opamp-based mic circuit idea

[dfuruta]

Saving the details* for later, does something like this look halfway workable inside a condenser mic?

I thought putting a jfet in front of the opamp might have lower distortion than going directly into the opamp + input due to the extremely high source impedance from the capsule;  perhaps it's not a big enough deal to justify the extra difficulty with biasing & feedback.

It's obviously simpler to put a few transistors in there and forget the opamp, but that's no fun, is it?

*EDIT:
by "details" I mean capsule polarization, power supply, blocking caps and output filtering, impedance balancing, etc

 

[abbey road d enfer]

Saving the details* for later, does something like this look halfway workable inside a condenser mic?

That would probably work somewhat, provided you adjust the DC voltages for proper operating point, but what are your expectations? Are you concerned with the intrinsic distortion of a stand-alone FET and want to eliminate it by including it in the NFB loop?

I thought putting a jfet in front of the opamp might have lower distortion than going directly into the opamp + input due to the extremely high source impedance from the capsule;

A FET opamp may work in direct connection to the capsule, but the noise figure would be appalling. A BJT opamp would not really distort; it would overload the capsule resulting in very low signal and poor LF response.

perhaps it's not a big enough deal to justify the extra difficulty with biasing & feedback.

It all depends what advantage you expect from this and what price you attach to it.

It's obviously simpler to put a few transistors in there and forget the opamp, but that's no fun, is it?

Perhaps not; making a discrete circuit working within the constraints of phantom power supply is not a simple task either. This is a problem you'll have to deal with; few opamps offer valuable performance with the meagre current allowance available.

 

[gyraf]

This schematic is supposed to show one of the early KMS-mics: http://www.gyraf.dk/schematics/Neumann_Handheld.GIF

Jakob E.

 

[abbey road d enfer]

This schematic is supposed to show one of the early KMS-mics: http://www.gyraf.dk/schematics/Neumann_Handheld.GIF

Jakob E.

No. The diagram presented by the OP is using an opamp to control the operation of an FET, thus eliminating the FET's distortion, but with the benefit of the ultra-high impedance and low EIN of the FET. Overall it's a VHZ voltage follower.
The Neumann schemo is a completely different paradigm. It's an illustration of the "charge amp", where the opamp gain is governed by the ratio of the NFB cap (C1) to the capsule's own capacitance. This circuit has some advantages such as reducing the influence of stray capacitance but has a major flaw: the linear relationship between displacement and output voltage is not constant anymore. That is the cause for 2nd-order distortion. Some think it's a bonus, some not.

 

[dfuruta]

That would probably work somewhat, provided you adjust the DC voltages for proper operating point, but what are your expectations? Are you concerned with the intrinsic distortion of a stand-alone FET and want to eliminate it by including it in the NFB loop?

Yes, that's the goal.  I'd like to come up with a relatively clean, low-distortion mic amp that I can use in a variety of different settings - I want to experiment with different capsules and bodies, and so I want a circuit that won't add much grunge of its own.  Very low distortion is more important to me than very low noise.

The OPA140 seemed like a possible candidate for the opamp.  Not sure about the fet.  I'm looking at an LT3014 or two for power regulation.  For capsule polarization, it seems like a CD40106 based relaxation oscillator into a Cockcroft-Walton multiplier might work - I don't have it running quite right on the bench, yet, but it seems to an extent that one can trade a high parts-count (more diodes and caps) for a very low operating current (lower voltage in the oscillator).

 

[abbey road d enfer]

The OPA140 seemed like a possible candidate for the opamp.

Seems adequate.

Not sure about the fet.

The constraints are the same as any FET used in conjunction with a condenser mic capsule. Low capacitance and low noise.

 

[gyraf]

(aah - different thing, I see..)

(Jakob E.)

 

[dfuruta]

The constraints are the same as any FET used in conjunction with a condenser mic capsule. Low capacitance and low noise.

Possibly a PN4416?  Looks like Linear Systems is still making those.  It's not so easy to find JFETs, nowadays.  I was also looking at the J310.  Neither has the lowest noise but both have low capacitance.

The 2N3819 seems to have disappeared into thin air.  2SK170s are still available from Linear Systems, but the capacitance is probably higher than I want to play with.

I am (wrongly?) assuming that a through-hole jfet is preferable, so I can wire the high-impedance part of the circuit in the air and not worry too much about schmutz on the pcb screwing up my circuit.

 

[bremusound]

The AKG C460B uses an opamp after the (J305) JFET.

http://www.coutant.org/akgc460b/service.pdf

Maybe that helps.

 

[dfuruta]

The AKG C460B uses an opamp after the (J305) JFET.

http://www.coutant.org/akgc460b/service.pdf

Maybe that helps.

Very interesting, thanks!  I hadn't seen that schematic.

 

[abbey road d enfer]

The AKG C460B uses an opamp after the (J305) JFET.

http://www.coutant.org/akgc460b/service.pdf

Maybe that helps.

Very interesting, thanks!  I hadn't seen that schematic.

That's exactly what you intended. Note that the opamp reference is undisclosed. It looks like a low-current type (could be a TL062...), maybe factory-selected for noise. Anyway the overall noise performance is governed by the capsule's FET (T1) and the bipolar in the S&K filter (T3).
It is interesting to note that the C480, which can be considered the successor of the C460, uses a similar approach with all discrete components.

 

[dfuruta]

Here's my next take, with a few more of the details filled in.

I'm not quite sure how to size C3.  I'm guessing a couple hundred pF is probably ok, but it's not clear to me how to do the math.  I'm also tempted to make the 5.5V reference adjustable with a trim pot, but I'm not sure if that's a good idea for stability over time (or how much it matters).

I am planning to use this circuit with K47/M7 or CK12 type capsules, thus the polarization voltages.

Now I'm going to try to work out the dc-dc converter for polarization.

 

[PRR]

What does R4 do?

 

[usekgb]

The constraints are the same as any FET used in conjunction with a condenser mic capsule. Low capacitance and low noise.

Possibly a PN4416?  Looks like Linear Systems is still making those.  It's not so easy to find JFETs, nowadays.  I was also looking at the J310.  Neither has the lowest noise but both have low capacitance.

The 2N3819 seems to have disappeared into thin air.  2SK170s are still available from Linear Systems, but the capacitance is probably higher than I want to play with.

I am (wrongly?) assuming that a through-hole jfet is preferable, so I can wire the high-impedance part of the circuit in the air and not worry too much about schmutz on the pcb screwing up my circuit.

What about trying the 2SK880?  Supposedly a low noise suggested replacement for the 2SK170.

 

[dfuruta]

What does R4 do?

Adds decorative flair.

 

[dfuruta]

What about trying the 2SK880?  Supposedly a low noise suggested replacement for the 2SK170.

The input capacitance of 2SK880 is 3x higher than PN4416 - noise is great, but I'm not too worried about noise for this time.

I'd also like to wire the high impedance nodes in the air, so through-hole would be nice.  Maybe that's a silly reason...

 

[abbey road d enfer]

Here's my next take, with a few more of the details filled in.

I'm not quite sure how to size C3.  I'm guessing a couple hundred pF is probably ok, but it's not clear to me how to do the math.

I wouldn't know either! Above all I hate writing never-ending equations that lead to a dubitative conclusion. You have to experiment . I made a sim but I don't have the OPA140 so I made do with an OPA2134 and found that about 50pF seems to tame the phase response. But I guess the real circuit with all its stray capacitances and whatnot will give a different optimization. However 47pF is a good start.

I'm also tempted to make the 5.5V reference adjustable with a trim pot, but I'm not sure if that's a good idea for stability over time (or how much it matters).

I would suggest you make it Adjustable Under Test with fixed resistors.

 

[dfuruta]

I wouldn't know either! Above all I hate writing never-ending equations that lead to a dubitative conclusion. You have to experiment . I made a sim but I don't have the OPA140 so I made do with an OPA2134 and found that about 50pF seems to tame the phase response. But I guess the real circuit with all its stray capacitances and whatnot will give a different optimization. However 47pF is a good start.

Yes, I'll probably have to put it together!  I have the circuit running happily on my workbench, with the values changed to work with a TL071 and J201 (don't have the correct parts yet, but I just wanted to make sure it flies in real life).

I would suggest you make it Adjustable Under Test with fixed resistors.

Sounds right.  I'm not sure how much adjustment will be necessary - with a PN4416A it should be guaranteed to bias - but I suppose it would be good to pick the best values for each fet.

 

[usekgb]

What about trying the 2SK880?  Supposedly a low noise suggested replacement for the 2SK170.

The input capacitance of 2SK880 is 3x higher than PN4416 - noise is great, but I'm not too worried about noise for this time.

I'd also like to wire the high impedance nodes in the air, so through-hole would be nice.  Maybe that's a silly reason...

Through hole fet's seem to be going the way of the dinosaurs.  I have had to modify many designs in order to use SMT, where I used to use through hole.  The parts seem excellent, but they are much harder to measure and sort.

 

[dfuruta]

Got a capsule polarization circuit working on the breadboard.  I feel a little silly about the number of components it takes, but I'm getting + and - 60V out from 8V@750μA in.

For now, I'm looking at a relaxation oscillator using a 40106 cmos inverter running around 75kHz -> two more inverters as buffers, each driving an 8-stage cockcroft-walton multiplier made with 1nF caps and 1N4148 diodes.  On the output of each multiplier I've got a 470nF cap for filtering.

Output impedance of the multipliers is high;  to measure I'm using an old Keithley electrometer.  I don't imagine it will be a problem, but I'll try later to hook up a capsule and measure with the (negligible?) load.  I still need to figure out what the ripple looks like - I'll try to do that tomorrow.

A few concerns I have:
75kHz might be too slow for the oscillator;  I don't want it to be audible, obviously.  It's not easy to get it faster without using more power, though.  Right now I've got the timing set with a 1M resistor and a 22pF cap.  These values probably aren't ideal.  I tried a bunch of different values and made some graphs for myself, and those were the best I could find.

My hope is that the amount of power being thrown around is small enough that the oscillator won't get into the audio path.

I need to try the heat gun and see how much the voltage changes with temperature.

This uses a lot of caps and diodes - 32 of each.  The cost is low, but I have to decide if it's a good trade to use the board space to save a mA.  I think I'm probably going to do it.


I'd appreciate any opinions, of course!