JFET shootout

[k brown]

The only time mic self noise is an issue for me is nature recording in very quiet places. My trusty old Panasonics were always a problem. AT853 were almost useable. Best have been Primo EM23s and the newer Primo EM273. Even then the EM23s I use with K660s, and EM273 obvious contain a FET with diodes on the gate, and those are some of the quietest capsules out there.

But like you say, using the diode FETs are a different story with electrets.

 

[kingkorg]

O

The only time mic self noise is an issue for me is nature recording in very quiet places. My trusty old Panasonics were always a problem. AT853 were almost useable. Best have been Primo EM23s and the newer Primo EM273. Even then the EM23s I use with K660s, and EM273 obvious contain a FET with diodes on the gate, and those are some of the quietest capsules out there.

But like you say, using the diode FETs are a different story with electrets.

No matter how quiet those little capsules are, they are still going to be way noisier than something like NT1 capsule polarized to 100v.

The thing totally overlooked here, is how much capacitance swing a capsule can produce and how much voltage is baked into, or or applied to it. Not to bash op's process here, the thread is called fet shootout, but replacing capsule with a capacitor and measuring whole system noise makes zero sense. But yet again, the thread is called fet shootout, so in that sense we get fet noise figures.

Townsend Labs sphere is specified to 7db of self noise and uses tl070 opamp with 60v applied to the capsule. Raise the voltage and you get even less noise. WITH AN OPAMP!

The most hated answer. It's not the part, it's how it's used. IF4500 for example with it's 35pF of input capacitance!

 

[abbey road d enfer]

Suggest a capsule with voltage Vin and capacitance Cc and a JFET with input capacitance Cin. At the JFET gate you get Vgate = Vin*Cc/(Cc+Cin) because of trivial voltage divider rule. Now inject a noise current In at the JFET gate node. It will produce a noise voltage Vn = In/(jw(Cc+Cin)). Calculate
Vn/Vin = In/(jwCc). Cin is eliminated and therefore does not influence the signal to noise ratio..

This is true for the KTC noise voltage, but I think the FET's noise is constant, so reducing the effective signal decreases S/N.
When gain decreases, the noise of the rest of the circuit (including the mic preamp to which it is connected) becomes more audible.
Most readings of talented mic designers indicate they tend to favout minimum capacitive loading of the capsule, for this reason, and also for minimizing distortion. Either by choosing low capacitance FET's or bootstrapping/cascoding them.

NTE457 is highly claimed as best performer in the sensor industry. Used in several capacitive sensor devices, but not often seen in condenser mics..

I don't know which sensors you're referring to, but the ones that I know are usually of higher capacitance than the 15-100pF of a condenser mic. 2SK170 (input C=30pF, reverse C=6pF) is commonly used with piezo sensors, because it's very low noise. So indeed, comparing apples to oranges.

 

[abbey road d enfer]

replacing capsule with a capacitor and measuring whole system noise makes zero sense.

I think iot's a pretty harsh comment, in regard with the fact that the OP has conducted his tests with 3 capacitor values that are pertinent to small, medium and large capsules.

Townsend Labs sphere is specified to 7db of self noise and uses tl070 opamp with 60v applied to the capsule. Raise the voltage and you get even less noise. WITH AN OPAMP!

k brown mentioned "nature recording in very quiet places". You want to use a Townsend Sphere to record mosquito farts?

The most hated answer. It's not the part, it's how it's used.

Sure hated when it's proferred in such a condescending way...

 

[abbey road d enfer]

My point was that mic self noise is something that can get overly obsessed about - in practical use it is, in the vast majority of situations, inaudible in the final recording.

I tend to agree, although it can be debated. A moderately trained musician can easily distinguish the continuous white noise of electronics from the pink-ish noise of air-con circulation, and from the non-continuous noises of respiration and body moves. It may not be audible in the mix, particularly if spectral cleaning and heavy compression has been used, but the performing talent is quite sensitive to electronics noise.
Now, if you're recording wild animals, I don't think they care...

 

[kingkorg]

I think iot's a pretty harsh comment, in regard with the fact that the OP has conducted his tests with 3 capacitor values that are pertinent to small, medium and large capsules.

k brown mentioned "nature recording in very quiet places". You want to use a Townsend Sphere to record mosquito farts?

Sure hated when it's proferred in such a condescending way...

I don't really see how i was either being harsh, or condescending. But here i'll draw my non native English speaker card, and hope no one got offended.

1. Op did use 3 fixed values, but what's absent is how much a capsule swings in capacitance which will determine output of the capsule, and s/n ratio. Also how and how much the capsule is polarized. It would paint a greater picture of how really irrelevant fet noise performance would be in greater picture.

2. I provided example of TL Sphere as a mic that uses an opamp and has great noise performance. Curcuit like that can easily fit inside a sdc.

3. I gave an excellent example of a fet with high input capacitance which is used cleverly in one of the quietest circuits ever. No matter how many times i read the original sentence i can't see how it's condescending.

Thanks once again to the OP for a shootout of couple of fets under these specific testing conditions.

 

[abbey road d enfer]

I don't really see how i was either being harsh, or condescending.

"measuring whole system noise makes zero sense" is derogatory to the work the OP has conducted.
"It's not the part, it's how it's used" implies you think you know better, which may be true, but remember, some of the members here are not total idiots.

But here i'll draw my non native English speaker card

You're not alone. Actually, two mods out of three are non-native english speakers.

1. Op did use 3 fixed values, but what's absent is how much a capsule swings in capacitance which will determine output of the capsule, and s/n ratio. Also how and how much the capsule is polarized.

Actually, there is not so much variations within types of capsules. A typical LDC capsule's sensitivity is about 10mV/Pa +/-6dB and a SDC about 12dB less. Indeed there are exceptions, like B&K's 200V polarization and undamped diaphragms. So I still think that analyzing specifically electronics noise is a valid endeavour.

It would paint a greater picture of how really irrelevant fet noise performance would be in greater picture.

What is your estimate of the contribution of FET noise to the global noise in typical cases? Do you think it's negligible? If it was, veteran mic designers would not have wasted time on improving it.
Of course requirements are different between e.g. close-miced vocals and distant micing a flute in a church.
I would invite you to start a study where electronics noise is compared to capsule noise. That would produce objective assessment.

2. I provided example of TL Sphere as a mic that uses an opamp and has great noise performance. Curcuit like that can easily fit inside a sdc.

Are you sure the capsule is connected directly to a TL07x?
7dB(A) implies an electrical noise at the capsule level of about 0.5-1uV. A TL07x has 2.5uV noise.

3. I gave an excellent example of a fet with high input capacitance which is used cleverly in one of the quietest circuits ever

What circuit?

 

[rankot]

Did you try PF5102? I have them in my stash, but no conditions to do the tests like you did. Ciss max is 16pF, so I presume they usually have less.

 

[kingkorg]

"measuring whole system noise makes zero sense" is derogatory to the work the OP has conducted.
"It's not the part, it's how it's used" implies you think you know better, which may be true, but remember, some of the members here are not total idiots.

There's whole first part of sentence missing there. This is how i talk and think, i can see how you could interpret this as harsh, but it still makes no sense to me. You can not replace the capsule with a fixed value cap in a certain circuit and extrapolate how fet might behave in some other circuit with a real capsule.

You're not alone. Actually, two mods out of three are non-native english speakers.

Then i just guess you are not particularly fond of how i express myself, and this is fine.

Actually, there is not so much variations within types of capsules. A typical LDC capsule's sensitivity is about 10mV/Pa +/-6dB and a SDC about 12dB less. Indeed there are exceptions, like B&K's 200V polarization and undamped diaphragms. So I still think that analyzing specifically electronics noise is a valid endeavour.

Actually there is, and with all due respect what you stated here is wrong. The OP's original post shows relatively small differences in fet noise performance which can be easily compensated by choosing different capsule, or polarizing it using higher voltage. The whole point of my statement why you can't just replace a capsule with a capacitor. It is a valid endeavour, and i didn't dispute the value of performed test, i just point to overlooked stuff.

Are you sure the capsule is connected directly to a TL07x?
7dB(A) implies an electrical noise at the capsule level of about 0.5-1uV. A TL07x has 2.5uV noise.

Yes, and i could extrapolate here that you consider me an idiot, but i won't. There are many low noise opamp mics out there.

What circuit?

You answered it yourself, so i can't put it any better. Also excellent example of how input capacitance could be irrelevant, and how you use a fet is the key.

Looks like there's nothing magic here. The input FET is indeed a VLN type, with the associated high capacitance, but in an LDC it is not redhibitory. The FET input capacitance is bootstrapped by NFB, and the feedback capacitance is not a big issue since the voltage gain of the FET is quite low in this configuration, so I would say the combination of minimal attenuation and VLN FET explains almost everything.

At no point was i trying to bash @MicUlli thread. He is a valuable contributor here!

 

[kingkorg]

Did you try PF5102? I have them in my stash, but no conditions to do the tests like you did. Ciss max is 16pF, so I presume they usually have less.

This will depend on circuit you intend to use it in. Ciss could be something that matters, or not.

 

[rankot]

This will depend on circuit you intend to use it in. Ciss could be something that matters, or not.

We're talking about microphone circuits since the start of post.

 

[kingkorg]

We're talking about microphone circuits since the start of post.

Yes, i am aware of that. There are many different microphone circuits, and ways fets are used within them.

 

[abbey road d enfer]

There's whole first part of sentence missing there. This is how i talk and think, i can see how you could interpret this as harsh, but it still makes no sense to me.

OK, it's a matter of perception. After all if MicUlli does not feel offended, why should I bother? But saying that someone else's endeavour "makes zero-sense" is on the verge of discourtesy, which is against this group's rules.

You can not replace the capsule with a fixed value cap in a certain circuit and extrapolate how fet might behave in some other circuit with a real capsule.

Really? Many researches are conducted using the theorem of superposition.

Then i just guess you are not particularly fond of how i express myself, and this is fine.

I just mentioned that to affirm that we are familiar with language deviations and rather good at discriminating what comes from lack of mastery of english and what comes from bad manners.

Actually there is, and with all due respect what you stated here is wrong.

This is your comment in reply to mine, where I say "Actually, there is not so much variations within types of capsules." Do you actually negate it? Do you have examples of common LDC's with a sensitivity higher than 20mV/1Pa? I suppose you're familiar with the risks associated with increasing teh polariztion voltage of a capsule that's not designed for it.

The whole point of my statement why you can't just replace a capsule with a capacitor. It is a valid endeavour, and i didn't dispute the value of performed test, i just point to overlooked stuff.

So you say one thing and its contrary in the same sentence...

Yes, and i could extrapolate here that you consider me an idiot, but i won't.

That is a weird answer to my comment: "7dB(A) implies an electrical noise at the capsule level of about 0.5-1uV. A TL07x has 2.5uV noise."Do you have an objective answer to that?

There are many low noise opamp mics out there.

I don't doubt it, but a capsule directly into a TL07x, come on...

You answered it yourself, so i can't put it any better.

My question was "What circuit?" Where have I answered it? You must confuse me with someone else.

 

[kingkorg]

OK, it's a matter of perception. After all if MicUlli does not feel offended, why should I bother? But saying that someone else's endeavour "makes zero-sense" is on the verge of discourtesy, which is against this group's rules.

Really? Many researches are conducted using the theorem of superposition.

I just mentioned that to affirm that we are familiar with language deviations and rather good at discriminating what comes from lack of mastery of english and what comes from bad manners.

This is your comment in reply to mine, where I say "Actually, there is not so much variations within types of capsules." Do you actually negate it? Do you have examples of common LDC's with a sensitivity higher than 20mV/1Pa? I suppose you're familiar with the risks associated with increasing teh polariztion voltage of a capsule that's not designed for it.

So you say one thing and its contrary in the same sentence...

That is a weird answer to my comment: "7dB(A) implies an electrical noise at the capsule level of about 0.5-1uV. A TL07x has 2.5uV noise."Do you have an objective answer to that?

I don't doubt it, but a capsule directly into a TL07x, come on...

My question was "What circuit?" Where have I answered it? You must confuse me with someone else.

Abbey at this point i'm really considering someone might have hacked your account.

I'm not going to continue this discussion, as it seems to me you are negating anything i say even after quoting your own words, with a direct link to the quote and the original post.

I made a typo, it's tlc070, not tl070. The point remains the same. There is whole thread here devoted to that mic, with images!

There is also a thread with me experimenting with higher voltages, one where i burn the capsule, and several where i successfully polarize existing stock capsules to higher voltages. So yes there are many higher sensitivity capsules, one of which is found in Røde NT1.

I'll refrain myself from posting further in this thread, for anyone interested summary of my point:

Original post includes a valid test for fet noise under these particular conditions. However fet is rarely bottleneck in well executed mic circuit noise wise. Fet noise performance will depend on how it's used in a circuit.

While the original test is valid, the slightly, i believe not purposefully deceiving "What is the best JFET for diy mic builders?" is wrong, simply because the tested fets might not be the best solution in different circuits. Example of fet with high ciss is if4500, yet Røde NT1 is one of the quietest microphones ever made, and can be made even quieter by raising the capsule voltage to 100v, which i have done, reported, and documented here on this forum.

Peace out!

 

[Ricardus]

OK, it's a matter of perception. After all if MicUlli does not feel offended, why should I bother? But saying that someone else's endeavour "makes zero-sense" is on the verge of discourtesy, which is against this group's rules.

At the risk of trying to provide some context here, and as a native English speaker, Abbey is not wrong.

"Makes zero sense" is an absolute and makes it sound like you think you know more than anyone else on this topic. Saying something like "Makes no sense to me" would sound orders or magnitude less arrogant.

It's just a question of polite conversational English. I would only have worded it your original way if I was trying to be provocative.

 

[kingkorg]

At the risk of trying to provide some context here, and as a native English speaker, Abbey is not wrong.

"Makes zero sense" is an absolute and makes it sound like you think you know more than anyone else on this topic. Saying something like "Makes no sense to me" would sound orders or magnitude less arrogant.

It's just a question of polite conversational English. I would only have worded it your original way if I was trying to be provocative.

Thank you so much for the clarification, I sincerely wasn't aware of this. I definitely meant this made no sense to me. These are nuances I'm not able to pick up. I also hope the technical point I was trying to make got across.

 

[jarvis]

Hey,

I'm not sure that a discussion on whether a JFET's noise performance actually matters in the context of even higher capsule noise leads anywhere, there's certainly nothing substantial to gain from it.

Now there's been a discussion about different circuits calling for different types of JFETs but I've yet to see anyone but @ricardo elaborating on other possible criteria apart from the forward transconductance and the capacitance that are mentioned in the PDF in post #1. So I'm curios what the other parameters are that different circuits call for to obtain low noise.

The type of circuit also influences matters. The common Schoeps circuit needs a FET with high Idss for low noise and some FETs are inadequate.

My question is here: Do the aforementioned 2 other parameters play a lesser role within the Schoeps-circuit compared to Idss? Or is Idss just an additional criteria?

Then I have another questions concerning @MicUlli 's findings: Even though capacitance seems to be a problem, as @ricardo describes here and @MicUlli also states in the PDF,

The reason many RF FETs are used in commercial mikes is that the input capacitance (and more importantly Cgd) reduces the 'gain'. This is one reason why K118 is often better (quieter) than K117 in a real mike.

it doesn't seem to be a problem in the the tests that have been conducted. Since the JFET with the highest capacitance scored best in noise performance irrespective of the capacitance used as a capsule replacement. Why is that?
And if that test is somehow flawed, as Ricardo insinuates, which parameters of a real capsule would invalidade the findings?

Best
Jannis

 

[analogguru]

I would be interested to see how a 2SK192A (or other RF-FET), a GaAs-FET or a DG-Mosfet (like the BF981) are performing under the same test-setup.

 

[godzich]

For those of you not familiar with the book "The art of electronics (by P. Horowitz and Winfred Hill) I would warmheartedly recommend reading Chapter 8 (Low Noise Techniques). There is also a chapter (8.6.1, ED3) about Low Noise Design with JFETS and a quite comprehensive list of the lowest noise J-FETs available (page 516 in ED3). Generally thorough and well-written text and much about JFET noise optimization. Highly recommended read!

 

[Gold]

I have what might be fake 2N3819’s. One channel in my lathe electronics was noisier than the other. It would also intermittently cause a low frequency noise. After chasing connections and checking for dirty pots I zeroed in on the Feedback Amplifier. It is a high gain circuit like a microphone preamp. I figured the 2N3819 at the front end might be the culprit.

I looked in my stash and I saw some that looked like the had an A suffix. That usually means low noise so I figured I’d give it a shot. No bueno. Didn’t pass signal. Fake? I put in a regular 2N3819 and it worked and was quieter than the previous 2N3819. I changed all the bjt’s also and that made it even quieter. Now the two channels match well noise wise.