JFET for electret mic follower

[precaud]

I have a mid-70's electret microphone, Sony ECM33P that went dead long ago. I stuffed it away in a box and forgot about it. Just checked it out and it has a bad JFET. This is the old style electret where the JFET is separate from the capsule, see attached schematic.

The original FET is a 2SK39A, made by Sony. I'm guessing the critical specs are the input capacitance and noise, but neither are given. It crosses to a 2N/PN4302 (3pF Cis), which is also unobtanium.

Looking at a list of low-noise, low-capacitance N-JFETs, PN4117A, PN4220A, 2N5457-58-59. I think I have some 5458's here.

I've not spec'ed JFETs before so am looking for some input on the best replacement FET.

[TimFox]

Mouser has some 2N4338s in stock at $12.72 ea. from Interfet.  Check the datasheet to see if suitable:
https://www.mouser.com/ProductDetail/InterFET/2N4338?qs=OxRSArmBDfwtFbxHgqekmw%3D%3D&mgh=1&gclid=EAIaIQobChMI_qTc2tOE9gIV3smUCR0mFwQvEAQYASABEgJaavD_BwE

[precaud]

Very interesting, thanks. Noise looks great. I missed that one because I set my scan parameters for 6pF Cis max, and that one is 7. I'm unsure if there's room for a TO-18 in there.

[SeanB]

SMD version would fit no problem though, and would very likely work even better there.

[Kleinstein]

If SMD is OK, there quite a lot of moden low noise JFETs to choose from (e.g. Sk3557). In the electret microphone case the tricky part is getting the DC level at the input right, as the main paths for DC are the JFETs gate leakage and the sensor leakage. Chances are one would need a JFET with a rather low threshold. There are also a few fets specially made for electret microphone capsules.

The extra capacitor that is there to switch in suggests that the input capacitance may not be that critical. In the source follower it would be more Crss and not Ciss that matters.

[TimFox]

The switchable capacitor plus the source-follower's input capacitance (approximately C_DG) form a capacitive load on the electret, whose output impedance is also capacitive.  Thus, the switched extra capacitance decreases the gain, but doesn't affect the low-frequency response (much).  With the large resistor on the source terminal, the current is (sort of) defined by the battery voltage and the resistor, so the gate (should) float to the gate-source voltage required to maintain that current.  The "-8" on the original part number may be a sorting bin on I_DSS to obtain a reasonable gate-source voltage in the circuit.

[precaud]

The switchable capacitor plus the source-follower's input capacitance (approximately C_DG) form a capacitive load on the electret, whose output impedance is also capacitive.  Thus, the switched extra capacitance decreases the gain, but doesn't affect the low-frequency response (much).

Yes. I assumed the FET Cis = 3pF, and 33 + 3pF is an 8dB divider, then the capsule would be roughly 54pF.

[quote  With the large resistor on the source terminal, the current is (sort of) defined by the battery voltage and the resistor, so the gate (should) float to the gate-source voltage required to maintain that current.  The "-8" on the original part number may be a sorting bin on I_DSS to obtain a reasonable gate-source voltage in the circuit.[/quote]

They would be sorting for low Idss, then?

I could probably make SMD fit.

[TimFox]

If it were my design, I would try to run the device at a drain current that gave roughly half the supply voltage across each of the drain-source and the source resistor (your diagram doesn't show voltages).  Therefore, one wants I_DSS less than that value (but not a whole lot less).

[Conrad Hoffman]

I hadn't looked at fets in a while and was surprised at how few there are, how expensive they are and how long the lead times are. I'd be tempted to get a couple really cheap electrets with built-in fets and steal the fets out of them.

[precaud]

I'd be tempted to get a couple really cheap electrets with built-in fets and steal the fets out of them.

Clever idea! I may have one here.

If it were my design, I would try to run the device at a drain current that gave roughly half the supply voltage across each of the drain-source and the source resistor (your diagram doesn't show voltages).  Therefore, one wants IDSS less than that value (but not a whole lot less).

That makes sense. I was looking at PN4302 specs, it does appear to be the one to get. And I discovered that Tektronix used one in the 7B53A timebase. I may have a junker I could steal it from, will look manana.

[Uunoctium]

tl;dr
how to adjust the source resistor after FET replacement for minimum THD and optimum gain?
Neumann marked this R always with an * = factory selected.
Awesome tutorial about:


2SK117GR 

[Terry Bites]

EEVblog #609 - Condenser Microphone Design Tutorial

[precaud]

Awesome tutorial about:

2SK117GR 

Excellent video, thx for posting. I wish this mic had as much room on the pcb to work with as his mic had...

[precaud]

Well I'm in luck. I have a Tek 7B53A timebase with bad mechanical damage. Q596 on the main pcb is a 2N4302, and it tests good. Mic preamp surgery scheduled for this weekend.

Interesting, it's easy to fool the M328 transistor tester into seeing it as a BJT. Just connect lead 2 (center) to the drain. Is this a known idiosyncrasy of these testers?

[precaud]

Sucesss. It works.

I didn't mention this before, but I bought a pair of these ECM33P's decades ago knowing that one was bad. The working one was modified by the prior owner; I recall liking it on acoustic guitar. He used a 2SK369 with a different circuit configuration which I haven't sketched out yet. He was tweaking on this one too but apparently gave up. I see why he failed. He forgot to provide a path to ground for the capsule body.

I wanted to restore this one as close to stock as possible to judge if his mod is worthwhile.

I haven't listened to or measured the revived one yet, but on a scope the sensitivity of both is pretty much the same and I see no noise difference between them.

[floobydust]

2SK596 or KSK596 is "the usual" electret mic JFET, with low noise and input capacitance 3.5pF
Oddly enough, it's usually cheaper to take apart an electret capsule and remove the JFET.

If you want high quality, Linear Systems has several JFET remakes too. Most manufacturers are no longer making JFETS.
Although TI has the new Burr-Brown JFE150 but 24pF Ciss and SOT23-5. It looks worth trying out for a condenser mic.

[precaud]

Interesting. Specs on it (KSK596) are very much like the original and the 2N4302.

I can see why alternatives were being explored. They have very high THD. 1.2% best-case. No distortion specs given for the JFE150, either.

[precaud]

EEVblog #609 - Condenser Microphone Design Tutorial

Actually this one is more relevant to electrets:
EEVblog #611 - Electret Microphone Design