Comparison of JFETs for mic applications

[JohnRoberts]

If you are still working in the Audio field, they are not. Trust me on that.

I'm not, trust me on that.

That's terrible Latin. If you MUST be trite, please use:

NOLI PATI A SCELESTIS OPPRIMI

Thor

Quid agis hodie?

My apologies for dog latin, but I studied it 60 years ago....

JR

 

[thor.zmt]

Quid agis hodie?

litore habitantis
potator cervisie
persequens puellarum
excogitatoris singulari soni apparatu

My latin lessons are also long time ago.

Thor

 

[k brown]

k brown, you are right that the IBM/Lenovo ThinkPads are by far the most reliable. That's what I've been using so far but the last one died about a year ago. The T61 is the most powerful that had XP drivers. The later Lenovo TP series are not as reliable.

Which model ThinkPads have you got? If you can spare one from the early 90s, dis beach bum would be eternally grateful. Do any have XP?

Not sure how I can re-compense you unless you are making a Tetrahedral Ambisonic mike. If so, I can do a full alignment and calibration that you can compare with da 'general purpose' programmes .. or perhaps even Thors SupaDupa one

Something a mystery here; I had three, and the only one I can find now is an A31. It won't boot up because the mem batt is dead; I used to be able to manually put in the date and it would boot, but now I cen't get it do that.

The other two I can't find anywhere; my wife may have taken them to electronics recycling without my knowledge.

 

[k brown]

No, the J-Fet Library posted at the LT-Spice group of course. The one you need to sign up to access.



In the 2sk660 pdf.

View attachment 143448
As you can see, once we have > 2V across the FET, we are in saturation.



But we are not operation at 1V, are we?

View attachment 143449

I ran BAV99 in the sim, the results match @MicUlli 's test results, good enough for me.



Can do. If you insist. Not that it will change things.



There is nothing here not reasonably well understood for at least 5 decades. I'm still puzzled why we even have a debate.



Other than the Brownian motion noise of the capsule, which I will tackle once my office is set back up, I have confidence in my sims. The kind of confidence that comes from having simmed many dozens of circuits, taken them onto a PCB for mass production and having tested the result with AP2.

So thank you for concern, but it's misplaced.

Meanwhile, please post the circuits you want me to use for the "SimpleP48" and the "SimpleP48RCA" clfor a build in diode J-FET with Id(ss) = 370uA @ 10V.

I was digging around at the io group but gave up. Very poor SNR.

Thor

I purposely retain a certain amount of noise in my gear because recorded music is only emotionally moving because of Brownian Motion.

 

[Khron]

I purposely retain a certain amount of noise in my gear because recorded music is only emotionally moving because of Brownian Motion.

This one / this kind, right?

 

[micolas]

This one / this kind, right?

 

[JohnRoberts]

litore habitantis
potator cervisie
persequens puellarum
excogitatoris singulari soni apparatu

My latin lessons are also long time ago.

Thor

but still teaching vocabulary...

I only recognize about half those words. None of them are JFET.

JR

 

[thor.zmt]

I only recognize about half those words. None of them are JFET.

You asked what I am doing...

Thor

 

[ricardo]

No, the J-Fet Library posted at the LT-Spice group of course. The one you need to sign up to access.

In the 2sk660 pdf.

View attachment 143448

Thanks Thor. I've already got this. But this isn't ... build in diode J-FET with Id(ss) = 370uA @ 10V.

Please post your model with Idss = 370uA @ 10V

I'll have a look at the model in the LTspice JFET library

---

It display can be customized a lot, but it displays either input or output noise density in V|/Hz .. across an arbitrary bandwidth.

Can it plot this against log frequency?

V/rt(Hz) is OK if we have log frequency scale like moamps B&K curve (which is Constant Relative Bandwidth .. but we'll pass on that if it's too difficult for TINA )

Or are you saying we have to do V/rt(Hz) in a bunch of separate 'bandwidths' and plot them ourselves???

After all, the noise advantage is my 'only' claim for both Zephyr's circuit and SimpleP48. I've already conceded da THD battle to Thor's SupaDupa circuit though I'd like another go with SimpleP48RCA .. which is in SimpleP48.pdf as one of the two recommended versions.

 

[ricardo]

Something a mystery here; I had three, and the only one I can find now is an A31. It won't boot up because the mem batt is dead; I used to be able to manually put in the date and it would boot, but now I cen't get it do that.

The other two I can't find anywhere; my wife may have taken them to electronics recycling without my knowledge.

Thanks for taking the trouble to have a look k brown

 

[thor.zmt]

I purposely retain a certain amount of noise in my gear because recorded music is only emotionally moving because of Brownian Motion.

This an interesting topic.

Easy experiment. Take a super clean acoustic recording, with very little noise.

Now add pink noise, violet noise and white noise at -60dB based on signal envelope, to make separate "noisy" tracks, meaning the noise will always be 60dB below the signal peak level and in principle mased by the signal.

Do a blind listening comparison.

Not ABX, but preference based.

And try to find descriptive terminology for each file without knowing the spectrum of the added noise.

Similar fun can be had with HD, compare a classic even order distortion dominant "picket fence" at -60dB added to an (overly?) "clean" track, to a distortion spectrum at -60dB without the even order harmonics (odd order only).

These levels are not that far from what many classic microphones produce.

Thor

 

[JohnRoberts]

You asked what I am doing...

Thor

The words I did recognize appear on brand for you... I do not need to feed this veer.

JR

 

[JohnRoberts]

This an interesting topic.

Easy experiment. Take a super clean acoustic recording, with very little noise.

Now add pink noise, violet noise and white noise at -60dB based on signal envelope, to make separate "noisy" tracks, meaning the noise will always be 60dB below the signal peak level and in principle mased by the signal.

Do a blind listening comparison.

Not ABX, but preference based.

And try to find descriptive terminology for each file without knowing the spectrum of the added noise.

Similar fun can be had with HD, compare a classic even order distortion dominant "picket fence" at -60dB added to an (overly?) "clean" track, to a distortion spectrum at -60dB without the even order harmonics (odd order only).

These levels are not that far from what many classic microphones produce.

Thor

I vaguely recall some old psychoacoustic research that suggested a high noise floor can create the impression of extended HF response .

Human audition is not exactly linear.

JR

 

[rogs]

This one / this kind, right?

Trouble is, once you've seen this you cant 'unsee' it!... I shall now have this picture in my head every time I hear the word 'Brownian' -- probably forever!

 

[micolas]

I vaguely recall some old psychoacoustic research that suggested a high noise floor can create the impression of extended HF response .

Human audition is not exactly linear.

JR

Some intentionally introduce noise into the music or mix, for artistic/psychoacoustic or technical purposes. When mastering, for example, we apply dithering (we add intentional noise to a signal in order to reduce quantization distortion when reducing the bit depth of a file, when bouncing from 24 or 32-bit down to 16-bits)

 

[JohnRoberts]

Some intentionally introduce noise into the music or mix, for artistic/psychoacoustic or technical purposes. When mastering, for example, we apply dithering (we add intentional noise to a signal in order to reduce quantization distortion when reducing the bit depth of a file, when bouncing from 24 or 32-bit down to 16-bits)

I do not know or understand the phenomenon but this predates digital dithering.

FWIW HF bias in magnetic tape recording assists low signal level resolution a similar result to dithering. But neither is related to perceived HF response .

JR

 

[thor.zmt]

Thanks Thor. I've already got this. But this isn't ... build in diode J-FET with Id(ss) = 370uA @ 10V.

Correct, the datasheet shows the curves for 250uA Idss @ 10V. They obviously shift with Idss which is a bit variable, especially in older J-Fets.

The difference is ~ 50%.

My computer and office/lab is in boxes, until I unpack and set up, no model.

Can it plot this against log frequency?

It can plot against any axis format you elect to use. The stuff you dis.iss as "unnecessary bells and whistles".

After all, the noise advantage is my 'only' claim for both Zephyr's circuit and SimpleP48.

Can you leave Zeph out until.he starts using nbers that actually make sense and are within reasonable variations of reality?

I have zero confidence in what is in this doc.

I've already conceded da THD battle to Thor's SupaDupa circuit

IT IS NOT A SUPA DUPA CIRCUIT.

For Fox sake dial it down.

though I'd like another go with SimpleP48RCA .. which is in SimpleP48.pdf as one of the two recommended versions.

I repeat, as there is no way to simply locate circuits at the micbuilders group and I lack the time to wade through all of that.

PLEASE post the circuits you want me to simulate as "SimpleP48" and "SimpleP48RCA", with all values according to the capsule and J-Fet as mentinoned.

So far you have been endlessly prevaricating and sandbagging and moving goalposts, after first a lot of big gob trash talking.

Otherwise I will have to use my own interpretation, which you may complain is not what you wanted.

So, to summarise, we agree that in all metrics, except electronic noise "SimpleP48" is not competitive with a very basic Schoeps derived design (I do not claim it as independent design of mine).

If the electronic noise difference is material in the context of a real microphone outside of Microsofts anechoic chamber is actually the real point of debate.

I already conceded that the theoretical electronic noise limit of the Schoeps derived circuit will be as much as 10dB greater, simply due to resistor noise. The tradeoff is lower HD (dramatically so) and the ability to handle high SPL's, which is what I emphasised in my design.

So if you want me to tell you that your circuit has lower noise (how much in a real mic remains debatable), I accept that.

The other problems of your circuit however make it exceedingly hard to recommend as a high quality recording microphone or for general universal use.

If you want to use a modern electronic (not transformer) input mic pre (build into a USB soundcard) and you are not interested in low distortion at typical recording levels for music, but only in low noise, for very low sound levels.

You win. You have a great simple microphone for that.

My only problem (and that of 90% plus of recording microphone users) I have no application for such a microphone. Zero. Zip. Zilch. Nada. It would make a so ic mess out of anything I might record, probably going full overload.

I never even considered this use case for my design, it is typically within 12" of a vocalists mouth, of an acoustic guitar, of a 12" X 4 Marshall Cabinet, an Ampeg SVT 10" X 8 Cabinet, within a few feet of a Grand Piano etc.

So, you win the trophy for a low noise microphone circuit design with 2 parts plus FET, that has extremely limited practical use to most people and is utterly useless to me.

That way we can save a lot of time.

And once you actually put all the relevant health warnings into your document, I have no point of criticism whatsoever.

Otherwise peeps will think they have an amazing guru designed Recording Mic and be disappointed.

Thor

 

[ccaudle]

But we are not operation at 1V, are we?

With a 2SK660 type device, is Vgs always close to 0? I'm trying to understand how much adjustment of bias point is available by modifying either the source resistor, or current feedback into the source. I am having trouble getting my head wrapped around how that works with a non-linear device connected between the source and gate. Do you really just analyze it like a very high impedance resistance, so you always have ~0V Vgs?
If that is the case how do you reconcile the two points that the source/drain current should be close to constant because Vgs is constant, but since you have a source follower the current should be determined by the voltage across the source resistor? Just like any other bootstrapped source follower, and treat the diode like a multi G resistor?

 

[thor.zmt]

With a 2SK660 type device, is Vgs always close to 0?

Precisely. It operates at 0V Vgs.

The PN junction of the J-Fet acts as the forward diode (conducts significant current with ~ +0.5V Vgs) and the added NP (viewed from gate) junction acts as reverse diode (conducts significant current with ~ -0.5V Vgs).

The Diode Junctions as "many gigaohm resistors" malarkey only really works at Vgs ~ 0V.

I'm trying to understand how much adjustment of bias point is available by modifying either the source resistor, or current feedback into the source.

Non, whatsoever.

Imagine for a moment a J-Fet with a 1GOhm resistor connected between gate & source internally.

I am having trouble getting my head wrapped around how that works with a non-linear device connected between the source and gate.

It only works if Vgs is essentially at 0V.

Do you really just analyze it like a very high impedance resistance, so you always have ~0V Vgs?

If you have significant Vgs from the signal, then you have a significant problem.

You will have to keep Vgs essentially near constant by appropriate circuit design.

If that is the case how do you reconcile the two points that the source/drain current should be close to constant because Vgs is constant, but since you have a source follower the current should be determined by the voltage across the source resistor?

You have (in "my" circuit) 5V across the source resistor, 10V across the J-Fet + Cascode and 5V across the drain resistor.

Thus the gate sits at +5V in this case.

Just like any other bootstrapped source follower, and treat the diode like a multi G resistor?

Yup.

Thor

 

[k brown]

I've always been curious about this method of biasing the 2SK660; it's from the Audio Technica AT8533 power module; so different from the way FETs are usually biased in microphones.

The only other thing I've seen like it is the PM for the old Shure SM90/91 boundary mics (which use a different FET).

In casual listening tests this simple circuit always wins out over other circuits designed for mics using the K660.