Comparison of JFETs for mic applications

[jp8]

Submit it to a fading low-frequency sinewave. I used to do that with a simple potentiometer, but today I can use a tone generated by a DAW.
I believe this method is documented in a Studer paper.

Or did you mean Jensen? See chapter 4.2 here.

Jan

 

[thor.zmt]

Or did you mean Jensen? See chapter 4.2 here .

Bill is in part correct.

I found the results far les subtles than he intimates, back in the 80's.

Anyway, a module from a mixing desk to be serviced could not pass basic acceptance tests with a magnetized transformer. Many that came had some gotten "magged". Must have been these wicked western imperialists magnetizing upstanding socialist Transformers of the people.

Secondly, what we are trying to do is to "re-randomize" the magnetic dipoles in the magnetic material after they have become aligned into a specific polarity. This condition is kinda self-reinforcing. It usually takes to beyond saturation (hence my implied suggestion of 5Hz +20dBu) to overcome.

Another issue is that the fade must be very slow, an excessively fast fade doesn't guarantee sufficient randomization. We "Go gray" pretty slowly.



As the signals drops so does the number of magnetic dipoles moved. Adding a significant amount of white noise to the VLF sinewave can be more effective and allows faster fade out.

Magnetized cans be PITA to clear, but commercial TV Tube demaggers (if they can be found) will do.

Thor

A man in a desolate train station
A suitcase at his side
Cold, fixed eyes
Show fear when he
Turns to hide

Feel the rain like an English summer
Hear the notes of a distant song
Coming out from behind a poster
Hoping life wasn't so long

Turning gray

 

[dai h.]

Or did you mean Jensen? See chapter 4.2 here.

Jan

hi,

(check out page 50) :

https://www.worldradiohistory.com/A...ive-Studio-Sound/80s/Studio-Sound-1987-05.pdf

(and page 5--this is the same pdf from the Studer ftp link in the old thread below (which no longer works)) :

http://revoxsammler.ch/revoxsammler-neu/quellen/SwissSound/english/Swiss-Sound-18-1987-02-e.pdf


(the Studer info has been referenced a few times as in these 2009 threads) :

https://groupdiy.com/threads/studer-169-269-de-magnetizing-input-transformers.32851

https://groupdiy.com/threads/some-strange-audio-transformer-response-curves.31454/#post-382253

p.s.

this link also has the same Swiss Sound issue, but also lots of back issues as well:

https://www.reeltoreel.nl/studer/Public/SwissSound/

 

[ricardo]

For the rest...

View attachment 142938
Predicted is an unweighted 14dB 20dB self noise 20Hz-20kHz, so perhaps 10dB(A) 16dB(A) without accounting for Brownian Motion Noise and acoustic resistances etc. I would say "low enough for music recording" but not phantastically low.

Thanks for this Thor.

It's unfortunate your starting point is a particularly bad version of BM800. Some of its sins have passed into your circuit which would be noisier than Zephyr's version of the Schoeps from above 1kHz culminating in more than 11dB extra in the aurally important 4kHz and above.

I don't think you would get 16dB(A)

This BM800 circuit responds well to Muntzing for better performance. Your regulator improvements are worthwhile but replacing it entirely with cheapo resistors & capacitors gives even better performance ... unless you need the mike to work in P24 & P12 situations. There is at least 9.2dB S/N improvement available on top of this.

I would keep the HP filter. Condensor mikes need subsonic filtering.

Zephyr's mike with 34mm ISK 'C12' (2011 vintage) clone capsule is a very quiet mike but even here, the electronics contribute at least 1dB extra noise. So we are still far from Acoustic Resistance noise or Brownian angels dancing on the diaphragm.

Schoeps size SDCs would probably have enough acoustic resistance to dominate even Zephyr's circuit ... which has up to 6dB noise advantage for variable pattern LDCs in some modes.

Of course if the mike is only used for singers in a booth, these noise improvements are moot. But there would be certainly appreciated if they were the main Blumlein pair in the Royal Albert Hall for the Proms.

I hesitate to point out SimpleP48 has even better noise ...

Shut up ricardo ! Just SHUT UP !!!

 

[thor.zmt]

It's unfortunate your starting point is a particularly bad version of BM800.

Not at all. Here is what I make from it:

Some of its sins have passed into your circuit which would be noisier than Zephyr's version of the Schoeps from above 1kHz culminating in more than 11dB extra in the aurally important 4kHz and above.

I am not sure where you get that impression from.

This BM800 circuit responds well to Muntzing for better performance. Your regulator improvements are worthwhile but replacing it entirely with cheapo resistors & capacitors gives even better performance ...

Not in a compact format. We have a passive 100k/10uF filter to the base of a low noise, super beta npn transistor, a transistor others have used in MC Pre-Pre's.

The Zener Diode will have perhaps 10uV noise 20-20k, our RC filter will have 0.16Hz turnover, so -40dB at 16Hz, I wager the zenner noise is filtered out well enough, to not be of concern in the audio band.

unless you need the mike to work in P24 & P12 situations.

This Mic circuit I show is P48 only.

There is at least 9.2dB S/N improvement available on top of this.

I rather doubt that, at least where the power supply is concerned.

I would keep the HP filter.

I would ABSOLUTELY NOT keep it. Especially the exact one as shown.

If I need (V)LF filtering, I do it elsewhere.

I hesitate to point out SimpleP48 has even better noise ...

Does it now. Just how much better, as we are doing a contest?

Let's agree on the C-12 capsule. I'll pull the tech data.

2SK660 used for both? How do we bias the capsule in the Simple48P?

I find the TINA simulator quite close to reality, I'll run the numbers.

Noise, THD @ 94, 104, 114, 124 and 134dB if it still makes sense. Would you like something different.

Thor

 

[ricardo]

View attachment 143243
This Mic circuit I show is P48 only.

I hesitate to point out SimpleP48 has even better noise ...

Does it now. Just how much better, as we are doing a contest?
I find the TINA simulator quite close to reality, I'll run the numbers.

SimpleP48 has about a 1dB noise advantage over a properly designed classic Schoeps mike in the cases where the 'electronic' noise is sorta verging on being prominent. It has MUCH more advantage over poor implementations like BM800.

Let's agree on the C-12 capsule. I'll pull the tech data.

2SK660 used for both? How do we bias the capsule in the SimpleP48 ?

Alas, SimpleP48 only works with electrets

Can we use a hypothetical capsule with the same capacitance as CK12 and assume Vp 60V ?

Noise, THD @ 94, 104, 114, 124 and 134dB if it still makes sense. Would you like something different.

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

I'm interested in THD @ 94, 104, 114, 124 & 134dB spl of SimpleP48. But I concede your SupaDupa circuit will have lower THD.

The SimpleP48 contender for THD is SimpleP48RCA on page 12 of SimpleP48.pdf

As this is a sim. challenge, could we use that instead? THD of SimpleP48RCA would be of great interest to many including myself. Even then, I think it is 'good' only to 130dB spl.

Let me dream up the values for my incredibly complicated circuit

 

[ricardo]

Thor, who's 2sk660 SPICE model are we using?

I'm worried about the 'reverse diode' across G-S. Guru Scott Wurcer points out the reverse current at low voltage should be about 1pA.

No maker specifies anything that low as it's very temperature & voltage dependent. But that's what it needs to be.

eg MicUlli says only Infineon & NXP BAV199s are good enough for this purpose but their datasheets show reverse current in the nAs.

---

And forgive my asking but does TINA simulate resistor noise properly? We can test this by doing noise spectrums with just various resistors like I do on page 10 of Zephyr.doc

---

Can we use a hypothetical capsule with the same capacitance as CK12 and assume Vp 60V ?

This is just to get a 'sensitivity' figure. We could use -66dBV/Pa which would be within 2.5dB of Zephyr's Schoeps variant with the ISK 34mm capsule.

---

I want the 'capsule + FET' side to be LN enough to clearly see the differences in the rest of the electronics ... as is the case with Zephyr's mike.

This would be useful to all and sundry ... as would THD at different spls for this 'hypothetical' capsule with SimpleP48RCA.

Much better than a "mine is bigger than yours!" shootout Much better use of your time and effort too for which I'm sure everyone here will be grateful.

 

[ricardo]

Does anyone have a pukka Windows XP install disc they could send to a beach bum in Cooktown, Oz?

I need a real XP machine to run my own software in case Thor asks mi mo hud kwestens. My last one has died.

Moderators, please move this cry for help to an appropriate forum if required.

 

[thor.zmt]

Thor, who's 2sk660 SPICE model are we using?

LT spice J-Fet Library.

I'm worried about the 'reverse diode' across G-S.

Must be external. It is not part of the J-Fet model.

Guru Scott Wurcer points out the reverse current at low voltage should be about 1pA.

No maker specifies anything that low as it's very temperature & voltage dependent. But that's what it needs to be.

The diode is not modelled, I added a BAV99.

When you diffuse a J-Fet you are implicitly making a low leakage diode (an N-Channel J-Fet is basically a diode with two cathodes and one anode (Gate).

I expect the Bias Diode made on the exact same diffusion process to be equal to the G-S/D diode.

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

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

This is just to get a 'sensitivity' figure. We could use -66dBV/Pa which would be within 2.5dB of Zephyr's Schoeps variant with the ISK 34mm capsule.

-66dBV!!!???

That is, 0.5mV/94dB?

That's around 40dB less than I expect.

Normally the bigger capsules are all around -34dBV/94dB @ 60V.

Well implemented Schoeps has ~6dB gain, so I expect -28dBV/94dB (1PA) not -66dBV/94dB.

---

I want the 'capsule + FET' side to be LN enough to clearly see the differences in the rest of the electronics ... as is the case with Zephyr's mike.

In my circuit there are not really any such sources of noise. The J-Fet load resistors are the main noise source.

The BSS84 has 1.5nV|/Hz with a bit of excess 1LF noise rise worse than J-Fets.

As we have two of these balanced, around 2nV|/Hz for the MOSFET follower, let's be generous for the 1LF noise rise and call it 0.5uV 20-20k unweighted.

The powersupply for the J-Fet frontend and phasesplitter is basically the noise of the BC337-40, or under 1.4nV|/Hz.

It will only add to the inverted phase of the circuit. So it's contribution on the output will be swamped by the BSS84 followers. Let's include in the 0.5uV output noise figure snd allow 3dB more, so 0.7uV noise.

The 2SK660 is specified as 1uV A-weighted typically and 30k of split load inverter resistors give around 3uV. The J-Fet is a pretty high impedance, so this noise will dominate.

Eyeballing all noise components I'd call ~ 2uV A-weighted.

My Capsule sensitivity is -34dBV/94dB (20mV) @ 60V bias and in my case 35V bias drop this to ~ 12mV.

So SNR ~ 76dB(A) @ 94dB SPL, or 18dB(A). I'd give it +/-3dB tolerance. My samples tested somewhere around there, I was aiming for < 20dB(A), which is lower than many a Chinese made sub 1k "Professional Recording Microphone". And no, that doesn't mean BM-X00.

More bias claws back a few dB SNR and looses SPL handling at high SPL.

Lowering the load resistors increases HD and lowers maximum SPL, but claws back a few dB SNR. I was more focused on SPL handling than lowest noise.

My actual mic's use a E202 CRD to feed the zener diode, so current draw is constant. I actually use P68 instead of P48 which restores capsule bias to ~ 60V. It's nonstandard, but making 68V where commonly you see 48V is usually not that hard (was not on my Yamaha 01V96MK2)

As noted, just a rainy Sunday afternoon project meat to be better than the common "China Mike" stuff.

This would be useful to all and sundry ... as would THD at different spls for this 'hypothetical' capsule with SimpleP48RCA.

In the middle of moving, I'll get back to it later.

Let's just settle on the new 34mm Electret Capsule, brass case. No bias needed.

Much better than a "mine is bigger than yours!" shootout Much better use of your time and effort too for which I'm sure everyone here will be grateful.

I usually simply show what I recommend and expect considered constructive criticism.

If others promote something, I will look and consider if there are weak points and possibly issues. And I feel free to point them out.

On the SimpleP48 I see the unbalanced current draw as the main deal breaker, the not at all great HD levels as a second. It will pass a signal, but then, a lot will.

Does anyone have a pukka Windows XP install disc they could send to a beach bum in Cooktown, Oz?

What is windows XP!? !? !?

Run it on 11, set the software compatibility to XP. So far works for me every time.

Thor

 

[gyraf]

You can get winXP for virtualbox - e.g. https://archive.org/details/xp51_20191108

 

[thor.zmt]

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

No. Incidentally, all the numbers there are science fiction. Behringer ECM8000 = -60dB sensitivity? What is Zephyr Smoking? 156dB undistorted SPL?

Behringer ECM8000 = 8mV/94dB/1PA and thus sensitivity= -42dBV/PA.

All numbers in there need an 18dB correction factor. And who knows how reliable those are.

I'm not sure I want to entertain something where so basic fundamental assumptions are inaccurate. I lack any confidence anything else will be of better congruence with reality.

The DOC is dated 12 years ago!?

In all that time nobody picked up on the basic errors in this? For Fox sake!

AND I BLEEDIN' WELL HAD TO SIGN UP TO IO GROUPS FOR THAT! Thank you Richard for wasting my time.

Thor

 

[thor.zmt]

Can we use a hypothetical capsule with the same capacitance as CK12 and assume Vp 60V ?

Let's stick with something real.

I used these 26mm Cardioids capsules in a front address mic:

https://a.aliexpress.com/_okEySpP

Sensitivity: -35dBV/PA (+/-2dB)
Capacity: 42pF +/- 4pf

Let's work with design center values.

These are reasonably close based on my experience.

They are overall also not that bad, not LDC, but decent MDC... I literally used the same circuit minus bias but on a narrow, long PCB.

It would actually be worthwhile for someone else making a decent basic mic to use this capsule.

The mic body I used was a copy of the one they use on "Voice of Scheena".

Lots of Mongolian Throatsinging I expect. I never watch Chinese Children's Television (CCTV), so I wouldn't know.

Nice suspension integrated. Nice cast body.

Result reminded me a lot of the EV BK-1 which I loved in the 80's as vocal mic's.

Thor

 

[micolas]

Let's stick with something real.

I used these capsules in a front address mic:

I just found this on AliExpress: US $28.00 | 2PCS 26mm Diaphragm Condenser MIC Capsule Microphone Cartridge Electret Cardioid https://a.aliexpress.com/_okEySpP

Sensitivity: -35dBV/PA (+/-2dB)
Capacity: 42pF +/- 4pf

Let's work with design center values.

These are reasonably close based on my experience.

They are overall also not that bad, not LDC, but decent MDC... I literally used the same circuit minus bias but on a narrow, long PCB.

It would actually be worthwhile for someone else ting a decent basic mic to use this capsule.

The mic body I used was a copy of the one they use on "Voice of Scheena".

Lots of Mongolian Throatsinging I expect. I never watch Chinese Children's Television (CCTV), so I wouldn't know.

Nice suspension integrated. Nice cast body.

Result reminded me a lot of the EV BK-1 which I loved in the 80's as vocal mic's.

Thor

Hi Thor, I'm sorry to bother you,
the AliExp link doesn't work for me

 

[GVK]

Works for me. Copy it from https onwards

 

[thor.zmt]

Hi Thor, I'm sorry to bother you,
the AliExp link doesn't work for me

Corrected.

 

[micolas]

Corrected.

Okay Mr. Thor!
Thanks!

 

[thor.zmt]

Okay Mr. Thor!
Thanks!

I would also like to clarify what I propose.

Buy a BM-X00 microphone. Use it for the mechanical parts. Repaint the nasty anodised pink alu (somehow the cheapest are always pink) in hammertone spray paint.

Strip the BM-X00 circuit board completely. Rebuild to my design. Make sure the parts are genuine, for SMD capacitors, resistors and FET's that pretty much means mouser.

Test the circuit board properly.

Use either my bias chain and a 34mm china capsule to taste, or this (and similar) 25...26mm upper end electret capsules with a suitable saddle.

Make a complete, decent looking and competent performance microphone that will give a lot out there a run for the money, with minimal effort.

Make a bunch with a bunch of different capsules and patterns, who said you must limit yourself to two Mic's?

Will it be better than a vintage or true modern replica of a C-12, U87 etc. et al? Nope. It will be different. The one with a K47 capsule copy and Neumann Style head basket will sound different to the one with the C-12 capsule copy and AKG style headbasket.

Can we do better? Of course.

Can we do easier? Of course.

Can we do cheaper? Of course.



Thor

 

[ricardo]

You can get winXP for virtualbox - e.g. https://archive.org/details/xp51_20191108

Alas. I need an actual Windows XP machine. Machines that have the correct drivers for XP are all at least 10 yrs old.

It's complicated as all my serious software was written in DOS days. I run a virtual Windows 98 on XP and run a DOS window from that.

Versions of Windows after XP don't have the facilities in their virtual machines to do what I need.