Comparison of JFETs for mic applications

[thor.zmt]

It seems that 2SK170 is one of the most counterfeited items.

Together with OPA2604 and NE5532 (yeah, fake 5532's exceeded my credulity too, until I had some send to me).

Oh, add WIMA capacitors.

And about any other common item, where a few cent can be made making a cheap fake.

I'll save the war stories from Shenzhen SEG electronic mall and the supercute gorgeous girl I met at a wine and cheese event who's family had a big business selling fake components at SEG (and she was importing and selling Wine as hobby) and what happened after we hit off and hopped on the good foot back to my hotel.

Thor

 

[soulsonic]

Hi @soulsonic !
You have practically tested and you can
confirm that 2sk880 is the smd equivalent of 2sk117?

I won't say that it is an exact equivalent - the tests earlier in this thread should show that it isn't - but I can say that I was able to drop them into existing circuits that used the 2SK117 and the performance was very good for my purposes. But, bear in mind that these were guitar distortion and booster circuits, so I can't comment on their performance in a "clean" circuit like a microphone preamp. There may very well need to be changes to component values to optimize them, depending on the circuit topology, but in the case of the circuits I use them in, I didn't have to.

 

[Voyager10]

Hi @soulsonic !
You have practically tested and you can
confirm that 2sk880 is the smd equivalent of 2sk117?

Looking back at the various data tables & graphs in this thread (posts #33, #38, #49 and #53), the 2SK117-BL I tested comes out very similar to the 2SK209-GR and 2SK880-GR in pretty much all the measurements. (The 2SK880 appears to be electrically identical to the 2SK209, it's just a smaller package). It's as close a substitute as you're likely to get.

 

[soulsonic]

Looking back at the various data tables & graphs in this thread (posts #33, #38, #49 and #53), the 2SK117-BL I tested comes out very similar to the 2SK209-GR and 2SK880-GR in pretty much all the measurements. (The 2SK880 appears to be electrically identical to the 2SK209, it's just a smaller package). It's as close a substitute as you're likely to get.

I wish I had noticed the 2SK209s back when I chose the 880s, so I wouldn't have to struggle so with soldering these tiny things.

 

[MicUlli]

2SK117 (GR or BL) already available in Germany from Reichelt. They claim that it is "Toshiba". 0,95€ per piece.
My favourite is 2SK209 (GR or BL) "Toshiba", 0,49€ per piece. Because it is specified for VDS up to 50V, hardly destroyable by P48 phantom power.
If lower voltage can be guaranteed NSVJ3910 brings about 1dB lower noise.
Happy new year
MicUlli

 

[SoundBlaster]

What’s funny is that I have nearly 1000 2SK147BL’s and 2SJ72BL’s that I find difficult to sell due to skepticism.. I picked them up with a bunch of Hitachi lateral Mosfets back in the early 90’s. At this point I figure they’re like bitcoin. Just keep going up in value…and genuine!

 

[Voyager10]

For the sake of completeness, I've dusted off the "ADK / MXL 2001" circuit from post 33 and tested with all the FETs which have turned up since the original post.

Summary table​

Device	Idss	Vgs @ 0.5mA	Rbias	Gain (1nF)	Gain (68pF)	THD (1nF)	THD (68pF)	Ein (1nF)	Ein (68pF)
BF256B	8.05​	-2.07​	4.14​	6.08​	4.98​	0.071​	0.014​	-120.7​	-118.7​
J305	5.66​	-1.63​	3.26​	5.97​	4.96​	0.017​	0.020​	-120.7​	-118.8​
2SK246-BL	8.42​	-3.02​	6.04​	5.72​	4.38​	0.110​	0.043​	-120.5​	-118.5​
2SK369-V	15.4​	-0.35​	0.7​	6.74​	2.73​	0.046​	0.270​	-121.5​	-118.4​
2SK373-GR	3.4​	-0.85​	1.7​	6.36​	4.63​	0.180​	0.022​	-118.8​	-116.9​
2SK709-V	23​	-1.76​	3.52​	7.03​	5.36​	0.025​	0.130​	-119.0​	-115.2​
2SK3557-6	15.7​	-2.02​	4.04​	7.04​	4.8​	0.018​	0.190​	-121.3​	-114.1​

• The Rbias figure is the calculated value for the VR1 trimmer, to set 0.5mA drain current.
• Gain and THD are measured with a 1KHz, 100mV RMS sine wave.

The 2SK369-V has pretty similar numbers to the 2SK170 I measured earlier - similar value of bias resistor, high gain & linearity with 1nF source capacitance but dropping rapidly with lower source capacitances. Of all the FETs I've looked at this is the closest to a drop-in replacement for the 2SK170, if that's actually a good thing...

I've attached noise plots for these, although (as has been copiously discussed earlier) the intrinsic FET noise isn't very relevant with this circuit; the BF256B, J305, 2SK256 and 2SK369 do as well as any other FET will. The 2SK3557 is a surprise - it measured very well in another circuit (post #90) but has a huge (> 10dB) extra LF noise here. It's begging for a bit more investigation.

 

[thor.zmt]

For the sake of completeness, I've dusted off the "ADK / MXL 2001" circuit from post 33 and tested with all the FETs which have turned up since the original post.

What might be illuminating here is to include some of the FET's parameters, such as Ciss (Gate to Source capacitance) and Crss (Reverse transfer Capacitance Drain to gate) as well transconductance and IF POSSIBLE a proxy for transconductance linearity with signal.

For now with casual glances, it seems that in this circuit the "Key Figure Of Merit" (KFOM) is Crss.

Thor

 

[micolas]

For the sake of completeness, I've dusted off the "ADK / MXL 2001" circuit from post 33 and tested with all the FETs which have turned up since the original post.

Summary table​

Device	Idss	Vgs @ 0.5mA	Rbias	Gain (1nF)	Gain (68pF)	THD (1nF)	THD (68pF)	Ein (1nF)	Ein (68pF)
BF256B	8.05​	-2.07​	4.14​	6.08​	4.98​	0.071​	0.014​	-120.7​	-118.7​
J305	5.66​	-1.63​	3.26​	5.97​	4.96​	0.017​	0.020​	-120.7​	-118.8​
2SK246-BL	8.42​	-3.02​	6.04​	5.72​	4.38​	0.110​	0.043​	-120.5​	-118.5​
2SK369-V	15.4​	-0.35​	0.7​	6.74​	2.73​	0.046​	0.270​	-121.5​	-118.4​
2SK373-GR	3.4​	-0.85​	1.7​	6.36​	4.63​	0.180​	0.022​	-118.8​	-116.9​
2SK709-V	23​	-1.76​	3.52​	7.03​	5.36​	0.025​	0.130​	-119.0​	-115.2​
2SK3557-6	15.7​	-2.02​	4.04​	7.04​	4.8​	0.018​	0.190​	-121.3​	-114.1​

• The Rbias figure is the calculated value for the VR1 trimmer, to set 0.5mA drain current.
• Gain and THD are measured with a 1KHz, 100mV RMS sine wave.

The 2SK369-V has pretty similar numbers to the 2SK170 I measured earlier - similar value of bias resistor, high gain & linearity with 1nF source capacitance but dropping rapidly with lower source capacitances. Of all the FETs I've looked at this is the closest to a drop-in replacement for the 2SK170, if that's actually a good thing...

I've attached noise plots for these, although (as has been copiously discussed earlier) the intrinsic FET noise isn't very relevant with this circuit; the BF256B, J305, 2SK256 and 2SK369 do as well as any other FET will. The 2SK3557 is a surprise - it measured very well in another circuit (post #90) but has a huge (> 10dB) extra LF noise here. It's begging for a bit more investigation.

Hi @Voyager10 !
Happy New Year!
Many thanks for all your hard work, for the generosity with which you share the test results.
I have a side questions: in the circuit in the attached picture, from the post33 ,
You connected a capsule, have you tested it in real life, how does it behave compared to the classic Shoeps topology, which jFET do you prefer here?

 

[thor.zmt]

how does it behave compared to the classic Shoeps topology

Compared to Schoeps it is a rather questionable design. I shall leave it at that, following my Mon's dictum that lest you have something nice to say STFUAGPS.

First, it has significant gain, theoretically ~ 20dB and in practice ~ 15dB, where we normally do not need gain. There is no sensible way to lower the gain significantly.

The circuit contains a number of unnecessary sources of excess noise.

And it needs substantial value output coupling capacitors.

The Oktava/Microtech Gefell style circuit would be an improvement.

The best option for this is to increase the gain maximally (short R3, connect C2 negative pole to Q2 Emitter, use highest transconductance with low Crss J-Fet) and then apply a "magic" 33pF capacitor from the output side of C7 (place R 9 after the capacitor) to the capsule (place directly at the capsule connection), converting the whole circuit into a charge amp with a gain of ~ 2 (6dB) for a 68pF Capsule.

Replacing BC559C with a BSS84 P-MOS would likely improve the circuit further, allowing higher value resistors for R6/7 and a C0G ceramic or Film capacitor for interstage coupling.

A cascode for Q1 AND (say) an E-501 CRD as load would help to maximise gain and linearity more

In terms of making a "neutral" Mic, this would likely be a better choice than even Schoeps with BJT followers.

Schoeps will have more character at higher SPL. Oktava/MTG will likely be inbetween.

As always, capsule dominates the sound and usually the HD.

Thor

 

[micolas]

Compared to Schoeps it is a rather questionable design. I shall leave it at that, following my Mon's dictum that lest you have something nice to say STFUAGPS.

First, it has significant gain, theoretically ~ 20dB and in practice ~ 15dB, where we normally do not need gain. There is no sensible way to lower the gain significantly.

The circuit contains a number of unnecessary sources of excess noise.

And it needs substantial value output coupling capacitors.

The Oktava/Microtech Gefell style circuit would be an improvement.

The best option for this is to increase the gain maximally (short R3, connect C2 negative pole to Q2 Emitter, use highest transconductance with low Crss J-Fet) and then apply a "magic" 33pF capacitor from the output side of C7 (place R 9 after the capacitor) to the capsule (place directly at the capsule connection), converting the whole circuit into a charge amp with a gain of ~ 2 (6dB) for a 68pF Capsule.

Replacing BC559C with a BSS84 P-MOS would likely improve the circuit further, allowing higher value resistors for R6/7 and a C0G ceramic or Film capacitor for interstage coupling.

A cascode for Q1 AND (say) an E-501 CRD as load would help to maximise gain and linearity more

In terms of making a "neutral" Mic, this would likely be a better choice than even Schoeps with BJT followers.

Schoeps will have more character at higher SPL. Oktava/MTG will likely be inbetween.

As always, capsule dominates the sound and usually the HD.

Thor

Hi @thor.zmt !
Happy New Year!
I am really pleasantly surprised for these analyses, considerations, it is obvious that you are very passionate, and pedantic, perfectionist and altruistic.
(the praise that my friend @Emmathom gave you is perfectly justified, you helped him a lot with interesting projects)
(A concrete scheme speaks a thousand words, but that would be the topic of another thread)
We thank you for all the information!

 

[thor.zmt]

(A concrete scheme speaks a thousand words, but that would be the topic of another thread)

Oktava / MTG is discussed here:

https://groupdiy.com/threads/if-only-two-mics.89368/page-11#post-1182101

I forgot which Microphone it was, but I discussed the charge amp in this context.

Here a visualisation of the most basic level of what I wrote:


Changing C2 negative from ground to the Emitter of Q2 creates a "bootstrap" for R4, it becomes a very high virtual value.

Removing R3 removes degeneration, so the full gain of the circuit is available.

The 2SK209 @MicUlli mentioned seems a good compromise. It would have ~ 6mA/V transconductance and with an effective drain load of (say) 200kOhm the circuit (open loop) gain exceeds 60dB.

The 33pF capacitor closes the feedback loop around the entire circuit, including C7 (C8 remains an issue, but connecting C4 (needs higher voltage) to the top of C8 would also pull that inside the feedback loop.

Anyway, gain is set by the ratio between C_in (which simulates the capsule) and the 33p Capacitor.

The closed loop gain approaches C_in / C_fb.

Really quite trivial circuitry, all of which was already an old hat when I was young.

Thor

 

[OneRoomStudio]

Oktava / MTG is discussed here:

https://groupdiy.com/threads/if-only-two-mics.89368/page-11#post-1182101

I forgot which Microphone it was, but I discussed the charge amp in this context.

Here a visualisation of the most basic level of what I wrote:

View attachment 142555
Changing C2 negative from ground to the Emitter of Q2 creates a "bootstrap" for R4, it becomes a very high virtual value.

Removing R3 removes degeneration, so the full gain of the circuit is available.

The 2SK209 @MicUlli mentioned seems a good compromise. It would have ~ 6mA/V transconductance and with an effective drain load of (say) 200kOhm the circuit (open loop) gain exceeds 60dB.

The 33pF capacitor closes the feedback loop around the entire circuit, including C7 (C8 remains an issue, but connecting C4 (needs higher voltage) to the top of C8 would also pull that inside the feedback loop.

Anyway, gain is set by the ratio between C_in (which simulates the capsule) and the 33p Capacitor.

The closed loop gain approaches C_in / C_fb.

Really quite trivial circuitry, all of which was already an old hat when I was young.

Thor

Very interesting circuit! How are you supplying capsule polarization voltage?

 

[micolas]

Oktava / MTG is discussed here:

https://groupdiy.com/threads/if-only-two-mics.89368/page-11#post-1182101

I forgot which Microphone it was, but I discussed the charge amp in this context.

Here a visualisation of the most basic level of what I wrote:

View attachment 142555
Changing C2 negative from ground to the Emitter of Q2 creates a "bootstrap" for R4, it becomes a very high virtual value.

Removing R3 removes degeneration, so the full gain of the circuit is available.

The 2SK209 @MicUlli mentioned seems a good compromise. It would have ~ 6mA/V transconductance and with an effective drain load of (say) 200kOhm the circuit (open loop) gain exceeds 60dB.

The 33pF capacitor closes the feedback loop around the entire circuit, including C7 (C8 remains an issue, but connecting C4 (needs higher voltage) to the top of C8 would also pull that inside the feedback loop.

Anyway, gain is set by the ratio between C_in (which simulates the capsule) and the 33p Capacitor.

The closed loop gain approaches C_in / C_fb.

Really quite trivial circuitry, all of which was already an old hat when I was young.

Thor

Well ,Thor, I'll consider that a nice New Years present!
Many thanks!

 

[JF]

Late to the party ... I have good experience with MMBF4392 J-Fet. Also there is a 2SK879 which should be better fitted as 2SK880.
But also MMBF J202 (201 doesn't fit in many circuits) can be good when used in the right environment.

I never use these input J-Fet with gain. Build an optimised impedance converter and follow with a buffer when high output voltage is needed. The input circuit of a C3000 is pretty good, just add a buffer of your liking. And don't do such an attenuator ...

 

[Voyager10]

Compared to Schoeps it is a rather questionable design.

There are so many odd design choices. Apart from the FET stage's noise & THD problems, in the original circuit:
- the emitter-follower stage is poorly biased, and clips easily.
- there is a feedback network slavishly copied from the U87, which is almost totally ineffective due to the low open-loop gain.
- the polarisation voltage is taken from a point 3V lower than it needs to be.

So why does it turn up in so many mics? Did everybody just Google "condenser mic circuit" and picked the first answer?

 

[k brown]

I remain puzzled as to why 2SK660 has not been included in these FET tests, since it's specifically designed for use in condendser microphones.

 

[Voyager10]

Primarily because the only source seems to be from randos on eBay, at ~$5 each.

 

[k brown]

Primarily because the only source seems to be from randos on eBay, at ~$5 each.

Ali, as well, but you suspect they are fakes?

2SK596 is very similar, and available from Mouser and DigiKey. It's what's found in most BM700/800s.

Also used in some (many?) electret capsules with built-in FET.

 

[ccaudle]

Not many applications use T0-92 now. Shouldn't there be a small smt equivalent available? Do those small capsules with a jfet built in use a bare die? I always assumed they had a SOT package inside, but I never disassembled one to check.