2SK660 in ext polarized mic?

[k brown]

If you don't want to believe datasheets and what can be labelled as common knowledge I can recommend you to do a search for "HP4070D", invest about $ 20,--, buy such an Multimeter and you will be able to measure resistances up to 2 GOhm.

Why are you saying I don't believe data sheets - the input resistance isn't in 2SK660's data sheet. Is there some reason to assume the '660 and the '596 are the same?

 

[analogguru]

Is there a good reason to believe that they are significantly different ? (Hope and wishful thinking are not a good reason)

 

[thor.zmt]

Not capacitance but maybe input resistance of only about 30MOhm...

How do you get "30M input resistance"?

Not from the datasheet and not from the 2SK596 Datasheet.

Be careful to not infer that the datasheet equivalent schematic is real or that there is a real resistor.


It reads "equivalent circuit". The resistor is actually the parasitic current in the second diode (counting L->R) while the first diode is actually the J-Fet Junction.

So the precise internal schematic is this:



Or to infer indeed that the -1dB worst case 110Hz attenuation stated with 15pF source impedance is indicative of the input impedance (if it was, it would be ~ 500MOhm).

A lot of the Datasheet values are "limit values" that are guaranteed across all operating condition (unless stated differently in the small print). Like at elevated temperatures where electron mobility increases and parasitic diode currents increase or with very large Vds which causes increased leakage currents.

There is a reverse biased diode on the input (and the diode of J-Fet Junction). Intention is to operate the J-Fet at Idss with zero V Vgs.

These diodes provide a non-linear resistance, which up to ~ +/-0.4V is near infinity (depends on electron mobility to sure exactly how low the parasitic currents will be).



Precise measures are hard to come by, but at normal signal levels it is extremely high.
As we have a guaranteed limit of ~500MOhm for the 2SK596, we can assume "typically GOhm" fairly safely.

As the circuit starts "limiting" at ~ 0.4V Peak (or 28dB above 10mV which is ~ 94dB SPL output on most larger capsules and thus at 122dB absolute SPL) it may be desirable to "bootstrap" the input.

At the same time, the non-linear input impedance above 122dB may be desirable and if not we can always add a capacitive pad if we need higher SPL's.

And of course the J-Fet can be used with non-electret capsules, but in circuit the reverse biased diode must be accounted for.

Generally "Non ECM" J-Fets lack this diode. So the "ECM" J-Fets cannot directly drop into circuits designed for normals J-Fet's.

Thor

 

[analogguru]

How do you get "30M input resistance"?

Not from the datasheet and not from the 2SK596 Datasheet.

Did you really have a look at the top of page 2 of the 2SK596 datasheet I posted ?
There you can find it:

Input resistance (Zin @ f 1kHz): min. 25 MOhm

Attached you will find another datasheet from a similar FET.

 

[thor.zmt]

Did you really have a look at the top of page 2 of the 2SK596 datasheet I posted ?
There you can find it:

Input resistance (Zin @ f 1kHz): min. 25 MOhm

Note it states IMPEDANCE and a FREQUENCY. And NOT Resistance.

The input impedance of a grounded source J-Fet is in effect the input resistance (say > 1GOhm) in parallel with the J-Fet input capacitance (Cgs aka Ciss) and miller amplified reverse transfer capacitance.

The 2SK596, assuming standard J-Fet Geometry has 3.5pF Cgs & 3.5pF Cgd. With a gain of -3dB (0.7) Cgd will be ~ 6pF.

Now what is the IMPEDANCE of 6pF @ 1kHz?

And what is the IMPEDANCE of 6pF @ 10Hz?

Thor

 

[analogguru]

Note it states IMPEDANCE and a FREQUENCY. And NOT Resistance.

IMHO I can read in both datasheets RESISTANCE and I can spot a resistor symbol, not a capacitor.... especiallly in the second datasheet of the TF202 where only one diode is shown like in your post above.

You will not find these components in a datasheet of a "normal" JFET.... where your calculation from above should apply too....

But I don't really care.... Why should I think about using such an el-cheapo-FET as long as I can use a decent FET ?

 

[ccaudle]

as long as I can use a decent FET

A "decent" FET will still have gate-to-source and gate-to-drain capacitance, that is just intrinsic to how devices are constructed (any physical device, even tubes have intrinsic capacitance between the terminals). I think you missed the point of what Thor was trying to point out to you.

 

[analogguru]

A "decent" FET will still have gate-to-source and gate-to-drain capacitance, that is just intrinsic to how devices are constructed (any physical device, even tubes have intrinsic capacitance between the terminals). I think you missed the point of what Thor was trying to point out to you.

No I didn't.... If you would have read my post more carefully, you would have been able to recognize that I even reffered to it:

.... a "normal" JFET.... where your calculation from above should apply too....

And for your personally a clarification:
When I wrote about "decent" FETs I meant: ...as long as I have enough 2SK30A ((> 1000), 2SK117, 2SK118, 2SK170, 2SK246 and a lot of others in my storage I do not intend to think about using a 2SK596, 2SK660 or 2SK1109 for a "true" capacitor microphone...

BTW, did you notice that these ECM-devices have a Vds and a Vgdo of only 20V ?

And now I am out of this time stealing discussion.

 

[ccaudle]

as long as I have enough 2SK30A

A 2SK30A JFET is specified as having 8.2pF input capacitance, so at 1kHz it will only have 19M Ohm input impedance as well. I think that is the point you missed, the 30M Ohm spec did not refer to the effective resistance of the 2SK660 (which Thor pointed out is not actually a discrete resistor, but an effective resistance due to the leakage current across the device and would be near 1GOhm at room temp). The 30M Ohm was defined very specifically at 1kHz, and at 1kHz the effective input impedance of your "decent" JFET is even lower. It is a perfectly fine device, but having a higher input impedance than a 2SK660 is certainly not the reason, since it does not per the datasheet values.

 

[thor.zmt]

IMHO I can read in both datasheets RESISTANCE and I can spot a resistor symbol, not a capacitor.... especiallly in the second datasheet of the TF202 where only one diode is shown like in your post above.

Resistance with frequency is IMPEDANCE, as it is AC. I suspect that the "resistance" is actually a poor translation from japanese to engrish.

You will not find these components in a datasheet of a "normal" JFET.... where your calculation from above should apply too....

They do apply.

Let's take a 2SK170GR with a "design center" 4.55mA Idss, grounded source, 1k drain load and 9V supply. This will work with 6.4V to 2.5V across the FET.

Gain will be ~ 31.6 (30dB).

For capacitances Ciss ~ 30pF & CRSS ~ 10pF and Crss is of course Miller amplified to 316pF, for a total effective Cin of 346pF which gives a Zin/Rin of 460kOhm at 1kHz.

But I don't really care.... Why should I think about using such an el-cheapo-FET as long as I can use a decent FET ?

Depends. I would use it because:

1) I can do away with the high value resistor, without any real penalty
2) Low input capacitance provides less loading for the Capsule compared to most other J-Fets
3) The Diodes on the input provide soft and frequency dependent limiting at very high input SPL
4) The Idss is ideal for use in a Schoeps circuit without needing to adjust bias
5) Used common source with a follower and output transformer we get a very nice, tube like sound without tube

You are of course welcome to use an expensive FET that is less well suited to the specific application.

This may be worth reviewing:

https://www.edn.com/condenser-microphone-uses-dc-coupled-impedance-converter/

Look at noise levels and consider where all the noise comes from? Not the FET itself.

Thor

 

[MicUlli]

These JFETs with builtin diode(s) are very well suited for capsule capacitances around 15pF where the current noise dominates. If these devices are used with higher capsule capacitances around 50..80pF their voltage noise ruins the noise figure dramatically.
BR MicUlli

 

[thor.zmt]

These JFETs with builtin diode(s) are very well suited for capsule capacitances around 15pF where the current noise dominates. If these devices are used with higher capsule capacitances around 50..80pF their voltage noise ruins the noise figure dramatically.

Hmm really? These FET's have a Gm of around 1mA/V and thus roughly self noise equivalent to a 1k Resistor. That would be around 0.5uV giving 86dB SNR Just for the J-Fet.

I'm not sure what noise figure you are targeting. At 10pF input these J-Fets as system show around 4uV noise.

With a 60pF capsule I'd expect noise in the region of 0.7uV. Signal ~ 10mV @ 1Pa.

Apply rootsumsquare and we get overall noise at 0.86uV. even at 1uV total noise I refuse to loose sleep as it would be 80dB below 10mV or 14dB absolute.

My Choeps (China Shoeps) with 2SK660 (NOS) and P-MOS outputs runs the capsule at ~40V bias directly from phantom power (via RC filtering naturally). So we get ~6.7mV/Pa and 3dB worse noise.

Even there I do not feel that noise is high compared to commercial "budget'ish" (sub 1k) china made "recording microphones".

These Mic's are super simple to make (buy a BM-800, restuff the existing PCB, all SMT, all from Mouser except NOS 2SK660, fit a 797 made 32mm capsule, done).

If someone made Mic's like that in China, they could sell under 200 USD and smoke A LOT of commercial mic's. I remember cost per mic, buying parts from Taobao came a few 100 RMB each.

These could be made by any number of cheap microphone assemblers in Guangdong province, like in Enping for a song.

Thor