Comparison of JFETs for mic applications

[thor.zmt]

Is that Vgs of M1 in your circuit?

The Threshold voltage of 1.7V is a datasheet value. My BSS84 Model seems to indicate 2V Vgsth.

What voltage do you sim at the Drains of M2 & 3?

Especially for the Guru, a sim with all DC voltages and currents probed and spelled out:



Incidentally, I made up an A-Weight + 5th order 20khz lowpass network (all passive, VCVS & Ohm level resistance's) which suggests an 89dB(A) SNR @ 94dB SPL (or 5dB(A) SPL Self Noise), which is pretty Horrorshow in my books, except capsule noise scotches this big time.

HD @ 94dB SPL & 1kHz at 0.006% (or -83dB of pure H2) is also pretty pukka, innit?

Again, the goal was/is to reuse all of the BM-800 China Mike (except the saddle & Capsule where we go 26mm or 34mm Electret or 32/34mm LDC) including the PCB, mostly use commodity SMD Parts that fit the PCB.

And to make a studio recording Mic that betters relatively affordable (sub 500 USD) China made "recording microphones".

Not to make some supa dupa world record setting something. I leave that to Guru's.

I'm too pragmatic to be a guru.

I will show in a separate the thread the LDC Capsule version, with Bias from P48. It looks....



Thor

 

[Matador]

I for one appreciate the back and forth, even accounting for the snark.

Especially for the Guru, a sim with all DC voltages and currents probed and spelled out:

Why are the independent voltage source(s) added? Is it to monitor branch currents?

I'm curious what would change if you abandoned P48 power, and were free to set the MOSFET currents to whatever "optimum" you felt was best?

 

[thor.zmt]

Why are the independent voltage source(s) added?

Voltage source? Only 48V is source. You mean the Schematic symbol for a Meter with a "V"?

These are virtual volt meters. This is what I like about TINA, I can throw on Meters and probes galore to my desire.

I'm curious what would change if you abandoned P48 power, and were free to set the MOSFET currents to whatever "optimum" you felt was best?

Ok, take a step back.

The point of this design is to reuse existing PCB's and plug into a mixing desk or ADC/Pre with P48.

Once we abandon these limitations, we have other options.

A lot then will depend on the design goals again.

Say we want a "colour Mic", use a 6AK5 without Gigaohm resistors and with a decent but small output transformer (Neutrik has one). There is a thread on such a DIY Mic. It will saturate early, may need a switchable C-Pad, but it will generally give "good sound".

As alternative for "super low noise, super low distortion, super high SPL" (let's set the goal to beat anything else with a similar capsule on all metrics), perhaps a 2SK2145 paralleled plus Op-Amp as charge amp with flyback diode for bias and +/-18V rails, with enough gain to go directly to a 10k line level input on the AD. Say 146dB = +20dBu out and < 4dB(A) SPL self noise. Who has a 146dB Dynamic ADC in their pocket?

Or just go fully digital with a large (say 64pcs) array of PDM output MEMS mic elements and a suitable DSP, direct output to Wifi/USB, nothing analog whatsoever. DSP for directivity, frequency response and emulation of classic Microphones (distortion/noise etc. vs level vs frequency on top of directivity vs frequency).

In 2025 we can make almost anything, but still one person cannot make everything.

Thor

 

[Cqwet Dbdfte]

These are virtual volt meters. This is what I like about TINA, I can throw on Meters and probes galore to my desire.

LTspice is very intuitive, used it over 20 years, met Mike Engelhart 2004, now with QSPICE, since he left Analog Devices.
You can just click on a "wire" and it will plot its voltage, customize keybinds to match PCB layout etc.
There is a very large usergroup for LTspice.
I tried TINA, its like another commercial product. Did not find any redeeming qualities.At least it is free.
Mike Engelhart was part of the original group at Berkeley that invented SPICE.
Though beware of Voltage sources, they default to Zero impedance, mathematically correct, but unobtanium. Give them a realistic value for useful circuit operation.

 

[thor.zmt]

LTspice is very intuitive

Coming from PSPICE I disagree. Let's leave it there.

You can just click on a "wire" and it will plot its voltage

TINA does that too, or click on a component and it tells you Voltage and Current and so on.

But with TINA I can just show all voltages I want on a screenshot in a very easy way, the way LTS doesn't work.

Though beware of Voltage sources, they default to Zero impedance, mathematically correct, but unobtanium. Give them a realistic value for useful circuit operation.

Unless of course you want the "ideal" state for your simulation.

Thor

 

[Cqwet Dbdfte]

Unless of course you want the "ideal" state for your simulation.

Nahh... I don't need the "ideal state", it is best left on the chalkboard.

 

[Matador]

perhaps a 2SK2145 paralleled plus Op-Amp as charge amp with flyback diode

What is "flyback" in this context? It is using a diode as the gate-bias element?

 

[thor.zmt]

What is "flyback" in this context? It is using a diode as the gate-bias element?

Yes.

It is essentially "fly-back" in that clamps any excursion not clamped by the J-Fet's inherent diode.

We could get clever with bootstraping this and all sorts of jiggery pokery.

Thor

 

[Voyager10]

FETs for 'ECM' applications​

I got curious about the 2SK660 and related devices, so for your entertainment here's some measurements for:

• 2SK660 from a random AliExpress seller (literally called "Shop5789782")
• 2SK596 from an equally random eBay seller
• OnSemi MMBFJ201 (surface-mount version of the J201, 15p each from CPC)

These FETs all have low Idss values (< 1mA) which means they're expected to operate at Vgs ~ 0V, as determined by the bias network (diode / high value resistor) between gate and source, at least on the 2SK660 and 2SK596.

The J201 doesn't have this built-in, but it can easily be adapted by wiring half of a BAV199 diode across it (anode to source, cathode to gate). I did this using a double-sided SOT23 package adapter, with the FET on one side and the diode on the other, with a track break to disconnect the other BAV199 diode from the drain terminal.

It looks like this:

Common-source amplifier​

First up is a common-source amplifier, which is the circuit for which they give measurements, in the '660 and '596 datasheets.



R1, C1 and Q1 are all socketed. The VDD was heavily filtered (schematic attached at the bottom, not very exciting) to allow a bench power supply to be used, and the output went via my White follower buffer to drive the soundcard, so the output is essentially unloaded.

Circuit gain​

Gain is affected by the FET and the R1 and C1 values. Who knew?

Here are the plots:

• With R1 as 1K, 2K2, 3K3 and 6K8, and C1 fixed at 68pF

• With R1 fixed at 1K, and 10pF, 22pF, 47pF, 68pF and 100pF for C1
  

Noise​

This is the equivalent input noise, measured using REW (i.e. feed a known signal level into the circuit input, then calibrate REW with that signal level; I'm using a 5mV RMS level so my audio interface can work at max gain).

• C1 = 68pF, R1 = 1K (for comparison against other FETs & circuits in this thread)

• C1 = 10pF, R1 = 1K (e.g. small electret capsule)

There is some mains pickup at 100Hz here, it's in the wiring rather than from the PSU, but it doesn't appear to affect the numbers too much. The J201 also appears to have a lot of LF noise, but subjectively it isn't particularly noticeable.

Distortion​

This is where it gets a little ugly. The input signal voltage appears across the input diodes, and as soon as they begin to conduct there is a lot of distortion.

Summary table​

	2SK660	2SK596	MMBFJ201
Idss (mA)	0.20​	0.21​	0.50​
Gain (1K / 68pF)	0.89​	0.90​	1.35​
Gain (1K / 10pF)	0.57​	0.60​	0.72​
Ein, 68pF (dBV, A-weighted)	-120.8​	-120.6​	-124.6​
Ein, 10pF (dBV, A-weighted)	-109.9​	-108.9​	-112.8​
Vin max (THD < 0.5%, mV RMS)	7​	7​	14​

Observations​

• Input capacitance of the devices appears to be commendably low (as shown by relatively flat gain-vs-Cin curve), particularly 2SK596 and 2SK660
• There's very little observed difference between 2SK660 and 2SK596 in most parameters
• The measured values for gain, noise and THD are consistent with the original 2SK660 and 2SK596 datasheets, so perhaps I did get genuine devices after all.
• THD levels for the common-source circuit are significantly higher than other impedance-converter circuits in this thread.

 

[micolas]

FETs for 'ECM' applications​

I got curious about the 2SK660 and related devices, so for your entertainment here's some measurements for:

• 2SK660 from a random AliExpress seller (literally called "Shop5789782")
• 2SK596 from an equally random eBay seller
• OnSemi MMBFJ201 (surface-mount version of the J201, 15p each from CPC)

These FETs all have low Idss values (< 1mA) which means they're expected to operate at Vgs ~ 0V, as determined by the bias network (diode / high value resistor) between gate and source, at least on the 2SK660 and 2SK596.

The J201 doesn't have this built-in, but it can easily be adapted by wiring half of a BAV199 diode across it (anode to source, cathode to gate). I did this using a double-sided SOT23 package adapter, with the FET on one side and the diode on the other, with a track break to disconnect the other BAV199 diode from the drain terminal.

It looks like this:

View attachment 145569View attachment 145570

Common-source amplifier​

First up is a common-source amplifier, which is the circuit for which they give measurements, in the '660 and '596 datasheets.


View attachment 145571
R1, C1 and Q1 are all socketed. The VDD was heavily filtered (schematic attached at the bottom, not very exciting) to allow a bench power supply to be used, and the output went via my White follower buffer to drive the soundcard, so the output is essentially unloaded.

Circuit gain​

Gain is affected by the FET and the R1 and C1 values. Who knew?

Here are the plots:

• With R1 as 1K, 2K2, 3K3 and 6K8, and C1 fixed at 68pF

View attachment 145573

• With R1 fixed at 1K, and 10pF, 22pF, 47pF, 68pF and 100pF for C1
  View attachment 145574

Noise​

This is the equivalent input noise, measured using REW (i.e. feed a known signal level into the circuit input, then calibrate REW with that signal level; I'm using a 5mV RMS level so my audio interface can work at max gain).

• C1 = 68pF, R1 = 1K (for comparison against other FETs & circuits in this thread)

View attachment 145575

• C1 = 10pF, R1 = 1K (e.g. small electret capsule)

View attachment 145577

There is some mains pickup at 100Hz here, it's in the wiring rather than from the PSU, but it doesn't appear to affect the numbers too much. The J201 also appears to have a lot of LF noise, but subjectively it isn't particularly noticeable.

Distortion​

This is where it gets a little ugly. The input signal voltage appears across the input diodes, and as soon as they begin to conduct there is a lot of distortion.

View attachment 145579

Summary table​

	2SK660	2SK596	MMBFJ201
Idss (mA)	0.20​	0.21​	0.50​
Gain (1K / 68pF)	0.89​	0.90​	1.35​
Gain (1K / 10pF)	0.57​	0.60​	0.72​
Ein, 68pF (dBV, A-weighted)	-120.8​	-120.6​	-124.6​
Ein, 10pF (dBV, A-weighted)	-109.9​	-108.9​	-112.8​
Vin max (THD < 0.5%, mV RMS)	7​	7​	14​

Observations​

• Input capacitance of the devices appears to be commendably low (as shown by relatively flat gain-vs-Cin curve), particularly 2SK596 and 2SK660
• There's very little observed difference between 2SK660 and 2SK596 in most parameters
• The measured values for gain, noise and THD are consistent with the original 2SK660 and 2SK596 datasheets, so perhaps I did get genuine devices after all.
• THD levels for the common-source circuit are significantly higher than other impedance-converter circuits in this thread.

Wow, so Ali 2SK660 from seller Shop5789782 is usable
I bought Ali 2SK30 and they are good, similar to the original ones
Thank you

https://fr.aliexpress.com/item/1005001937860978.html