Comparison of JFETs for mic applications

[jp8]

Do you want to post the circuit?

Thor

Uhhmm, not yet. I understand that it is easier for you to think along in solutions if you have the schematic, but for now I will keep it to myself. As I said in an earlier post: maybe later. And if I do, I'll not post it in this thread.

Jan

 

[thor.zmt]

Uhhmm, not yet. I understand that it is easier for you to think along in solutions if you have the schematic, but for now I will keep it to myself.

I might have have had a BMFBFOTO*.

Normally one side of the capsule terminates into a low impedance. This lowpasses the noise of the bias element.

If the microphone is single diaphragm and backplate, there is no low impedance path to "ground" (or to anywhere). So no lowpass for noise.

A capsule with dual diaphragms and a separate backplate (3 wire) with a low impedance bias supply to the backplate and a diaphragm to each input should kill the noise.

Thor

* Big Mutha Funking Blinding Flash Of The Obvious

 

[jp8]

I might have have had a BMFBFOTO*.

Normally one side of the capsule terminates into a low impedance. This lowpasses the noise of the bias element.

If the microphone is single diaphragm and backplate, there is no low impedance path to "ground" (or to anywhere). So no lowpass for noise.

A capsule with dual diaphragms and a separate backplate (3 wire) with a low impedance bias supply to the backplate and a diaphragm to each input should kill the noise.

Thor

* Big Mutha Funking Blinding Flash Of The Obvious

It's an externally polarized SDC capsule with a "floating" membrane and backplate (Superlux S502 MK2). So it is not your typical SDC capsule with the membrane connected to the shell.

The capsule connector's "collar", which I believe connects to the membrane, is decoupled to ground through a 22nF C0G decoupling capacitor. This capacitor connects to the polarization voltage oscillator through a 10M resistor. On the output of the CMOS oscillator is another 22nF reservoir capacitor.

The capsule backplate connects to the Gate of the JFET and to a 2G resistor, which feeds the bias voltage to the Gate. So the 2G resistor noise has a "low" impedance path through the capsule capacitance and the 22nF capacitor to ground. Maybe I have my BMFBFOTO too now and should inspect the 22nF capacitor connected to the capsule.

Thanks Thor.

Jan

 

[thor.zmt]

The capsule backplate connects to the Gate of the JFET and to a 2G resistor, which feeds the bias voltage to the Gate.

Maybe I misunderstood, I was under the impression the whole circuit was balanced and the microphone connected between the halves, obviously not then.

Maybe I have my BMFBFOTO too now and should inspect the 22nF capacitor connected to the capsule.

Probably worth doing. But it seems I still fo.not understand your circuit correctly from your descriptions.

Thor

 

[jp8]

Maybe I misunderstood, I was under the impression the whole circuit was balanced and the microphone connected between the halves, obviously not then.



Probably worth doing. But it seems I still fo.not understand your circuit correctly from your descriptions.

Thor

Understand your confusion. One "leg" is the actual impedance buffer and sends audio to the preamp. The other "leg" copies most of the circuit of the first, but is only there for impedance balancing, balancing the phantom power supply load, and supplying the current to the CMOS oscillator. Those few SMT parts used in the 2nd "leg" are much smaller than a bunch of elcaps, allowing the circuit to fit into a very small SDC without having to couple the audio signal into the preamp through too small elcaps which would affect low-end response. The input of the 2nd "leg" is properly decoupled to ground to avoid feeding its noise into the preamp. Second BMFBFOTO: I should also check this decoupling capacitor.

 

[ricardo]

You were quite open to doing a noise frequency analysis in #264 As my claim(s) were very specific, I would have thought this was important .. if not vital.

I said I'd see what we can display

Thank you. I take this to mean you are happy to bring back the noise shootout with Noise Spectrum display

I'll tell you what I do instead. I'll use a 2SK660 as diode, as this will match closely what was diffused on die.

I'm happy with that.

I'll illustrate various circuits of interest using the same 2SK660 fet (or 2SK170 + 1Gohm where the original specifies such) and their relative performance for noise, frequency response and distortion

Please don't change the goalposts.

My original claim is that your #245 circuit will be >10dB noisier than Zephyr's Schoeps variant with his 34mm ISK CK12 capsule at 4kHz and higher. I also claim SimpleP48 is perhaps 1dB quiter than his circuit. You issued a challenge in the same post which I accepted.

As the common version of SimpleP48 has known problems at high spls, I suggested SimpleP48RCA (the other recommended version) as it has been tested for real life spl performance. You graciously agreed to add this to the shootout in #264.

Let's concentrate on these.

But if what you propose is in addition to our shootout, I can post Zephyr's circuit which has some noise advantage compared to the usual Schoeps circuit and is good at high spl too.

---

Can you tone down the trash talking please? I've not sh*t on your stuff. I made a comment about noise which is what we are testing.

We all KNOW, cos you have told us ad nauseum in 3 different languages, that you think all Chinese engineers, especially itinerant ones in Oz, are idiots and their stuff is not worth reading. I'm not sure your Supa powers allow you to claim sh*t is or isn't in there without bothering to read.

I have serious reservations about your 'noise analysis' in #249, #264 etc but there's no point in putting my $0.02 in until we have at least some sim. results.

We can discuss the reasons for the results, rationally, when we get them and attempt to relate them to 'real life'