Comparison of JFETs for mic applications

[ricardo]

And forgive my asking but does TINA simulate resistor noise properly?

Yes. But... It has a funky way displaying it.

Can we show this funky display of the noise spectrum for just a couple of different resistors? 2k4, 4k7, 9k1, 18k & 36k would be nice. Ideal resistors are fine.

I didn't claim SimpleP48 would cure Global Warming and bring World Peace. Only that it would be quieter than your SupaDupa circuit in certain frequency bands.

So a good noise spectrum display is obviously essential for this shootout. Constant Bandwidth instead of the Constant Relative Bandwidth on page 10 of Zephyr.doc is OK But log frequency scale please. Linear Frequency will scrunch up the important parts of the display so we won't be able to see much.

Can TINA produce a noise spectrum curve like the one on page 10 of Zephyr.doc ?

If the answer is 'No', then we have to find a different way to investigate my claim that SimpleP48 is quieter than your SupaDupa circuit in certain frequency bands. Any suggestions for alternative methods?

---

I'll talk about the requirements for the 'THD with level' side of the shootout in separate posts. I'm still interested in this though I'm sure your SupaDupa circuit will have as great an advantage over SimpleP48 as it has in number of parts.

As a start, are you happy to use SimpleP48RCA for this (and the noise test if you wish) instead of the original SimpleP48? This equally complex circuit was in direct response to people who wanted lower THD & higher max. spl handling over the original. Should not be a problem as this is a sim. shootout.

 

[ricardo]

Much of today's unserious software runs rings around what was possible under DOS.

Moderators, I beg your indulgence for some wi**y wan*ing, an opportunity not often available to beach bums like me in such august company

When I emerged after nearly a decade in the bush circa 2008, I found that though bells, whistles and fancy pics/animations bla bla were now commonplace, the ability to do useful stuff in software was still pretty rare.

In the meantime, I found the power of even cheapo computers had increased so much such that stuff I'd only dreamed off was now, not only possible but easy even with my primitive DOS compilers.

I present 2 examples.

Circa 1990, I dreamt up, in theory, Professore Farina's Log sweep to measure Response & THD in the theoretically shortest possible time. At that time, the computing power, and especially the A/Ds were too $$$ as I wanted lots for Factory Test.

More than 1.5 decades later, Angelo visited dis beach bum in Cooktown and was surprised to be shown Jurassic code that replicated his method. I use my own versions of Angelo's method to measure speakers, microphones bla bla

The other example is How to measure, align and dream up Digital Filters for a Tetrahedral Ambisonic mike. I know this is something among Thor's many Supa powers. I met up with Peter Craven, the co-inventor of the Soundfield Mike, at AES, San Francisco 2006?, and presented the results and also the commercial manifestation of these musings.

Stan Lipshitz made useful comments on my digital juju there too

You may like to ask Prof. Farina, Peter Craven (whom Michael Gerzon said was the smartest person he knew) and Stan about their opinion of the ravings of the "Mad Chinaman"; one of my nicknames in my previous life.

Alas, Michael Gerzon has passed on so I can't ask him for a reference. I met him and Peter circa 1980 when I was doing the Mk4 Soundfield; IM not so HO, the best microphone of the 20th century. Certainly the most $$$ and complex which should surely gladden Thor's heart

PERHAPS you should consider to be dragged kicking and screaming into the windows age. Or Linux for that.

Alas. I've tried to do this on 3 separate occassions this Millenium but was thwarted by computers dying on me. This is my 4th attempt but I need to have something Jurassic working to check what my software should do.

---

Now back to software mike shootouts ...

 

[thor.zmt]

When I emerged after nearly a decade in the bush circa 2008, I found that though bells, whistles and fancy pics/animations bla bla was now commonplace, the ability to do useful stuff in software was still pretty rare.

Well, you are 100% wrong. Just saying. And all software managed to run on Windows 7, no problems.

I guess anyone looking for useful computer software to do useful work will find it. And anyone looking to have his prejudice confirmed that only programs under DOS can do useful work will find confirmation.

Alas. I've tried to do this on 3 separate occassions this Millenium but was thwarted by computers dying on me. This is my 4th attempt but I need to have something Jurassic working to check what my software should do.

I just packed up my little workshop and moved it.

My "Labrat" Komputermaschiene is an antique Lenovo Thinkpad T440. For lab use, recording and DJing I always buy an ex lease 3 year old minimum spec T series Thinkpad.

Extremely reliable. Extremely Serviceable, even a 10 year old T series Thinkpad will have most any FRU (Field Replaceable Unit) available.

The keyboard will survive a coffee spill and the whole machine is easily folded up and can be used to beat latter day luddites to death, then opened again and continue emails.

As T series Thinkpads (at least up the one I got) have replaceable CPU, Memory and HDD's, pay 100 Bux for the machine, pay another 100 Bux for a totally obsolete top of the line Intel CPU fitting it, 16-32G in memory sticks, a 256G SSD and you are off to the races.

Mine runs of course all the sim software and EDA (though real work is done on the big workstation). Attached to this 15" Screen laptop are:

Quantasylum Audio Analyser QA404

RIGOL 4 Channel 100 MHz DSO

RIGOL Function Generator

MINIDSP EARS calibrated against G.R.A.S H.A.T.S with Calibration file

MINIDSP UMIK USB measurement microphone Calibrated against Earthworks with calibration file

Electronic Load for testing powersupplies and batteries

What can be done with small setup that fits a small work desk is mind-blowing to a guy who started out in the Comecon.

Packed up it easily fitted a 100 liter box on wheels including my old-fashioned 3 Channel 5A Lab supply.

The abilities modern computers and computerized, computer controlled test gear are off scale compared to what we made do in the 80's, extremely affordable.

And modern simulation software goes as far as simulating passives with S-parameters if you are thusly inclined.

We absolutely live in a golden age when it comes to what computers do for us to support audio design work. My simple desk can do stuff almost instantly that a complete engineering department would take month in the 80's.

Thor

 

[thor.zmt]

Can we show this funky display of the noise spectrum for just a couple of different resistors? 2k4, 4k7, 9k1, 18k & 36k would be nice. Ideal resistors are fine.

TINA displays either noise density (spectrum) or SNR with reference to an arbitrary level across bandwidth.

I didn't claim SimpleP48 would cure Global Warming and bring World Peace. Only that it would be quieter than your SupaDupa circuit in certain frequency bands.

Which frequency bands would that be?

The numbers for 20-20k SNR re 94dB did not favour SP48.

If the answer is 'No', then we have to find a different way to investigate my claim that SimpleP48 is quieter than your SupaDupa circuit in certain frequency bands. Any suggestions for alternative methods?

Simple inductive circuit analysis?

SP48 operates the J-FET as transconductance element with no degeneration into in effect 6.8kOhm load, giving with a 1mA/V transconductance J-Fet a gain of ~ 7.

With a transconductance of 1.2mA/V we expect an equivalent noise resistance of our J-Fet of 1V/1.2mA or around 833 Ohm, giving ~ 0.5uV noise in the flat band, or around 4nV|/Hz.

We have the full influence of the remaining input capacitance and miller capacitance, attenuating the microphones signal for SP48.

With 1.2pF Crrs and a gain of 7 we get 8.4pF plus 4.5pF Ciss, for ~ around 13pf load capacitance on a 44pF microphone, causing ~ 2.3dB attenuation.

So SNR of THD raw J-Fet is degraded by 2.3dB. The 6.8k load(s) contribute ~ 1.5uV each, or ~ 2.1uV combined.

The 0.5uV noise of the J-Fet sees an input referred load resistor contribution of around 0.4uV, for a total noise of 0.65uV

The cascoded split load inverter in my circuit in effect eliminates any capacitive effects. The gain to each load resistor is 1, but each resistors contributes it's noise.

As these resistors need to be scaled for ~ 5V according to the actual Idss, I used the highest Idss class with 10k loads.

The load noise will be ~ 1.8uV per resistor.

So purely Ein at the gate will be around the 1.8uV, theoretically an 8dB noise disadvantage, IF NO OTHER SOURCES CAUSE EXCESS NOISE.

---

I'm still interested in this though I'm sure your SupaDupa circuit will have as great an advantage over SimpleP48 as it has in number of parts.

It does. And my design was made for studio recordings, not live in a concert hall.

Of course, adding a single capacitor (say 33pF S-G for a 44pF capsule) could improve that a lot.

As a start, are you happy to use SimpleP48RCA for this (and the noise test if you wish) instead of the original SimpleP48?

Moving goalposts already Richard?

How about NO.

I am ok to ADD this.

Please post the exact circuits you want me to use, assuming a J-Fet with ~ 370uA IDSS and a Yfs of ~ 1.6mA/V (this is the model I have to work with).

And a capsule with 44pF capacitance and -35dBV Sensitivity and 26mm Diameter (11mm Radius for Brownian motion noise).

Thor

 

[gyraf]

I have several pieces of essential software that requires me to keep hardware alive for both DOS/Win3.11 and winXP.

Some programs simply won't run on modern hardware (or rather modern abstraction layers). Try to real-time a parallel port or mount a paranoid protection dongle, and you'll know what I mean..

@ricardo - are you just looking for a xp install cd, or for the whole computer?

/Jakob E.

 

[thor.zmt]

Some programs simply won't run on modern hardware (or rather modern abstraction layers). Try to real-time a parallel port or mount a paranoid protection dongle, and you'll know what I mean..

I had this problem with an AP2 that needed an ISA Card and an antique version of windows. Eventually the last PC with an ISA slot in the part of Scheena where I was died. So now I had a 10k USD boat anchor.

Solution, USB to AP2 interface, use USB 2.0 isolator and also kill the ground loop that plagues the the AP2. Run a laptop. Could have done years earlier, burlt wasn't on the old inertia thing if it ain't broke.

As was, the AP was down for Several weeks.

At this point in time I have invariably found alternative (and usually better) more modern software, hardware or suitable bridge hardware, to move forward and get off unsupportable legacy hardware.

It seems easier to keep old hardware/software alive, beyond a certain point I find it counterproductive and borderline latter day ludditism.

And I'm really conservative on hardware / software replacement to start with.

Example, using an old pentium to get hardware control of a parallel port, vs running an I5 10 gen with a USB 2.0 480M/sec to parallel dongle and emulation.

Which REALLY responds faster? Virtual machines and multiboot allow configurations that remove componts that drag down performance.

For example, I find having a second boot configuration on my laptop that is maximally slimmed down for DAW / DJ Software use. That is literally a second install of the OS, double boot, configuration for best performance, low latency etc.

Anyway, I'm preaching to those who probably don't want to listen. But if you rely on a legacy system enough to keep it alive well past it's "Recycle by" date, you probably should start looking hard enough for a replacement or way out. And if not, let it die, move on.

Otherwise we'd all still 50MHz tube scopes.

Thor

 

[rh001]

Updated: see post #33 for more measurements
Good morning all!

I thought I'd share with you some measurements I've been making of various different JFETs, for use in impedance converter / amplifier applications in condenser mics. Specifically, how do the JFETs compare in the very widely-copied "FET amplifier + emitter follower" transformer-based design. RecordingHacks attributes this to the ADK A-51 originally (see http://recordinghacks.com/microphones/ADK/S-51) and it can be found in the MXL 2001 and V67 (https://groupdiy.com/threads/mxl.84887/), sE 2000 (https://groupdiy.com/threads/se-electronics.70694/) and relatives, Groove Tubes GT-55 (https://groupdiy.com/threads/groove-tubes.45261/) and many, many others.

I used the following circuit:

View attachment 122184


The core of the circuit is the FET itself, Q1, drain resistor R4 and source resistor R3 (which set the AC gain), and variable bias resistor R2+VR1, which set the DC operating point. Almost all the clones use R4=20K. R3 is usually in the range 2K2-4K7, I'm using 2K2 for the maximum gain, as this will emphasise differences between the FETs.

There's an input socket J1, and the board allows the input coupling capacitor C_IN and the FET itself to be easily changed. For simplicity, the circuit was powered from a bench PSU and I omitted the output transformer - the output goes directly to an audio interface (and/or scope) for measurement.

Here it is in its box:

View attachment 122185

The FETs I had available were:

• 2N3189
• 2SK117-BL
• 2SK118-GR
• 2SK170-BL
• 2SK208-GR
• 2SK30A-GR
• BF244B
• J112

Next post: basic DC/AC measurements

 

[ricardo]

TINA displays either noise density (spectrum) or SNR with reference to an arbitrary level across bandwidth.

Noise density spectrum please. Preferably constant relative bandwidth, eg 1/3 8ves. Like moamp's B&K curve or my curve on page 10 of Zephyr.doc But constant bandwidth will do.

Which frequency bands would that be?

The frequencies above about 1kHz. Exact start depends on the '1G + capsule' chosen. You see this clearly on page 10 of Zephyr.doc

Please post the exact circuits you want me to use, assuming a J-Fet with ~ 370uA IDSS and a Yfs of ~ 1.6mA/V (this is the model I have to work with).

And a capsule with 44pF capacitance and -35dBV Sensitivity and 26mm Diameter (11mm Radius for Brownian motion noise).

Please have patience with an old beach bum as it takes him time to check dis new fangled stuff.

Is your FET the one in the LTspice library? If not, could you post the model here?

The capsule + FET needs to be quiet enough not to affect the 'electronic' noise at 4kHz ... like Zephyr's mike.

I'm not sure assuming 11mm radius for Brownian noise is quiet enough for this. Zephyr's mike is VERY quiet but still well above the limit of Brownian noise

What do you make the Env for this Brownian noise to be? Xcuse mi but doin' sums is hard 4 my small brain

 

[k brown]

Well, you are 100% wrong. Just saying. And all software managed to run on Windows 7, no problems.

I guess anyone looking for useful computer software to do useful work will find it. And anyone looking to have his prejudice confirmed that only programs under DOS can do useful work will find confirmation.





I just packed up my little workshop and moved it.

My "Labrat" Komputermaschiene is an antique Lenovo Thinkpad T440. For lab use, recording and DJing I always buy an ex lease 3 year old minimum spec T series Thinkpad.

Extremely reliable. Extremely Serviceable, even a 10 year old T series Thinkpad will have most any FRU (Field Replaceable Unit) available.

The keyboard will survive a coffee spill and the whole machine is easily folded up and can be used to beat latter day luddites to death, then opened again and continue emails.

As T series Thinkpads (at least up the one I got) have replaceable CPU, Memory and HDD's, pay 100 Bux for the machine, pay another 100 Bux for a totally obsolete top of the line Intel CPU fitting it, 16-32G in memory sticks, a 256G SSD and you are off to the races.

Mine runs of course all the sim software and EDA (though real work is done on the big workstation). Attached to this 15" Screen laptop are:

Quantasylum Audio Analyser QA404

RIGOL 4 Channel 100 MHz DSO

RIGOL Function Generator

MINIDSP EARS calibrated against G.R.A.S H.A.T.S with Calibration file

MINIDSP UMIK USB measurement microphone Calibrated against Earthworks with calibration file

Electronic Load for testing powersupplies and batteries

What can be done with small setup that fits a small work desk is mind-blowing to a guy who started out in the Comecon.

Packed up it easily fitted a 100 liter box on wheels including my old-fashioned 3 Channel 5A Lab supply.

The abilities modern computers and computerized, computer controlled test gear are off scale compared to what we made do in the 80's, extremely affordable.

And modern simulation software goes as far as simulating passives with S-parameters if you are thusly inclined.

We absolutely live in a golden age when it comes to what computers do for us to support audio design work. My simple desk can do stuff almost instantly that a complete engineering department would take month in the 80's.

Thor

True, that.

Every time one of my ~10-year old Mac laptops starts acting up, my wife says: "Well it's really old - that computer's not gonna last much longer." I have ThinkPads made in the early '90s that all still do everything they did when I got them. Not one has developed a single issue other than the intenal clock memory battery dying.

 

[rh001]

Hi Voyager10:

Nice work. I read as far as replies page 8, then got tired, but my takeaway is that any of the fets you tested will do...if this was sports a photo finish would be needed to tell who won. But omg can you imagine how one might do actual music listening tests to find "the best fet". That would be so hard to do and time consuming... how many different musical instruments to serve as a usable subset of all instruments and then, what about the player and the quality/characteristics of the instruments and the rooms. My hunch is that they will not sound significantly different as a mic capsule amp input but who knows cuz surprises happen.

But since you have the fets and the test jig you could at least, with a small perhaps crude capsule simulator interface on preamp input, pipe some different music recording through the circuit and record them and then AB the results. For reasons unknown one or a few might have a sonic edge over the others.

 

[thor.zmt]

Noise density spectrum please. Preferably constant relative bandwidth, eg 1/3 8ves but constant bandwidth will do.

TINA doesn't really do these, but we can still see comparative results.

Zephyr.doc

Please stop mentioning this, until all the gross inaccuracies are corrected. And no, I will not try to make curves that are fundamentally wrong.

The frequencies above about 1kHz. Exact start depends on the '1G + capsule' chosen. You see this clearly on page 10 of Zephyr.doc

If we use Diode bias and no "1G resistor" the LF noise is quite different. And please stop promoting something fundamentally wrong and instead get it fixed.

Please have patience with an old beach bum as it takes him some time to check dis new fangled stuff.

It's not "newly fangled", it's just a different way to represent the same phenomenon graphically.

The capsule + FET needs to be quiet enough not to affect the 'electronic' noise at 4kHz ... like Zephyr's mike.

At 4kHz the noise is not "electronic", in most cases it will be brown(ian motion) noise.

I guess I'll have to start making an accurate capsule model first, including noise sources that are inherent to it.

Than we can look at circuit.

That looks like degenerating I to work.

Work is the enemy of the drinking class.

Thor

 

[ricardo]

TINA displays either noise density (spectrum) or SNR with reference to an arbitrary level across bandwidth.

How does TINA display noise density (spectrum)? Can we try this out with some resistors?

I've no doubt TINA calculates resistor noise correctly. Just need it to display the noise spectrum in a manner useful for our shootout.

 

[ricardo]

Some programs simply won't run on modern hardware (or rather modern abstraction layers). Try to real-time a parallel port or mount a paranoid protection dongle, and you'll know what I mean..

@ricardo - are you just looking for a xp install cd, or for the whole computer?

Thanks for the offer Jakob. An XP install CD please.

In my latest attempt to enter the 21st century, I finally managed to get a another working XP machine but I mucked it up I need an XP install CD to repair it.

But if you have an XP machine spare ...

 

[JohnRoberts]

the benefits of entering the 21st century are over stated.

I hear that surf's up.... (illigitimi non carborundum).

JR

 

[ricardo]

the benefits of entering the 21st century are over stated.

I hear that surf's up.... (illigitimi non carborundum).

Thanks John

The reason I'm so slack in answering Thor is I'm trying to organise a couple of Dive trips out of Cooktown. We are at the Great Barrier Reef's doorstep