Simple head amp for electret capsules

[Voyager10]

Over in my FET measurement thread I've been playing with "electret microphone" FETs - representative of the sort found built in to some capsules. The main difference between these and other condenser-mic JFETs is a built-in bias "resistor" / diode, which means you don't get to set the bias current externally (the circuit has to run with Vgs close to 0V, so the operating current is basically the Idss value).

I think there's room in the world for a circuit somewhere between the "SimpleP48" and "Alice" designs, in terms of complexity. Design goals included (a) low and balanced output impedance and (b) lack of DC offset between 'hot' and 'cold', as well as the usual low noise and THD.

Here's the design:


It's suitable for capsules without a built-in FET (like Transound 2555) or "3-wire" capsules (e.g. Primo EM273) with a built-in FET, where source and drain are accessible separately. I've tested it with 2SK660 and 2SK596 FETs (obsolete, buy from Ali and Ebay, sorry), and a J201 / BAV199 diode combination (see thread linked above), which are still available. (Obviously, 3-wire capsules have their own FET for Q1 above).

Basic theory of operation:

• Q1 is a source-follower with a very high load resistor, to keep drain current nearly constant.
• R1 is chosen according to the Idss of the FET, so that it has about 20V across it.
• Q2 buffers the source follower and feeds the output via C1.
• Collector current for Q2 comes through R2-R5 and LED D1 (this can be any small red or green LED; needs to have ~ 2V forward voltage at say 4mA current)
• D1 and Q2 hold the drain-source voltage of Q1 at about 2.5V, which helps eliminate the effects of drain-gate capacitance.
• C2 and R4/R5 ensure the output impedance on 'hot' and 'cold' is (very close to) balanced.
• Supply current through the R2/R4 leg and R3/R5 leg is (very close to) identical, so no DC offset.
• D2 protects Q2 from damage if the output is shorted whilst C1 is charged.

Here's the prototype on some pad board:




Test signal input is on the left. There's a (socketed) capacitor to model the capsule capacitance, and sockets to allow Q1 and R1 to be selected.

 

[Voyager10]

Measurements​

I've measured noise, gain and THD, for the three sample FETs, and for 68pF and 10pF source capacitances (typical for a LDC and small electret capsule, respectively):

Parameter	2SK660	2SK596	J201/BAV199
R1 value (K)	120​	120​	68​
Measured R1 voltage	23​	22.9​	20.5​
Idss (mA)	0.19​	0.19​	0.30​
Maximum input voltage (0.5% THD, 1KHz), V RMS	4.6​	4.3​	4.7​
Gain, (Cin=68pF, 1KHz, 100mV), dB	-0.23​	-0.16​	-0.28​
Typical THD, (Cin=68pF, 100mV RMS input), %	0.0011​	0.0009​	0.0005​
Equivalent Input noise, (Cin=68pF, A-weighted), dBV	-120.9​	-120.9​	-122.5​
Gain, (Cin=10pF, 1KHz, 100mV), dB	-0.4​	-0.42​	-0.45​
Typical THD, (Cin=10pF, 100mV RMS input), %	0.0011​	0.0018​	0.001​
Equivalent Input noise, (Cin=10pF, A-weighted), dBV	-109.8​	-109​	-114.5​

That's nearly 5V RMS output before actual clipping sets in. My (Focusrite Clarett) input stage has a reasonably high input impedance, so maybe other preamps won't fare as well, but it's still plenty.

Here's a typical distortion spectrum at 1V RMS input (0.015%, according to REW, almost all second-harmonic):



And here are the noise curves:

Footnote​

Hat tip to @MicUlli for his Versatile input stage for SDC mics circuit. This borrows a lot of ideas from that design.

 

[k brown]

I imagine many would like to see some filtering between the 10ks and the FET drain, and some RF protection on the outputs.