OTish Distortion in Conderser microphones

[martin griffith]

Found this, maybe of interest to audio dudes here
http://lavryengineering.com/forum_images/Condenser.pdf


martin

 

[[email protected]]

Found this, maybe of interest to audio dudes here
http://lavryengineering.com/forum_images/Condenser.pdf

I imagine that the effect is real, but there is way of using negative
feedback to get around it.

You've got to be able to drive the microphone sensing membrane as an
electrostatic speaker, and then you can use an AC-excited capacitance
bridge to measure the capacitance of the microphone, and compare the
instantaneous capacitance with the long term average.

Any deviation in measured capacitance then generates an electrostatic
feedback signal to drive the capacitance back to its nominal value.

J.J.Opstelten and N. Warmholtz describe a pressure-seinsor based on
this pronciple in App. Sci. Res Hague, volume B4 from page 329 in 1955,
and again with J.J. Zaalberg van Zelst in volume B6 from page 129 in
1956.

I'm quoting the references from my Ph.D, thesis - which dates back some
25 years before we moved to the Netherlands.

The problem with using this technique in a pressure gauge is that the
electrostatic field required to counteract more than about 0,2% of
atmospheric pressure is enough to induce conduction across the gap.
This shouldn't be problem with a microphone.

Obviously, the electrostatic restoring force/voltage is a linear
function of the presure being counteracted. You've got to use a bridge
excitation frequency a good bit higher than the higest frequency you'd
want to follow - well over 100kHz, ata guess - and you'd want to design
the feedback loop on the basis that you were sampling the capacitance
of the microphone twice per bridge-excitation cycle rather than
continuously.

As a project it would offer a microphone that would be expensive enough
to fascinate the audiophool exploitation industry - pity I can't patent
it.

 

[Genome]

I imagine that the effect is real, but there is way of using negative
feedback to get around it.

This am rubb coz musurment sytem do not account fur brake up ov bitz.

DNA

 

[Glen Walpert]

Found this, maybe of interest to audio dudes here
http://lavryengineering.com/forum_images/Condenser.pdf

The author should have read up on the subject before publishing his
simplistic and unrealistic (10% capacitance variation!) analysis of
1/x error in condenser microphones for two load cases (not two
distortion mechanisms as he claims). A much better analysis can be
found in Acoustical Engineering by Harry F. Olson, copyright 1957, for
example. Other distortion mechanisms related to diaphragm and air gap
dynamics limit diaphragm displacement with low distortion to the
extent that 1/x error is very small. Never nonexistent, however, and
I think his analysis of the effect of load capacitance on 1/x error is
correct even if it is so simplistic as to be virtually useless. Other
than that, I liked it .

 

[Ken Smith]

Found this, maybe of interest to audio dudes here
http://lavryengineering.com/forum_images/Condenser.pdf

Problems:

(1) assuming 100M of input resistance of the amplifier is unreasonable. A
good design will be at least 1G. With bootstrapping, it can be many G.

(2) 30Hz is not below the audio range and 60Hz certainly isn't. The
distortion will largist at the 2nd harmonic.

(3) That is a huge signal you are assuming. What are you trying to
record?

(4) 1pF is not an extreme case. A fairly simple input stage using 2SK170s
will get under 1pF of effective capacitance, if the drop on the source
resistor can be high enough. Values down to 0.25pF can be done with real
components.

(5) You need to review the math by which you come up with the distortion.
The capacitance varies as (1/X) but the voltage goes as (1/C) so the
voltage varies as X for the unloaded case.

(6) You left out all the interesting effects.

(7) Your yellow background is annoying.

 

[Phil Allison]

"Glen Walpert"
"martin griffith"

Found this, maybe of interest to audio dudes here
http://lavryengineering.com/forum_images/Condenser.pdf

The author should have read up on the subject before publishing his
simplistic and unrealistic (10% capacitance variation!)......

** Given that a condenser mic capsule is typically polarised to circa 60
volts DC, a +/- 5% variation in capsule C produces +/- 3 volts peak
variation. This equates to a just over 2 volts rms of signal available at
the gate of the FET used to buffer the capsule to the outside world.

Also, typical studio condenser mics have a rated output voltage of circa 10
mV at 94db SPL, = 1 volt at 134 dB SPL = 2 volts at 140dB SPL. Specified
( pre-amp) overload levels of 150 dB SPL or more are commonplace.

The pre-amp circuitry in the body of such mics is of unity or more often
*less* than unity voltage gain - but huge power gain of course. So, a +/-
5% variation in capsule C is a very reasonable estimate of what happens in
practice with high SPL sound sources.

However, the polarising voltage is fed to the capsule via a resistor of 1
Gohm (sometimes more ) - placing the 3dB down point at about 3 Hz for a
50 pF capsule. Hence, the second harmonic generation alluded to in the URL
can only occur at the very lowest audio frequencies and become at all
significant at extreme SPLs that are most unlikely to exist at those
frequencies from a musical source.

Poking the mic down the port tubes in a 15inch sub woofer going full tilt
not withstanding !



........ Phil

 

[Robert Baer]

Found this, maybe of interest to audio dudes here
http://lavryengineering.com/forum_images/Condenser.pdf


martin

One could feed the capacitive mike into a short circuit (summing
junction of an opamp)...

 

[[email protected]]

This is rubbish because the measurement sytem does not take account of the break up of the bits.

Condensor microphones are small. Admittedly, the instability of the
membrane is a big problem with electrostatic speakers, and part of my
(unsuccessful) job interview at Quad was a discussion of Quad's magic
"conductive"coating on the memebrane in its origianl electrostatic
speakers that could be relied on the give the 100M per square (IIRR -
this is twenty years ago) needed to prevent the membrane getting
small-pox.

I'd be less worried about a microphone membrane, The Dutch pressure
gauges used thin metal membranes, which are a lot tougher than speaker
membranes, if a lot heavier. The weight might conceivably limit the
high frequency response - I haven't crunched the numbers - but a decent
amount of negative feedback would counteract this too,

Bear in mind the intended market - the system doesn't have to work
well, but it must look impressive. A big glowing valve or two
compensates for a lot of technical short-comings.and valves are
reasonably neat devices for driving electrostatic speakers.

 

[Pieter]

Found this, maybe of interest to audio dudes here
http://lavryengineering.com/forum_images/Condenser.pdf


martin

The linearity changes with capacitance change. The capacitance does
not change when the voltage does not change. So to prevent voltage
changes!

To do this, the microphone must be short-circuited (can also be into a
DC offset). The way to do this is feed the signal into the - input of
an opamp that has capacitive feedback. This way this circuit does not
function as voltage amplifier, but as CHARGE amplifier. The charge the
sound delivers is amplified.

I designed an accelerationmeter based on a crystal device (capacitive)
that had a very large frequency range this way.

Another advantage is that the signal is very low-impedant, this means
low disturbances, low noise. And it is not sensitive to the length of
the cable (the cable only sees a short-circuit), and it isn't
sensitive to movement of the cable.

Feel free to ask any questions.

Pieter
email: [email protected] without the NOSPAM

 

[Robert Baer]

The linearity changes with capacitance change. The capacitance does
not change when the voltage does not change. So to prevent voltage
changes!

To do this, the microphone must be short-circuited (can also be into a
DC offset). The way to do this is feed the signal into the - input of
an opamp that has capacitive feedback. This way this circuit does not
function as voltage amplifier, but as CHARGE amplifier. The charge the
sound delivers is amplified.

I designed an accelerationmeter based on a crystal device (capacitive)
that had a very large frequency range this way.

Another advantage is that the signal is very low-impedant, this means
low disturbances, low noise. And it is not sensitive to the length of
the cable (the cable only sees a short-circuit), and it isn't
sensitive to movement of the cable.

Feel free to ask any questions.

Pieter
email: [email protected] without the NOSPAM

Tanks for your eloquent expansion of my suggestion.

 

[Phil Allison]

"Robert Baer"

Tanks for your eloquent expansion of my suggestion.

** Yep - he made a real omelette out of one of your turds.






......... Phil

 

[Ken Smith]

The linearity changes with capacitance change. The capacitance does
not change when the voltage does not change. So to prevent voltage
changes!

This statement is simply incorrect. It is also incorrect in a complex way
too but we'll ignore this for now.

To do this, the microphone must be short-circuited (can also be into a
DC offset). The way to do this is feed the signal into the - input of
an opamp that has capacitive feedback. This way this circuit does not
function as voltage amplifier, but as CHARGE amplifier. The charge the
sound delivers is amplified.

Actually, it is more likely to act as a slightly charge sensitive RF
oscillator with a lot of noise.

Another advantage is that the signal is very low-impedant, this means
low disturbances, low noise.

No, it means very high noise. The input noise voltage of the op-amp
divided by the impedance gives a noise current. Also the lower a resistor
is, the more its noise current. When it comes to noise, you can't win,
you can't break even and you can't get out of the game.

And it is not sensitive to the length of
the cable (the cable only sees a short-circuit), and it isn't
sensitive to movement of the cable.

Try tapping on a coax with a little bias voltage on it connected to such a
circuit. You will find that the circuit is sensitive to moving the cable.
I've had to shock mount cables.

Feel free to ask any questions.

If a train is leaving the station at 11 miles per hour ..........

 

[Pieter]

This statement is simply incorrect. It is also incorrect in a complex way
too but we'll ignore this for now.



Actually, it is more likely to act as a slightly charge sensitive RF
oscillator with a lot of noise.

Not when you do a proper design. The opamp is feedback capacitive,
making it very stable. A small series resistor at the input may be
wise.

Low sensitivity to outside disturbances.

No, it means very high noise. The input noise voltage of the op-amp
divided by the impedance gives a noise current. Also the lower a resistor
is, the more its noise current. When it comes to noise, you can't win,
you can't break even and you can't get out of the game.

There will always be noise.

Try tapping on a coax with a little bias voltage on it connected to such a
circuit. You will find that the circuit is sensitive to moving the cable.
I've had to shock mount cables.

The other circuits are sensitive. The one is decribe here isnt. When
then is a short-circuit, there is no voltage. Where there is no
voltage, that voltage can't change either. But it is best not to use
phantom feeding, but a separate power supply cable. The phantom's
voltage does interfere.

If a train is leaving the station at 11 miles per hour ..........

Two hours and 15 minutes.

Pieter

 

[Ken Smith]

[...]

Actually, it is more likely to act as a slightly charge sensitive RF
oscillator with a lot of noise.

Not when you do a proper design. The opamp is feedback capacitive,
making it very stable. A small series resistor at the input may be
wise.

Putting a feedback capacitor on a high performance op-amp does not make it
"very stable". You have to use a "unity gain stable" one in this sort of
application. Without the small resistor you suggest, the op-amp is fairly
likely to oscillate even if it is a "unity gain stable" one if the input
cable is very long. The cable and capacitor look like a tuned circuit.

Low sensitivity to outside disturbance.

No, this is not the case. The circuit you suggest is this:

Noise
!
---Cnoise
---
! A -------------
Signal -[Zgen]---+-------! Z Amplifier !---- GND
-------------


You can't improve the signal to noise at "A" by making Zamplifier low.
How ever much you lower the amplifiers input impedance, you must also
raise its voltage gain by too to maintain the same signal output.

Just for fun, lets assume that Zgen=Cnoise and the Signal(RMS) is equal
to the noise(RMS).

High Z amplifier case: SNR = 1:1

Low Z amplifier case: SNR = 1:1

The other circuits are sensitive. The one is decribe here isnt. When
then is a short-circuit, there is no voltage. Where there is no
voltage, that voltage can't change either.

When there's no impedance, the current is infnite and thus the noise is
infinite. It was exactly the sort of circuit you are suggesting that
required the shock mounting of cables.

But it is best not to use
phantom feeding, but a separate power supply cable. The phantom's
voltage does interfere.

Try tapping of a length of COAX even with "no voltage" on it. You will
still see spikes when you strike it. Just hook a BNC-BNC cable onto your
scope's input and wack it against the work bench.

 

[Terry Given]

[...]

Actually, it is more likely to act as a slightly charge sensitive RF
oscillator with a lot of noise.

Not when you do a proper design. The opamp is feedback capacitive,
making it very stable. A small series resistor at the input may be
wise.

Putting a feedback capacitor on a high performance op-amp does not make it
"very stable". You have to use a "unity gain stable" one in this sort of
application. Without the small resistor you suggest, the op-amp is fairly
likely to oscillate even if it is a "unity gain stable" one if the input
cable is very long. The cable and capacitor look like a tuned circuit.

Low sensitivity to outside disturbance.

No, this is not the case. The circuit you suggest is this:

Noise
!
---Cnoise
---
! A -------------
Signal -[Zgen]---+-------! Z Amplifier !---- GND
-------------


You can't improve the signal to noise at "A" by making Zamplifier low.
How ever much you lower the amplifiers input impedance, you must also
raise its voltage gain by too to maintain the same signal output.

Just for fun, lets assume that Zgen=Cnoise and the Signal(RMS) is equal
to the noise(RMS).

High Z amplifier case: SNR = 1:1

Low Z amplifier case: SNR = 1:1

The other circuits are sensitive. The one is decribe here isnt. When
then is a short-circuit, there is no voltage. Where there is no
voltage, that voltage can't change either.

When there's no impedance, the current is infnite and thus the noise is
infinite. It was exactly the sort of circuit you are suggesting that
required the shock mounting of cables.

But it is best not to use
phantom feeding, but a separate power supply cable. The phantom's
voltage does interfere.

Try tapping of a length of COAX even with "no voltage" on it. You will
still see spikes when you strike it. Just hook a BNC-BNC cable onto your
scope's input and wack it against the work bench.

that it does. unless its terminated.

I didnt do a very good job, but I tried to isolate both connector ends,
and only whack the middle. simply waving an end about picked up a lot if
noise. my crude attempt at isolation greatly reduce the resultant spike.
its extremely sensitive at the connector.

I wonder if its because the open end is moving thru an otherwise
stationary E field. so I wrap it in tin foil, and it makes no difference....

so whats going on here? is it simply VdC/dt, with dC/dt due to the
pressure wave propagating thru the cable?

Cheers
Terry

 

[Ken Smith]

Ken Smith wrote: [....]

Try tapping of a length of COAX even with "no voltage" on it. You will
still see spikes when you strike it. Just hook a BNC-BNC cable onto your
scope's input and wack it against the work bench.

that it does. unless its terminated.

I didnt do a very good job, but I tried to isolate both connector ends,
and only whack the middle. simply waving an end about picked up a lot if
noise. my crude attempt at isolation greatly reduce the resultant spike.
its extremely sensitive at the connector.

I wonder if its because the open end is moving thru an otherwise
stationary E field. so I wrap it in tin foil, and it makes no difference....

so whats going on here? is it simply VdC/dt, with dC/dt due to the
pressure wave propagating thru the cable?

If you hit it hard enough, you get triboelectric effects. This is the
voltage generated when you mechanically break most materials. At lower
hits, I believe, it is charges trapped in the plastic materials that is
doing it. At the terminated end, the parts can move with respect to each
other more easily.

 

[Terry Given]

Ken Smith wrote:
[....]

Try tapping of a length of COAX even with "no voltage" on it. You will
still see spikes when you strike it. Just hook a BNC-BNC cable onto your
scope's input and wack it against the work bench.

that it does. unless its terminated.

I didnt do a very good job, but I tried to isolate both connector ends,
and only whack the middle. simply waving an end about picked up a lot if
noise. my crude attempt at isolation greatly reduce the resultant spike.
its extremely sensitive at the connector.

I wonder if its because the open end is moving thru an otherwise
stationary E field. so I wrap it in tin foil, and it makes no difference....

so whats going on here? is it simply VdC/dt, with dC/dt due to the
pressure wave propagating thru the cable?

If you hit it hard enough, you get triboelectric effects. This is the
voltage generated when you mechanically break most materials. At lower
hits, I believe, it is charges trapped in the plastic materials that is
doing it. At the terminated end, the parts can move with respect to each
other more easily.

makes sense. same reason semiconductor physics tests all assume an
un-stressed material.

Cheers
Terry

 

[Joseph2k]

Ken Smith wrote:
[....]

Try tapping of a length of COAX even with "no voltage" on it. You will
still see spikes when you strike it. Just hook a BNC-BNC cable onto
your scope's input and wack it against the work bench.

that it does. unless its terminated.

I didnt do a very good job, but I tried to isolate both connector ends,
and only whack the middle. simply waving an end about picked up a lot if
noise. my crude attempt at isolation greatly reduce the resultant spike.
its extremely sensitive at the connector.

I wonder if its because the open end is moving thru an otherwise
stationary E field. so I wrap it in tin foil, and it makes no
difference....

so whats going on here? is it simply VdC/dt, with dC/dt due to the
pressure wave propagating thru the cable?

If you hit it hard enough, you get triboelectric effects. This is the
voltage generated when you mechanically break most materials. At lower
hits, I believe, it is charges trapped in the plastic materials that is
doing it. At the terminated end, the parts can move with respect to each
other more easily.

makes sense. same reason semiconductor physics tests all assume an
un-stressed material.

Cheers
Terry

Not just triboelectric effects, piezoelectric effects as well.
Picocoulomb/g accelerometer sensors require very special cables which
require very special connectors in turn. I have had to use these and
special microwave cables (vibration resistant) when testing waveguide relay
switches under vibration. I think i still have a made-up cable somewhere.
The tooling for terminating such an accelerometer cable is probably in
excess of US$10,000 by now, 20 years ago the cable was US$20/ft.

 

[Terry Given]

Terry Given wrote:

Ken Smith wrote:

[....]


Try tapping of a length of COAX even with "no voltage" on it. You will
still see spikes when you strike it. Just hook a BNC-BNC cable onto
your scope's input and wack it against the work bench.

that it does. unless its terminated.

I didnt do a very good job, but I tried to isolate both connector ends,
and only whack the middle. simply waving an end about picked up a lot if
noise. my crude attempt at isolation greatly reduce the resultant spike.
its extremely sensitive at the connector.

I wonder if its because the open end is moving thru an otherwise
stationary E field. so I wrap it in tin foil, and it makes no
difference....

so whats going on here? is it simply VdC/dt, with dC/dt due to the
pressure wave propagating thru the cable?


If you hit it hard enough, you get triboelectric effects. This is the
voltage generated when you mechanically break most materials. At lower
hits, I believe, it is charges trapped in the plastic materials that is
doing it. At the terminated end, the parts can move with respect to each
other more easily.

makes sense. same reason semiconductor physics tests all assume an
un-stressed material.

Cheers
Terry

Not just triboelectric effects, piezoelectric effects as well.
Picocoulomb/g accelerometer sensors require very special cables which
require very special connectors in turn. I have had to use these and
special microwave cables (vibration resistant) when testing waveguide relay
switches under vibration. I think i still have a made-up cable somewhere.
The tooling for terminating such an accelerometer cable is probably in
excess of US$10,000 by now, 20 years ago the cable was US$20/ft.

as a general rule of thumb, everything is far more complex than at first
suspected....

Thanks for some informative posts guys

Cheers
Terry