Microphone "emphasis circuit"

[pasarski]

Once again out of my depth on the subject, but here we go! I built a simple JFET microphone that uses a K47 capsule and U47 -style body. I made it a "vibe mic". Single fet, a transformer and no feedback. I think it sounds really nice, on male vocals at least. And biased for symmetrical clipping it has enough headroom for vocal use. THD is quite high - unacceptable some could say - but that's what I aimed for.

I compared it to an other K47 mic (a tube mic and my main vocal mic) which sounded quite similar, but had little less mud and little more treble. So for the heck of it I tried if I can match my mic to sound more like it. I messed around in LTspice and came out with this little feedback loop (R2, R3, C3). It's kind of like the U87 de-emphasis but does the opposite thing so I guess it's an emphasis circuit then



Fr after adding the filter:



For my surprise it works and the mic really sounds closer to the reference mic now. But there's one down side (you guessed it) - noise. The mic got quite a lot noisier after the mod. It was really quiet before. I didn't try to measure the noise at any point, but it was the quietest fet mic I've built this far. Now the mic has something like 7db more noise. I was kind of expecting the noise to increase, but didn't expect that big of a difference.

Being out of depth on the subject and the "design process" being "messing around with a simulator" I'm not quite sure how to reduce the noise without messing up the fr? I guess the circuit cancels out the advantage gained with the 5gig resistor(s)? I also guess the capsule is part of the filter?

If someone has an idea how to make it quieter it'd be fab! If not, there's a noisy microphone emphasis feedback circuit for your pleasure, be my guest

 

[RuudNL]

I suppose R3 (33M) is responsible for most of the noise...
(Didn't you also miss a lot of low-end after the modification?)
You start with a 5 G.ohm gate resistor and then place 33.1 M.ohm in parallel...

 

[pasarski]

I suppose R3 (33M) is responsible for most of the noise...
(Didn't you also miss a lot of low-end after the modification?)
You start with a 5 G.ohm gate resistor and then place 33.1 M.ohm in parallel...

Yes I lost a little something I don't miss for the vocal use.

And yes and of course you're right, the 33 meg is in parallel with the 5 gig and of course it cancels out the noise suppressing effect of the 5 gig resistor (which is quite huge to my ear btw)

But even without aknowledgin that fact (which is obvious now that you mentioned it) I tried the circuit with larger values for the 33 meg and couldn't get anywhere close to that frequency response and attributed that to the capsule being part of the filter, which I'm not sure it's true.

 

[RuudNL]

The Neumann design superimposes the de-emphasis feedback on the backplate of the capsule, not directly to the gate of the FET.
In this case you can keep the feedback resistor values a lot lower and keep the advantage of a high value gate resistor.

 

[pasarski]

Of course! Thank you so much, that's exactly the kind of answer I was looking for. Now back to messing around with the simulator...

 

[oscarhuang]

I am very interested in the k47 frequency increase circuit, but I have never found a suitable circuit. I look forward to your success and analysis.

 

[pasarski]

Okay, arrived at a fairly similar fr after some fiddling around. Feedback is now superimposed (love this word) to the backplate. I had to add a coupling cap C5. I also changed the source bypass cap for lower value to match the low end. It's really fascinating how everything affects everything in circuits like these. It has a little bit more "sub" now also. I'm going to test it tomorrow in rl.

 

[oscarhuang]

Inspired, I made a circuit for k47 based on the m49 circuit. Among them, c3 and R7 form a low-pass filter. Negative feedback is given to the capsule. According to my meager understanding of the microphone circuit, the K47 high frequency will be improved. I didn't do circuit simulation, but I made a k47 microphone, which did increase the high frequency.

 

[RuudNL]

C3 = 10 µF ???

 

[Cqwet Dbdfte]

Is that C10 75pF the capsule's capacitance?

 

[oscarhuang]

C3 = 10 µF ???

Yes. I used a 10uf capacitor for C3. This is the frequency response measured with REW. More like a slow low cut from 2000HZ. I will test Mr. Pasarski's drawings.

 

[pasarski]

Okay, arrived at a fairly similar fr after some fiddling around. Feedback is now superimposed (love this word) to the backplate. I had to add a coupling cap C5. I also changed the source bypass cap for lower value to match the low end. It's really fascinating how everything affects everything in circuits like these. It has a little bit more "sub" now also. I'm going to test it tomorrow in rl.

I modified the circuit, and this one works too, sounding similar to the original. However, I can't hear a significant difference in noise level, and I can't wrap my head around why. I'll leave it as is for now. It's not particularly noisy—definitely usable for many things—but it was really quiet before.

The reference mic I used is a tube mic with a self-noise specification of 12.4 dB(A). Before adding the emphasis circuit, mine was noticeably quieter—perhaps around 6 dB, judging by ear. After the modification, it was as noisy, or slightly noisier than the reference mic, maybe about 1 dB. I also compared it to my old U8i, whose noise level was about the same, and the sensitivity seems about the same as well.

(I use a very unscientific method for estimating noise: I record my singing—and silence—match the singing levels, then listen to the noise, adjusting the fader until the noise levels sound the same.)

My conclusion: if I were making this mic for my own use, I wouldn't add the emphasis circuit at all. I'd prefer to EQ it afterward and enjoy the silence. However, this one is for a friend who loved my main mic when we worked in the same studio, so I wanted to "match" the frequency response. It seems I also nearly matched the noise level too

If you want to try this, choose which ever one you like more and experiment with the values. If you add a similar feedback path to another inherently noisier circuit, it probably won’t make much of a difference since another noise source will likely dominate.

 

[oscarhuang]

I also modified my circuit diagram. I made and tested it according to Mr. Pasarski's circuit diagram. The purple curve is the frequency response curve of the microphone without any auxiliary circuit. Blue is Mr. Pasarski's frequency response curve. Yellow and green are the curves after I modified the circuit. You can see that changing R7 or C4 can change the curve of the microphone. I ended up choosing 1000pf and 3.3M. If you want to try a different eq, try changing the parameters of these two originals.

 

[pasarski]

I also modified my circuit diagram. I made and tested it according to Mr. Pasarski's circuit diagram. The purple curve is the frequency response curve of the microphone without any auxiliary circuit. Blue is Mr. Pasarski's frequency response curve. Yellow and green are the curves after I modified the circuit. You can see that changing R7 or C4 can change the curve of the microphone. I ended up choosing 1000pf and 3.3M. If you want to try a different eq, try changing the parameters of these two originals.View attachment 136268View attachment 136270

Seems more like a low cut I'm not able to understand and explain my circuit properly, but I can at least guess

Without the smaller resistor to ground (R3 in my schemo) it's a low pass filter rather than band pass. Low passed signal is used for the negative feedback, thus reducing low frequencies. That of course makes the high frequencies relatively louder.

In my circuit (as I understand it) a band passed signal is used for the negative feedback and thus the mid frequencies are attenuated creating a smiley curve. The component values have to be selected very carefuly so that the band that is passed is in the right spot - somewhere in the middle of the audible range. If the the center frequency is too low it behaves like low cut, if too high like a high cut. In my circuit I wanted the center frequency to be in the 300-400 hz range to attenuate the mud but keep the lows intact (human lows, not whale lows).

Also, there are more components involved in the frequency shaping than the three I added. At least the drain resistor (plate resistor in your case) and the capacitance of the capsule affect the curve.

You can add a smaller resistor to ground like I'm my circuit, but guessing the right values using a real mic would be super time consuming. Maybe you can simulate your circuit and play with the values in the sim?

 

[oscarhuang]

Seems more like a low cut I'm not able to understand and explain my circuit properly, but I can at least guess

Without the smaller resistor to ground (R3 in my schemo) it's a low pass filter rather than band pass. Low passed signal is used for the negative feedback, thus reducing low frequencies. That of course makes the high frequencies relatively louder.

In my circuit (as I understand it) a band passed signal is used for the negative feedback and thus the mid frequencies are attenuated creating a smiley curve. The component values have to be selected very carefuly so that the band that is passed is in the right spot - somewhere in the middle of the audible range. If the the center frequency is too low it behaves like low cut, if too high like a high cut. In my circuit I wanted the center frequency to be in the 300-400 hz range to attenuate the mud but keep the lows intact (human lows, not whale lows).

Also, there are more components involved in the frequency shaping than the three I added. At least the drain resistor (plate resistor in your case) and the capacitance of the capsule affect the curve.

You can add a smaller resistor to ground like I'm my circuit, but guessing the right values using a real mic would be super time consuming. Maybe you can simulate your circuit and play with the values in the sim?

You're absolutely right, currently my circuit is just a low cut circuit. rather than a trap circuit. More experiments and designs are needed to improve it.We need a band-pass filter circuit to complete it.

 

[oscarhuang]

Seems more like a low cut I'm not able to understand and explain my circuit properly, but I can at least guess

Without the smaller resistor to ground (R3 in my schemo) it's a low pass filter rather than band pass. Low passed signal is used for the negative feedback, thus reducing low frequencies. That of course makes the high frequencies relatively louder.

In my circuit (as I understand it) a band passed signal is used for the negative feedback and thus the mid frequencies are attenuated creating a smiley curve. The component values have to be selected very carefuly so that the band that is passed is in the right spot - somewhere in the middle of the audible range. If the the center frequency is too low it behaves like low cut, if too high like a high cut. In my circuit I wanted the center frequency to be in the 300-400 hz range to attenuate the mud but keep the lows intact (human lows, not whale lows).

Also, there are more components involved in the frequency shaping than the three I added. At least the drain resistor (plate resistor in your case) and the capacitance of the capsule affect the curve.

You can add a smaller resistor to ground like I'm my circuit, but guessing the right values using a real mic would be super time consuming. Maybe you can simulate your circuit and play with the values in the sim?

I just changed the parameters of the circuit. You can see that it is basically a high frequency boost circuit.

 

[kingkorg]

I just changed the parameters of the circuit. You can see that it is basically a high frequency boost circuit.View attachment 136284

Where's all that noise and ripple coming from in the graphs?

 

[RuudNL]

Yes. I used a 10uf capacitor for C3. This is the frequency response measured with REW.

Honestly I don't see how this could work. 1 M.ohm and 10 µF give a corner frequency of 0.015 Hz...

 

[oscarhuang]

Where's all that noise and ripple coming from in the graphs?

I use interface and REW as measurement tools. The headphone output port of the interface is used as the input signal for microphone measurement. There is a squeaking noise every time I connect it.

 

[oscarhuang]

Honestly I don't see how this could work. 1 M.ohm and 10 µF give a corner frequency of 0.015 Hz...

Maybe my circuit is wrong, but it works but not obvious. The high-frequency boost of this latest circuit is very obvious.