Comparison of JFETs for mic applications

[jp8]

I did not see any indication that the Toshiba devices include an on-die low leakage diode for biasing like 2sk660, 2sk596, etc.

I think I interpreted your comment about lack of suppliers a bit too broadly, because in your post you were indeed very specific about JFETs with on-die diode. And you're right, the 879 and 880 don't have that, indeed. It looks like the 2SK880 is just the same chip as the 2SK209 btw, only in a smaller SC-70 package. But ayway, why frantically search for JFETs with on-die diodes? I don't see a point in that, unless you want to build your own electret mic capsule.

If your circuit does call for a JFET with Gate diode you can just take a regular JFET and add an external diode, as demonstrated by e.g. @MicUlli. So I think my proposal to check out the 879 and 880 is still relevant in the context of this thread where we want to investigate suitability of different JFETs in various types of impedance converter topologies. OK, the 880 can be skipped as the similar 209 was already investigated by @Voyager10 , but I'd still be interested to see how the 879 would perform.

 

[Voyager10]

@jp8 - there is data for the 2SK880 alongside the '209 (quick links to relevant tables: post #33, post #38, post #49) and they are essentially identical, certainly given the expected natural variation between devices, and the measurement accuracy (0.1dB is only 1% variation, I'll be extremely pleased if the repeatability of the measurements gets anywhere near this).

I don't have 'charge amp' measurements for any surface-mount devices yet (the layout of the ORS87 board doesn't match the arrangement of my SOT23 adapters, so it'll need a new test fixture building), but it looks like that will be of interest...

 

[Voyager10]

Found it:
https://www.edaboard.com/attachments/3-15-design-ideas-pdf.132665/

Thor

Table 2 also mentions the 2SK2219 as a possible alternative to the '596 and '660. Sadly doesn't appear to be any easier to find.

Also, the output noise range (3.6-4.6uV, -109 to -107 dBV, A weighted) is a good 10dB higher than I've been measuring. If it's not the FET, is there something about the rest of the circuit?

 

[thor.zmt]

Also, the output noise range (3.6-4.6uV, -109 to -107 dBV, A weighted) is a good 10dB higher than I've been measuring. If it's not the FET, is there something about the rest of the circuit?

The text states noise was measured with 10pF Cx, that is, we expect noise around 6.8 times higher than with 68pF.

Thor

 

[thor.zmt]

I was aware of this circuit when it was first published.

I referenced this article STRICTLY because it includes actual noise Data on the kind of J-Fets I have been pontificating about.

I neither recommend, nor disrecommend this circuit, but would state that I would be unlikely to use it.

OTOH, it does bootstrap Cgd, but not Cgs.

Well, it's a follower, so Cgs is bootstrapped by the follower action and Q2.

While the actual emitter follower Q3 bootstrap's Cgd. It in effect maintains 2 x Vbe across the J-Fet. I feel a BSS84 MOSFET would be a better choice as follower, but that's just me.

Balancing the currents is relevant for transformer based Mic inputs. So it's not a "non-issue".

IMO it's good food for thought, but it has not been developped fully.

I suspect this was left to those "skilled in the art".

And it certainly does not lend itself to capacitive NFB (short for "Charge Amp with EXTREMELY LOW loop gain").

It is a follower! Of course there is no gain.

Why do you bring "charge amp" into this, totally out of context?

Thor

 

[Mr. Moscode]

In Spice simulation software, units are unnecessary. A resistor marked 100 is a 100 ohms resistor, a capacitor marked 1n is a 1nF capacitor, and so on, thus a capacitor marked 1f (note the lower-case f) is a 1 femtofarad capacitor.
Though it has (almost) no relevance to the circuit's operation, it's useful because it allows copying the circuit without needing to delete an element that is not actually in the circuit.
The very same reason for 1u resistors or zero-volt voltage sources that one may see in some schematics.
Schematics in simulation are noticeably different than standard schematics; there are generally no connectors, no switches, but there may be additional elements destined to provide measurement points or offering the capability of making them inactive.

Thanks. I did not know that as I've never used one. My spice simulator needs a soldering iron. The one in my kitchen uses oregano, pepper and salt.

 

[Paul W]

First, Brownian motion of the air molecules impacting on the diaphragm. This is the ultimate limit. And the larger the diaphragm, the more noise, this is white noise. SDC's are at advantage here.

SDCs are at a disadvantage relative to LDCs, right? Doubling the area of the diaphragm doubles the signal but less-than-doubles the noise, because the noise in one half of the diaphragm is uncorrelated with the noise in the other half, and partly cancels. The signal gets 6 dB louder but the noise only gets 3 dB louder, so you get a net 3 dB improvement in SNR.

Quadrupling the area, like when going from 12 mm SDC to 24 mm LDC diaphragms, gives you 6 dB improvement in SNR.

 

[ccaudle]

why frantically search for JFETs with on-die diodes?

Maybe not frantically, but I think it will be interesting to compare the performance since those devices should presumably have a diode which is optimized for low leakage, and tend to be designed (or selected) for relatively low IDSS, which may be useful if trying to minimize current consumption.
But point taken, it isn't that much more trouble to use two packages, just find a JFET and add a diode.

 

[ricardo]

Found it:
https://www.edaboard.com/attachments/3-15-design-ideas-pdf.132665/

I do this juju on a circa 1980 measurement mike we supplied to Klark Teknik. As I'm only a pre10 guru, my version only has 2 active devices but an EVIL transformer.

Details are in Calrec.doc in the same MicBuilders directory as Zephyr.doc

Well, it's a follower, so Cgs is bootstrapped by the follower action and Q2.
....
It is a follower! Of course there is no gain.

Why do you bring "charge amp" into this, totally out of context?

At the end of SimpleP48.pdf, I expound on Great Guru Baxandall's explanation of the 'equivalence' of source & drain followers and Charge Amps.

If your circuit does call for a JFET with Gate diode you can just take a regular JFET and add an external diode, as demonstrated by e.g. @MicUlli.

SimpleP48.pdf also has the criteria for this diode.

MicUlli, I vaguely remember you doing this but can't find the thread after looking at about 10 pages of your posts. Could you or jp8 link to it please.

---

I haven't posted links in case k brown is offended. To anyone similarly offended, please contact me and I'll return your monies.

 

[ccaudle]

MicUlli, I vaguely remember you doing this but can't find the thread

Is this the thread you were thinking of?
https://groupdiy.com/threads/replacing-gohm-resistors-with-diodes-the-theory-behind-it.83141/

 

[thor.zmt]

SDCs are at a disadvantage relative to LDCs, right?

Correct. Brainfade on my part. Must be getting old or something.

Doubling the area of the diaphragm doubles the signal but less-than-doubles the noise, because the noise in one half of the diaphgragm is uncorrelated with the noise in the other half, and partly cancels. The signal gets 6 dB louder but the noise only gets 3 dB louder, so you get a net 3 dB improvement in SNR.

Quadrupling the area, like when going from 12 mm SDC to 24 mm LDC diaphragms, gives you 6 dB improvement in SNR.

Absolutely correct.

And "how much noise"?

NOTE, originally the numbers were off due to a unit conversion error!

Measurements on a circular electret microphone with a radius of 1.8 cm and a sensitivity of 5 mV/µbar yield Brownian‐motion noise voltages of about 9 nV|/Hz in the frequency range between 1 and 5 kHz.

So an LDC capsule (36mm Diaphragm) with -46dB -26dB sensitivity had around 1.3uV noise from Brownian motion, or around -72dB -92dB.

Common LDC Capsules are around 20mV/Pa so noise will be also lower by a factor 2.5, for around 0.52uV noise or 3.6nV|/Hz. So any J-Fet with < 3.6nV|/Hz at ~ 5kHz and up will degrade the flat band noise by less than 3dB, above 5kHz.

Trivial to scale from here for capsule size and sensitivity to get a first approximation of Brownian Motion noise.

Capsule capacitance and Resistor noise give us the LF noise function. A low leakage diode for biasing leads to less noise than a 1GOhm resistor and is likely easier to purchase for a DIY Enthusiast.

If our capsule above is 65pF and our effective resistance 500M (1G Bias and 1G after a coupling capacitor like commonly seen) will produce 410uV noise, but with a lowpass of ~ 5Hz (65pF x 500MOhm = 32500uS = 4.89Hz).

So 3uV|/Hz without lowpass becomes 300nV|/Hz at 50Hz, 30nV|/Hz at 500Hz and 3nV|/Hz at 5kHz. So we intersect the flat line Brownian motion noise at around 2
5kHz. We usually call noise with this behavior "1/LF" noise.

So our circuit noise is 3.6nV|/Hz for the flat (Brownian motion) part of the noise spectrum rising from ~ 5kHz to 300nV|/Hz at 50Hz and 530nV|/Hz at 20Hz.

That is for capsule, 1G bias resistor, 1nF coupling capacitor and 1G to ground, for a 20mV/PA or -34dB sensitivity capsule

With 3.6nV|/Hz in the flat band above 5kHz I must say I have few concerns that any sensible J-FET noise whatsoever will be of consequence for the complete microphone.

Note, all noise figures unweighted. For A weighting overlay this curve (which shows why A-weighted noise in microphones so low despite there being a ton of it):



I guess we can kill Brownian noise by using our microphones in a Vacuum and the resistor noise by doing it at very low temperatures close to absolute zero... 555

Thor

PS, this is not a high precision analysis of microphone noise, more a quick first order approximation that allows us to get a handle on the general magnitude of noise, not an 0.001dB accurate prediction.

 

[abbey road d enfer]

Well, it's a follower, so Cgs is bootstrapped by the follower action and Q2.

Agreed. I did look the wrong way, how the output is taken from the drain.

Balancing the currents is relevant for transformer based Mic inputs. So it's not a "non-issue".

This is in many cases done by using equal value resistors for extracting power.

Why do you bring "charge amp" into this, totally out of context?

I didn't. MicUlli did in post #139.

 

[thor.zmt]

This is in many cases done by using equal value resistors for extracting power.

This causes extra losses. And precludes DC coupling, mind you, 99.9% of all transformerless inputs have coupling cap's anyway.

I didn't. MicUlli did in post #139.

Ahhh, well, I posted this article strictly to have actual measurements of noise of the J-Fet's that were not tested.

In an "ideal" charge Amp (Amplifier Gain -oo) any parasitic capacitance's at the amplifier input are cancelled. But I said that before.

Thor

 

[thor.zmt]

If our capsule above is 65pF and our effective resistance 500M (1G Bias and 1G after a coupling capacitor like commonly seen) will produce 410uV noise, but with a lowpass of ~ 5Hz (65pF x 500MOhm = 32500uS = 4.89Hz).

As @MicUlli measured how diodes instead of 1GOhm perform, let's have this here:

From: https://groupdiy.com/threads/replac...odes-the-theory-behind-it.83141/#post-1076024



It shows very graphically (pun intended) why the 1G resistors are actually pretty stoopeed, unless you like to have heavily pink noise dither in your recording, of course...

Thor

 

[abbey road d enfer]

Measurements on a circular electret microphone with a radius of 1.8 cm and a sensitivity of 5 mV/µbar yield Brownian‐motion noise voltages of about 9 nV|/Hz in the frequency range between 1 and 5 kHz.

Assuming noise-pressure density is more or less linear, that would result in about 1.2uV across the audio band, or 72dB below 94dBspl, or 22dB equivalent noise.
This cannot be chalked entirely to Brownian movement. Several SDC's have better noise performance. Microsoft have measured the sound pressure in their "most silent room in the world". The results were below -20 dBspl. Considering the most part of it is thermal agitation (Brownian noise), I tend to think that it's smaller than that.
Something doesn't add up. Maybe it's me...

 

[thor.zmt]

Assuming noise-pressure density is more or less linear, that would result in about 1.2uV across the audio band, or 72dB below 94dBspl, or 22dB equivalent noise.

I stand to be corrected.

Possible options, electret are more noisy.

Unit conversion error. All my noise values are 20dB too high. 1uBar = 0.1PA, so 5mV/uBar = 50mV/PA.

But the article I cribbed this from looked at how well Electrets self noise maches theory and found a good match.

February 01 1974 - Noise Due to Brownian Motion of the Diaphragm, of Electret Microphones

This cannot be chalked entirely to Brownian movement. Several SDC's have better noise performance. Microsoft have measured the equivalent sound pressure of thermal agitation at room temperature and results were below -20 dBspl.

Do you have a reference I can peruse?

Something doesn't add up. Maybe it's me...

Or me.

Thor

 

[abbey road d enfer]

https://edition.cnn.com/style/article/anechoic-chamber-worlds-quietest-room/index.html

 

[thor.zmt]

https://edition.cnn.com/style/article/anechoic-chamber-worlds-quietest-room/index.html

The final result? “The noise level measured inside is -20.3dBA. This means that the ambient noise in the chamber is 20.3dB below the threshold of human hearing.”

Alas, no indication HOW this was measured. Like Which microphone.

Also, this cannot include Brownian motion impact on microphone diaphragms.

Still, interesting place.



Thor

 

[abbey road d enfer]

Also, this cannot include Brownian motion impact on microphone diaphragms.

My understanding is that the minimum noise in a completely isolated enclosure is due to air molecules movements (a.k.a. thermal agitation or Brownian noise).
The type and size of microphone does not change the noise sound pressure.

What's the use of this grandiloquent cover, when the original is so beautiful?

 

[thor.zmt]

My understanding is that the minimum noise in a completely isolated enclosure is due to air molecules movements (a.k.a. thermal agitation or Brownian noise).

It DOESN'T SAY THAT. You are assuming something.

Brownian noise was covered in my reference from 1974 and was suggested as -100dB/PA (now I got my unit conversion fixed) in a physical sense (-5dB re 0.0002ubar).

This is much higher than -114dB/PA.

The type and size of microphone does not change the noise sound pressure.

Yes, but if a microphone has (say) 6dB(A) self noise, like the Microtech Gefell M 940, how can we possibly measure noise levels 26dB lower?

What's the use of this grandiloquent cover, when the original is so beautiful?

The original is very hippy, the cover is by far more modern, more angry and edgy. I like that.

Too many seasons of quiet rage
Too many young people just wasted away
Too many futures hanging in the balance
So much owing, nothing left to pay
A lonely flag flutters in a breeze
For the hardened hearts
who still want to believe

Am I the witness or am I the crime?
A victim of history or
just a sign of the times?
Across my heart questions
and shadows still fly

In the dead of the night
I know where the answer lies
Woman be my country,
until my country can be mine
I have no flag, I sing no anthem,
I no longer carry an Armalite

Thor