Piezo buffer

[mrkva]

Hello everyone!
I am working on a design of a piezo buffer for audio use. By piezo I mean something likes this:

Here are few requirement
- it should be powered via phantom power (48V, 10mA)
- it should have as little noise as possible

So far I have come up with this:


My question is - what do you think can improve the design? It is quite difficult to work with it on breadboard - it catches FM radio even! Plan is to fit in shielded XLR shell.

Thanks for any suggestions and help!
Jonas

[T3sl4co1l]

BF862?  Hell, you're probably getting not only FM radio, but everything from DC to light through that sucker!  Putting capacitance on the gate is, ironically, only making it worse: between the two bypass capacitors, you get something like a Colpitts oscillator at some obscene frequency (500MHz?).

Simplest, most likely remedy: add a series resistor at the gate, say 47 or 100 ohms.  Won't have a damn bit of change on your application (it could probably be up to 100k and still work), but should stop RF in its tracks.

What are you really doing, on a system level?  I don't get the impression piezo transducers are all that weak; is it because you're driving a long transmission line with it?  Do you need voltage gain as well?

Tim

[mrkva]

BF862?  Hell, you're probably getting not only FM radio, but everything from DC to light through that sucker!  Putting capacitance on the gate is, ironically, only making it worse: between the two bypass capacitors, you get something like a Colpitts oscillator at some obscene frequency (500MHz?).

Aha! But, for some reason, the capacitor there seems to help with the FM-catching. I have to mention that the radio seems to be related to the shielding somehow. When I touch the recorder (metallic case - direct connection to the "ground" in that circuit), I can hear little clicks & pops. Or when I get my hand close to the exposed part of the cable on breadboard.

Simplest, most likely remedy: add a series resistor at the gate, say 47 or 100 ohms.  Won't have a damn bit of change on your application (it could probably be up to 100k and still work), but should stop RF in its tracks.

Thanks, will try that! But, won't the resistor insert some noise to the system?

What are you really doing, on a system level?  I don't get the impression piezo transducers are all that weak; is it because you're driving a long transmission line with it?  Do you need voltage gain as well?

From my basic understanding, the problem is impedance and capacitance. When I plug a piezo directly to my recorder, the sound is very "high-passed" and I need to crank the gain way too much for my taste to get proper sound.
My recorded btw (so you have an idea of the input spec): http://www.sounddevices.com/products/702/specs/

Thanks again for the info! Jonas

[T3sl4co1l]

Mind that it will sound "high pass" naturally, because a piezo disc (if that's really what you're using) has a relatively high mechanical resonant frequency that filters (band pass, actually) what it's exposed to.  Correcting this is a rather aggressive and unrewarding task; you're better off with a wideband transducer, for whatever it is you're working with.

Tim

[mrkva]

Mind that it will sound "high pass" naturally, because a piezo disc (if that's really what you're using) has a relatively high mechanical resonant frequency that filters (band pass, actually) what it's exposed to.  Correcting this is a rather aggressive and unrewarding task; you're better off with a wideband transducer, for whatever it is you're working with.

Tim

That is not what I mean - as you say, that is resonant bandpass. But the signal is missing lower part of the spectrum, which is actually there when the piezo is properly buffered.
http://f.cl.ly/items/3v2Z242z3H1b0V2x0s0G/contact_mic_test.mp3

Here is a little recording for you with a variation of the circuit above (actually using J201). You can hear it sounds way more natural then just pure piezo recording. The problem is noise floor though.

[T3sl4co1l]

Right, of course... the element itself is a capacitor, and working against load resistance.  Simple high pass.

I don't see a spec on what that resistance is, unless it's normally very high, to the point where they don't bother to specify; in that case, I guess the phantom power supply resistor would be dominant.

To reduce the noise floor, you need low noise and high gain; low noise requires matched source and load impedances, which doesn't make much sense against a capacitor, so, you're kind of stuck there.  A JFET is the way to go, having an input resistance about as high as possible.  You're also stuck on the output match -- 6.8k or whatever the load is.  The remaining matters are gain, and selecting a device which matches best to its surroundings.

Really, the only thing missing in your circuit, I would guess: reduce or eliminate the source resistor, to maximize gain.

Tim