Comparison of JFETs for mic applications

[thor.zmt]

PPS, I will have a go to see how much of these 10dB I can claw back with "Guru" Level circuit, without compromising the rest, but I'm quite happy as is.

Ok, a quick adjustment of the original circuit, which still fits the BM-X00 PCB and is limited by this board. I removed all other artificial restrictions and instead optimised for the 26mm Electret Capsule (no need for Bias from P48).

Let's remember what the original looks like:






All images from https://audioimprov.com/

So, can we, in simulator improve both noise and HD/SPL handling for use with our 26mm goldsputtered diaphragm (basically a proper mic diaphragm, not capacitor foil) backplate electret Capsule?




We observe some HF boost we would like to EQ.

So, can we do better than my original Schoep's hack that I concluded quite a few years back were good enough, especially in view of the Mic-Preamp's and ADC's in my Yamaha 01V/96MKII desk? We can.



First, output stage current is increased to 1.84mA from 1.2mA, to push current starved operation to higher SPL's.

It uses the previously discussed bootstrap follower as phase splitter around M1. All Mosfet's are the previously discussed low noise P-Channel FET*, J-FET the previously discussed Mic-Fet**.

The miller feedback capacitor C2 acts as EQ with some HF roll-off to make the sound less bright.

Gain is just a smidgen below 2, it is corrected to exactly 2 by the VCVS to allow normalisation of all results. With a -35dB/PA capsule actual sensitivity will be -30dB/PA, also a pretty loud microphone.

Frequency response with EQ:



On the response plot above 20kHz looks like +3dB and 10kHz looks like +4dB

So the -2dB @ 10kHz just take a bit of the top end edge off without killing "air" and "sparkle". We should end up with +2dB @ 10kHz & -2dB at 20kHz. In practice C2 is best tuned on either analyser or by ear.

So, how about noise:




Quite a bit of improvement. Within 3...4dB of the flat band noise and SNR of the Guru Supa Simple. I can live with that VERY well.



HD at 35mV out (94dB SPL), 0.006% all H2. Let's try 20dB more:



At 353mV out (114dB SPL) is 0.065% THD and higher order harmonics show up.

Ok, 20dB more, still clipping:


But we are at < 5% THD now:


Overall this modified circuit simulates similar in results to the original, but is is better in some respects (7dB less wideband noise, 1/3rd of the HD at 134dB SPL input).

Honestly, I like the shape of the clipping waveform less, the doubled H2 I'm agnostic about. I like the lower noise though I still question if it is really beneficial.

In my own recording setup I most certainly doubt that. So I don't think I would change my original MIC's, once I retrieve them from China, but I will from now on recommend this modified circuit over the original and would probably build new Mic's on this circuit pattern.

Recent 26mm & 34mm Electret Capsules start to get decent, so I'd probably just use those.

Ultimately the circuit is limited by being derived via the "generic chinese Karaoke Mic" from Schoeps. A clean sheet design an shed these limitations

I will end this veer from the original thread topic here and apologise for taking it so far off topic (not as if I didn't have a a lot of help) , though I hope we all learned something.

Thor

PS, Richard, Muntz that! Not allowed is significant unequal current draw on the Phantom Power leg's, material increases of HD (say more than 6db per harmonic), reduced SPL handling (say 3dB).

PPS, * BSS84 P-Channel MOS, pinout is same as SOT23 BJT, ** 2SK660, other options exist

PPPS, zip file with the TINA file is attached.

 

[thor.zmt]

You, on the other hand, have said about a zillion times, "results better most of the china sourced commercial LDC's for a fraction of the cost."

And I absolutely stand by that. I could make a long list of brands and models all 300-500USD and up, but that would not be constructive. I'm not here to diss stuff.

You made other claims too but searching this thread for things you 'forgot' you said is becoming a PITA

Allow me to help. Claims are here:

https://groupdiy.com/threads/comparison-of-jfets-for-mic-applications.86559/page-12#post-1183455

And your claims are here:

https://groupdiy.com/threads/comparison-of-jfets-for-mic-applications.86559/page-13#post-1184093

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

I didn't get this. Sim suggests 14dB UNWEIGHTED . Si I call this one B...Something.

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

Yes, we calculated, tested this assertion in the simulator and agreed it has lower noise. As for the rest, just look at the results osted earlier.

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.

As my quick hack job was exactly for that, with an acoustic guitar at 1' probably the quietest source you agreed my quick hackjob was good enough.

Now who used the SimpleP48 as "main Blumlein pair in the Royal Albert Hall for the Proms" and are there accessible recording.

I know things have going downhill in the UK, but it's a real shame that the BBC had to sell off all the DPA, Schoeps, Soundfield Microphones in their Inventory to make ends and now use DIY Mic's. I guess King Charls's Family silver and Crown Jewles will be next on block.

Per "DaGoogle" AI in the glory Days of the Beeb it looked like that.

The BBC Proms uses a variety of microphones to record the concerts, including stereo bars, cardioid mics, and 3D microphone arrays. The BBC uses these microphones to create binaural mixes, which are 3D sound recordings that can be played back through headphones.

Microphone types

• DPA 4006 omni mics: A stereo bar with three sets of these mics is used to create the main mix
  
  
• Schoeps MK2H omni capsules: Mounted above the stereo bar, these mics provide different tonal options
  
  
• Schoeps MK4 cardioid mics: A closely spaced pair of these mics face backwards to capture the audience and arena atmosphere
  
  
• Soundfield ST-250 microphone: Used to record ambient soundscapes

Microphone arrays

• 3D microphone array
  Used to create binaural mixes, this array is rigged at the Royal Albert Hall
  
  
• Radio Tree
  A smaller version of the original Decca Tree design, this array uses DPA 4006 mics with foam windshields

Microphone layout

• The microphone layout has remained largely unchanged since 2016
• The layout is well suited to creating quality binaural mixes

Generative AI is experimental.

https://www.bbc.co.uk/rd/blog/2016-09-binaural-proms

Hard times in the UK I guess, well, a commie government, what do you expect. Let's hope things improve soon.

Please repeat your exercise with SimpleP48RCA.

I leave that to the reader. Simulation files and models etc are for downloading.

FWIW, HD is reduced around 20dB, still nothing to write home about, levels are more than 10dB down.

Err.rh! Have you changed your expectations from #355?

No. My original target had effective -39dB sensitivity, so 4dB less than the Capsule we agreed upon. So SPL handling will be down by at least as much.

Thor

 

[ricardo]

Claims are here:

https://groupdiy.com/threads/comparison-of-jfets-for-mic-applications.86559/page-12#post-1183455

Thanks for this Thor.

If we can clarify what you claimed in this #231 post. 134dB spl <0.5% THD (H3 dominates) with -34dB/Pa capsule.

Today, you said

No. My original target had effective -39dB sensitivity, so 4dB less than the Capsule we agreed upon. So SPL handling will be down by at least as much.

So which is it? No need to be ashamed of a brain fart. We just need to know your original design intention .. so we can assess how the design meets these intentions.

And your claims are here:

https://groupdiy.com/threads/comparison-of-jfets-for-mic-applications.86559/page-13#post-1184093

Thanks for this Thor. There are other claims you made in other posts but I'm sure we'll find them in the fullness of time.

Yes, we calculated, tested this assertion in the simulator and agreed it has lower noise. As for the rest, just look at the results osted earlier.

Thanks for this.

.. "main Blumlein pair in the Royal Albert Hall for the Proms" and are there accessible recording.

I know things have going downhill in the UK, but it's a real shame that the BBC had to sell off all the DPA, Schoeps, Soundfield Microphones in their Inventory

Please don't bring the BBC into this. My Jurassic history with the BBC and my mikes in the Albert Hall is long, involved and best in another thread. But I would be accused of Wi**y W*nking by certain people

Please repeat the exercise with SimpleP48RCA as you proposed.

I leave that to the reader. Simulation files and models etc are for downloading.

Are your reneging on your proposal?

That's not unexpected in view of your many backflips. We just need to know.

 

[ricardo]

View attachment 144227

Quite a bit of improvement. Within 3...4dB of the flat band noise and SNR of the Guru Supa Simple. I can live with that VERY well.

HD at 35mV out (94dB SPL), 0.006% all H2. Let's try 20dB more:
....
Ok, 20dB more, still clipping: (at 134dB spl )

Overall this modified circuit simulates similar in results to the original, but is is better in some respects (7dB less wideband noise, 1/3rd of the HD at 134dB SPL input).
...
PS, Richard, Muntz that! Not allowed is significant unequal current draw on the Phantom Power leg's, material increases of HD (say more than 6db per harmonic), reduced SPL handling (say 3dB).

Congrats Thor! You FINALLY spotted your EVIL 15k R3, 4 that led me to predict >10dB noise over Zephyr's above 4kHz

You WIN A FREE DIVE TRIP TO THE GREAT BARRIER REEF FROM COOKTOWN. Please claim within 12 mths as I'm not sure I can still do this as I'm getting on in years

Not sure why you only get 7dB noise improvement but I'll try to do better.

Alas, I don't think further improvements are possible with simple, even SupaMuntzing

You've even spotted why I wanted to keep Zephyr's current consumption over 5mA.

I make Zephyr's original circuit somewhat better than your original at high spl and my expected >10dB better noise above 4kHz.

I'll post my poor sims showing this and also SimpleP48RCA

But your NeoSupa circuit raises the bar on high spl so much that I can only grovel at your feet.

I may post another version of Zephyr's to compete with NeoSupa but this would only be a poor attempt to make it sound as good. Couldn't possibly measure as good.

 

[thor.zmt]

If we can clarify what you claimed in this #231 post. 134dB spl <0.5% THD (H3 dominates) with -34dB/Pa capsule.

OPERATED AT 35V Bias instead of 60V, which reduced it real sensitivity to -39dB!

Please don't bring the BBC into this.

You did. Nothing to do with me. I don't name drop all sorts if companies, people and institiutions to make myself important. What I do can stand by itself.

Are your reneging on your proposal?

Tempus Fugit, the beach calls, maybe later. I didn't propose. I refused to do it, but stated it could be added.

As it is changing one connection on the other design, it is trivial to run. I just did. It gives 0.082% THD at 94dB, ~ 0.65% at 114dB, better but still pretty poor, plus very unfavourable harmonic spectrum. Weighted by DEL Shorter's metric, it would likely be worse than the original.

Even the cheaper Chinese sourced Mic's usually do better than this. See the earlier test results in this thread which used a common "China Mike" circuit.

Maybe the 4.99 USD Karaoke "Professional Microphone" meant for 5V supply and SE input into the PC etc Mic inputs in SE do as badly, maybe worse. I wouldn't know.


Noise is identical in the flatband to the original, but 3dB more at LF.



Well, you insisted.

Thor

 

[thor.zmt]

Congrats Thor! You FINALLY spotted your EVIL 15k R3, 4 that led me to predict >10dB noise over Zephyr's above 4kHz

I already stated this in my noise analysis. MANY, MANY posts back.

My target was something like this spec sheet:



And I think I did beat it at a very low budget including a decent 34mm Capsule.

So my very simple Schoeps derived circuit was QUIET ENOUGH for my product definition.

I was aware I could lower noise by lowering resistors, but everything is a trade-off, what came out was "good enough for the job" so I let it be. It's called pragmatism.

You WIN A FREE DIVE TRIP TO THE GREAT BARRIER REEF FROM COOKTOWN. Please claim within 12 mths as I'm not sure I can still do this as I'm getting on in years

ThanX, but I prefer Thailand.

Not sure why you only get 7dB noise improvement but I'll try to do better.

The extra stage has similar noise to the Input J-Fet, so the compound has 3dB more noise than just the J-Fet alone. I would have thought this as obvious.

Alas, I don't think further improvements are possible with simple, even SupaMuntzing

I think there are, just not with something derived via BM-X00 PCB. And that was one of my limitations set. The other was originally that it had to be done on a rainy Guangdong Sunday afternoon for a first prototype.

You've even spotted why I wanted to keep Zephyr's current consumption over 5mA.

I did not need to "SPOT IT".

I KNOW IT.

As I said many times before, all designs are set of chosen compromises.

I compromised on some factors in favour of others. It's called adapting a design to requirements.

Of course, for that one needs to understand and define requirements.

For my requirement the massive noise with 15k X 2 (10k in my real microphones as I used 0.5mA Idss FET's) was still low enough, accounting for environment, Microphone Preamp and ADC.

Not everyone can afford 16 Channels of 130dB Dynamic range ADC's...

I make Zephyr's original circuit somewhat better than your original at high spl and my expected >10dB better noise above 4kHz.

You do that by using the original circuit, TL431 and all or changing it?

But your NeoSupa circuit raises the bar on high spl so much that I can only grovel at your feet.

Richard, as I said, the circuit is not Supa, just part of a dirty coyote ugly hack of a BM-800 Microphone into a decent (very decent) recording microphone.

Even the update was done in minimum time. Nothing is optimised. I just threw it on and made sure it biased at 5V across the Drain Resistor of M1. Then ran the Sim's.

I'm sure a few more hours work adjusting each operating point just so can get a bit more, but that's not my point.

Thor

 

[ricardo]

HD at 35mV out (94dB SPL), 0.006% all H2. Let's try 20dB more:

At 353mV out (114dB SPL) is 0.065% THD and higher order harmonics show up.

Ok, 20dB more, still clipping:

But we are at < 5% THD now:

I'm a bit confused by all this.

I make 134dB spl with our -35dBV/Pa capsule at 5dBV, 1.78V rms ie 2.49Vp

Could you confirm this Input Voltage please. Much easier to specify INPUT to da various circuits (for our desired spl) and see what the sims spit out for output.

 

[thor.zmt]

I'm a bit confused by all this.

I make 134dB spl with our -35dBV/Pa capsule at 5dBV, 1.78V rms ie 2.49Vp

Could you confirm this Input Voltage please. Much easier to specify INPUT to da various circuits (for our desired spl) and see what the sims spit out for output.

I set all circuits to 50mV P-P (25mV Peak, as TINA specifies AC voltages on generators as peak, not RMS as peak is waveform independent) or 17.67mV RMS input for 94dB/1PA.

This is in the downloadable files. You did download and check them, right?

I then set the output with this input to 35.35mV RMS using the VCVS to scale. That is 50mV Peak or 100mV P-P or a net cir unit gain on 2 / 6.04dB.

So all circuits can be compared in the schematics file side by side.

Thus 134dB SPL = V(94dB) + 40dB or 17.67mV * 100 = 1767mV RMS = 2499mV Peak = 5,000mV P-P input to the circuit being evaluated.

In practice set input to 25mV Peak (-29dBV) for 94dB SPL in, 250mV Peak (-9dBV) for 114dB SPL in, 2500mV Peak (11dBV) for 134dB SPL in.

Very easy. I must apologiess for using all sorts of different units in context for what are the same signals, I use peak, P-P and RMS (and dBV/dBu) completely interchangeable, as they are all the same.

I should probably be more consistent in unit use, to be more considerate of others.

Thor

 

[ricardo]

I'm a bit confused by all this.

I make 134dB spl with our -35dBV/Pa capsule at 5dBV, 1.78V rms ie 2.49Vp

Could you confirm this Input Voltage please.

Thus 134dB SPL = V(94dB) + 40dB or 17.67mV * 100 = 1767mV RMS = 2499mV Peak = 5,000mV P-P input to the circuit being evaluated.

In practice set input to 25mV Peak (-29dBV) for 94dB SPL in, 250mV Peak (-9dBV) for 114dB SPL in, 2500mV Peak (11dBV) for 134dB SPL in.

Thanks for this Thor. 5,000mV p-p is close enough to 2.49Vp

 

[Cqwet Dbdfte]

Voltage units are well interchangeable, but can obscure what is happening.
RMS values can be equal but have different crest factors, which for sound signals would be a problem.
RMS may have its origins in legacy meters, and give a sorta useful output.
The true energy value is good for the power company's metering and heating of components.
Peak to peak is what electric circuits experience and what is visible on the o'scope, which would reveal clipping, symmetric or otherwise.
But you knew this already.