today I tested the influence of increasing the source bypass capacitor from 22µF to 122µF.
(Just adding a 100µF in parallel...)
The effect was in the order of 0.1 dB @ 40 Hz.

 

[micolas]

Today I tested the influence of increasing the source bypass capacitor from 22µF to 122µF.
(Just adding a 100µF in parallel...)
The effect was in the order of 0.1 dB @ 40 Hz.

Neumann used 20uF in the vintage versions

 

[RuudNL]

It seems 22µF (20µF) is good enough...

 

[rock soderstrom]

Today I tested the influence of increasing the source bypass capacitor from 22µF to 122µF.
(Just adding a 100µF in parallel...)
The effect was in the order of 0.1 dB @ 40 Hz.

Clearly audible difference!

 

[micolas]

Today I tested the influence of increasing the source bypass capacitor from 22µF to 122µF.
(Just adding a 100µF in parallel...)
The effect was in the order of 0.1 dB @ 40 Hz.

I also did a voice test. About 2 years ago. I did not record, I did not measure.
I changed the 22uF capacitor with a 10uF one.
Something was missing from my voice. It seemed a little different to me.
I went back to 22uF and recognized my voice again.
Maybe I made a hi-pass somewhere in the 'very low-mid' area. I felt a difference anyway.
If I heard a difference of about 0.0** dB, I'm a pro!
I think that the frequency variation depending on the capacitor is not linear, the capacitor also influences the bias of jFEt and the gain of the circuit.

 

[OneRoomStudio]

Today I tested the influence of increasing the source bypass capacitor from 22µF to 122µF.
(Just adding a 100µF in parallel...)
The effect was in the order of 0.1 dB @ 40 Hz.

Yes, the simulation shows a difference of 0.15 dB (I exaggerated slightly to say a quarter of a dB). Either way, not a big difference in that direction. It can make a difference in terms of bass-cut though, if you go much below 20uF. The stock circuit (c5=33n, c6=220p, c8=22u) has 40Hz about 2.6dB down relative to 1kHz. If you decrease c8 to 2.2uF though, 40Hz would be 3.75dB down relative to 1kHz (so you'd lose more than 1dB in the low end).

To be clear though, now that I have fixed my simulation (with help from @Khron), I agree that adjusting c5 is a more effective way to sculpt the low end.

Lesson learned. Never doubt @RuudNL

 

[stoa]

Thank you guys for the very helpful insights!