"Robert Stanton" wrote in message
m

chung wrote in message
news:4844a$402154da$c247604

To test your theory, all you need to do is to measure THD of, say a
1KHz signal at various output levels, with and without ultrasonic
noise. Do you seriously believe the THD will be less if the noise is
present?

If the system is acting like it is quantized, and if that ultrasonic noise
acts as dither, it is a theoretical and observable fact that a
properly-dithered quantized system has zero nonlinear distortion.

I would have to verify this with actual measurements, but just from a
theoretical point of view, noise would slightly reduce crossover
distortion. It would move the signal to a more linear portion of the
transfer curve.

I've verified it with practical measurements... You can do experiments like
this with good DAW software like Audition.

It may be only a small improvement. Even if it caused only a *slight*
reduction in crossover distortion, why get rid of it?

Agreed.

Hysteresis is a memory effect, and it's very different from
cross-over distortion.

Add enough dither to swamp it, and nonlinear distortion due to it is
history.

The mechanism is different but the principle is the same. Bias moves
the operating curve of tape, to a more linear portion of its' range.

Agreed.

"Robert Stanton" wrote in message
m
chung wrote in message
news:817ac$402130d0$c247604

Your statement implys that linear op-amps are not available.

No, my statement implies that it is non-trivial to design an active
filter with 120 dB linearity.

120 dB = 0.0001% THD.

As close as I've seen it - 0.00015% THD+N for two stages including D-A &
A-D:

http://www.pcavtech.com/soundcards/LynxTWO/index.htm

Looking at the pictures...

Second and Third harmonics are about 130 dB below peak operating level, and
the rest of the harmonics are at least 10 dB below that.

If op-amps with better than 120 dB linearity are available, what is
the problem?

No practical problem given that -80 dB nonlinear distortion generally
suffices for audibly transparent operation.

Are you saying the capacitors used would cause
non-linearity?

If they are unbiased electrolytics... This test had at least 2
electrolytics in series with the signal path.

BTW If you know of op-amps with 120 dB linearity, let us know what
they are!! :-)

My recollection is that the LynxTWO uses OPA27s in a balanced configuration.
The measurements shown were made someplace around +4 dBu. So, each half of
the input and output stages were running around -2 dBu which is about 0.5
volt rms.

"Robert Stanton" wrote in message
m
chung wrote in message
news:817ac$402130d0$c247604

Your statement implys that linear op-amps are not available.

No, my statement implies that it is non-trivial to design an active
filter with 120 dB linearity.

120 dB = 0.0001% THD.

As close as I've seen it - 0.00015% THD+N for two stages including D-A &
A-D:

http://www.pcavtech.com/soundcards/LynxTWO/index.htm

Looking at the pictures...

Second and Third harmonics are about 130 dB below peak operating level, and
the rest of the harmonics are at least 10 dB below that.

If op-amps with better than 120 dB linearity are available, what is
the problem?

No practical problem given that -80 dB nonlinear distortion generally
suffices for audibly transparent operation.

Are you saying the capacitors used would cause
non-linearity?

If they are unbiased electrolytics... This test had at least 2
electrolytics in series with the signal path.

BTW If you know of op-amps with 120 dB linearity, let us know what
they are!! :-)

My recollection is that the LynxTWO uses OPA27s in a balanced configuration.
The measurements shown were made someplace around +4 dBu. So, each half of
the input and output stages were running around -2 dBu which is about 0.5
volt rms.

"Robert Stanton" wrote in message
m
chung wrote in message
news:817ac$402130d0$c247604

Your statement implys that linear op-amps are not available.

No, my statement implies that it is non-trivial to design an active
filter with 120 dB linearity.

120 dB = 0.0001% THD.

As close as I've seen it - 0.00015% THD+N for two stages including D-A &
A-D:

http://www.pcavtech.com/soundcards/LynxTWO/index.htm

Looking at the pictures...

Second and Third harmonics are about 130 dB below peak operating level, and
the rest of the harmonics are at least 10 dB below that.

If op-amps with better than 120 dB linearity are available, what is
the problem?

No practical problem given that -80 dB nonlinear distortion generally
suffices for audibly transparent operation.

Are you saying the capacitors used would cause
non-linearity?

If they are unbiased electrolytics... This test had at least 2
electrolytics in series with the signal path.

BTW If you know of op-amps with 120 dB linearity, let us know what
they are!! :-)

My recollection is that the LynxTWO uses OPA27s in a balanced configuration.
The measurements shown were made someplace around +4 dBu. So, each half of
the input and output stages were running around -2 dBu which is about 0.5
volt rms.

"Robert Stanton" wrote in message
m
chung wrote in message
news:817ac$402130d0$c247604

Your statement implys that linear op-amps are not available.

No, my statement implies that it is non-trivial to design an active
filter with 120 dB linearity.

120 dB = 0.0001% THD.

As close as I've seen it - 0.00015% THD+N for two stages including D-A &
A-D:

http://www.pcavtech.com/soundcards/LynxTWO/index.htm

Looking at the pictures...

Second and Third harmonics are about 130 dB below peak operating level, and
the rest of the harmonics are at least 10 dB below that.

If op-amps with better than 120 dB linearity are available, what is
the problem?

No practical problem given that -80 dB nonlinear distortion generally
suffices for audibly transparent operation.

Are you saying the capacitors used would cause
non-linearity?

If they are unbiased electrolytics... This test had at least 2
electrolytics in series with the signal path.

BTW If you know of op-amps with 120 dB linearity, let us know what
they are!! :-)

My recollection is that the LynxTWO uses OPA27s in a balanced configuration.
The measurements shown were made someplace around +4 dBu. So, each half of
the input and output stages were running around -2 dBu which is about 0.5
volt rms.

Robert Stanton wrote:
chung wrote in message news:4844a$402154da$c247604


To test your theory, all you need to do is to measure THD of, say a 1KHz
signal at various output levels, with and without ultrasonic noise. Do
you seriously believe the THD will be less if the noise is present?


I would have to verify this with actual measurements,

You do that.

but just from a
theoritical point of view, noise would slightly reduce crossover
distortion. It would move the signal to a more linear portion of the
transfer curve.

No, it would not. You are thinking of using sufficient DC bias so that
only one side of the complementary outputs stays on, always. That is
different than using noise.


It may be only a small improvment. Even if it caused only a *slight*
reduction in crossover distortion, why get rid of it?


Why, if it does not? All you getting is more noise.


Hysteresis is a memory effect, and it's very different from cross-over
distortion.

The mechanism is different but the prinicple is the same.

Mechanism different = principle different.

Bias moves
the operating curve of tape, to a more linear portion of its' range.


Hysteresis is a memory effect, and it's very different from cross over
distortion.

Bob Stanton

Robert Stanton wrote:
chung wrote in message news:4844a$402154da$c247604


To test your theory, all you need to do is to measure THD of, say a 1KHz
signal at various output levels, with and without ultrasonic noise. Do
you seriously believe the THD will be less if the noise is present?


I would have to verify this with actual measurements,

You do that.

but just from a
theoritical point of view, noise would slightly reduce crossover
distortion. It would move the signal to a more linear portion of the
transfer curve.

No, it would not. You are thinking of using sufficient DC bias so that
only one side of the complementary outputs stays on, always. That is
different than using noise.


It may be only a small improvment. Even if it caused only a *slight*
reduction in crossover distortion, why get rid of it?


Why, if it does not? All you getting is more noise.


Hysteresis is a memory effect, and it's very different from cross-over
distortion.

The mechanism is different but the prinicple is the same.

Mechanism different = principle different.

Bias moves
the operating curve of tape, to a more linear portion of its' range.


Hysteresis is a memory effect, and it's very different from cross over
distortion.

Bob Stanton

Robert Stanton wrote:
chung wrote in message news:4844a$402154da$c247604


To test your theory, all you need to do is to measure THD of, say a 1KHz
signal at various output levels, with and without ultrasonic noise. Do
you seriously believe the THD will be less if the noise is present?


I would have to verify this with actual measurements,

You do that.

but just from a
theoritical point of view, noise would slightly reduce crossover
distortion. It would move the signal to a more linear portion of the
transfer curve.

No, it would not. You are thinking of using sufficient DC bias so that
only one side of the complementary outputs stays on, always. That is
different than using noise.


It may be only a small improvment. Even if it caused only a *slight*
reduction in crossover distortion, why get rid of it?


Why, if it does not? All you getting is more noise.


Hysteresis is a memory effect, and it's very different from cross-over
distortion.

The mechanism is different but the prinicple is the same.

Mechanism different = principle different.

Bias moves
the operating curve of tape, to a more linear portion of its' range.


Hysteresis is a memory effect, and it's very different from cross over
distortion.

Bob Stanton

Robert Stanton wrote:
chung wrote in message news:4844a$402154da$c247604


To test your theory, all you need to do is to measure THD of, say a 1KHz
signal at various output levels, with and without ultrasonic noise. Do
you seriously believe the THD will be less if the noise is present?


I would have to verify this with actual measurements,

You do that.

but just from a
theoritical point of view, noise would slightly reduce crossover
distortion. It would move the signal to a more linear portion of the
transfer curve.

No, it would not. You are thinking of using sufficient DC bias so that
only one side of the complementary outputs stays on, always. That is
different than using noise.


It may be only a small improvment. Even if it caused only a *slight*
reduction in crossover distortion, why get rid of it?


Why, if it does not? All you getting is more noise.


Hysteresis is a memory effect, and it's very different from cross-over
distortion.

The mechanism is different but the prinicple is the same.

Mechanism different = principle different.

Bias moves
the operating curve of tape, to a more linear portion of its' range.


Hysteresis is a memory effect, and it's very different from cross over
distortion.

Bob Stanton

Robert Stanton wrote:

chung wrote in message news:817ac$402130d0$c247604

Your statement implys that linear op-amps are not available.

No, my statement implies that it is non-trivial to design an active
filter with 120 dB linearity.

120 dB = 0.0001% THD.

If op-amps with better than 120 dB linearity are available, what is
the problem? Are you saying the capacitiors used would cause
non-linearity?

Please read what I said: "It is non-trivial to design an active
filter with 120 dB linearity".


Bob Stanton

BTW If you know of op-amps with 120 dB linearity, let us know what
they are!! :-)

Sounds like you answered your own question.

Robert Stanton wrote:

chung wrote in message news:817ac$402130d0$c247604

Your statement implys that linear op-amps are not available.

No, my statement implies that it is non-trivial to design an active
filter with 120 dB linearity.

120 dB = 0.0001% THD.

If op-amps with better than 120 dB linearity are available, what is
the problem? Are you saying the capacitiors used would cause
non-linearity?

Please read what I said: "It is non-trivial to design an active
filter with 120 dB linearity".


Bob Stanton

BTW If you know of op-amps with 120 dB linearity, let us know what
they are!! :-)

Sounds like you answered your own question.

Robert Stanton wrote:

chung wrote in message news:817ac$402130d0$c247604

Your statement implys that linear op-amps are not available.

No, my statement implies that it is non-trivial to design an active
filter with 120 dB linearity.

120 dB = 0.0001% THD.

If op-amps with better than 120 dB linearity are available, what is
the problem? Are you saying the capacitiors used would cause
non-linearity?

Please read what I said: "It is non-trivial to design an active
filter with 120 dB linearity".


Bob Stanton

BTW If you know of op-amps with 120 dB linearity, let us know what
they are!! :-)

Sounds like you answered your own question.

Robert Stanton wrote:

chung wrote in message news:817ac$402130d0$c247604

Your statement implys that linear op-amps are not available.

No, my statement implies that it is non-trivial to design an active
filter with 120 dB linearity.

120 dB = 0.0001% THD.

If op-amps with better than 120 dB linearity are available, what is
the problem? Are you saying the capacitiors used would cause
non-linearity?

Please read what I said: "It is non-trivial to design an active
filter with 120 dB linearity".


Bob Stanton

BTW If you know of op-amps with 120 dB linearity, let us know what
they are!! :-)

Sounds like you answered your own question.

chung wrote in message news:dd41e$40369df3$c247604

BTW If you know of op-amps with 120 dB linearity, let us know what
they are!! :-)

Sounds like you answered your own question.

It sounds like you are saying there are no op-amps with 120 db
linearity. I would find that easy to believe.

Maybe we should confine discussion to realistic distortion
specificatations.

Bob Stanton

chung wrote in message news:dd41e$40369df3$c247604

BTW If you know of op-amps with 120 dB linearity, let us know what
they are!! :-)

Sounds like you answered your own question.

It sounds like you are saying there are no op-amps with 120 db
linearity. I would find that easy to believe.

Maybe we should confine discussion to realistic distortion
specificatations.

Bob Stanton

chung wrote in message news:dd41e$40369df3$c247604

BTW If you know of op-amps with 120 dB linearity, let us know what
they are!! :-)

Sounds like you answered your own question.

It sounds like you are saying there are no op-amps with 120 db
linearity. I would find that easy to believe.

Maybe we should confine discussion to realistic distortion
specificatations.

Bob Stanton

chung wrote in message news:dd41e$40369df3$c247604

BTW If you know of op-amps with 120 dB linearity, let us know what
they are!! :-)

Sounds like you answered your own question.

It sounds like you are saying there are no op-amps with 120 db
linearity. I would find that easy to believe.

Maybe we should confine discussion to realistic distortion
specificatations.

Bob Stanton

chung wrote in message news:dfbb6$40369db0$c247604
I would have to verify this with actual measurements,

You do that.

Yes, I will and I will post the results, good or bad.



but just from a
theoritical point of view, noise would slightly reduce crossover
distortion. It would move the signal to a more linear portion of the
transfer curve.

No, it would not. You are thinking of using sufficient DC bias so that
only one side of the complementary outputs stays on, always. That is
different than using noise.


I see that you do not expect noise to reduce crossover distortion. It
is very possible that you are right. The difference between you and me
is, I have thought of something new, and I will try it to see if it
works. You on the other hand, reject the idea because it is outside of
your range of experience.


It may be only a small improvment. Even if it caused only a *slight*
reduction in crossover distortion, why get rid of it?


Why, if it does not? All you getting is more noise.


That is a big "if". You are not getting more noise, that noise is
already there on SACD outputs.


Bias moves
the operating curve of tape, to a more linear portion of its' range.


Hysteresis is a memory effect, and it's very different from cross over
distortion.


As the old Wendy's ad said: "parts is parts". Distortion is
distortion, whether it is caused by the non-linearity of magnetic tape
or by the non-linearity of semiconductors. The linearity curve of
unbiased mngnet tape looks very much like the curve of an unbiased
push-pull stage.

Bob Stanton

chung wrote in message news:dfbb6$40369db0$c247604
I would have to verify this with actual measurements,

You do that.

Yes, I will and I will post the results, good or bad.



but just from a
theoritical point of view, noise would slightly reduce crossover
distortion. It would move the signal to a more linear portion of the
transfer curve.

No, it would not. You are thinking of using sufficient DC bias so that
only one side of the complementary outputs stays on, always. That is
different than using noise.


I see that you do not expect noise to reduce crossover distortion. It
is very possible that you are right. The difference between you and me
is, I have thought of something new, and I will try it to see if it
works. You on the other hand, reject the idea because it is outside of
your range of experience.


It may be only a small improvment. Even if it caused only a *slight*
reduction in crossover distortion, why get rid of it?


Why, if it does not? All you getting is more noise.


That is a big "if". You are not getting more noise, that noise is
already there on SACD outputs.


Bias moves
the operating curve of tape, to a more linear portion of its' range.


Hysteresis is a memory effect, and it's very different from cross over
distortion.


As the old Wendy's ad said: "parts is parts". Distortion is
distortion, whether it is caused by the non-linearity of magnetic tape
or by the non-linearity of semiconductors. The linearity curve of
unbiased mngnet tape looks very much like the curve of an unbiased
push-pull stage.

Bob Stanton

chung wrote in message news:dfbb6$40369db0$c247604
I would have to verify this with actual measurements,

You do that.

Yes, I will and I will post the results, good or bad.



but just from a
theoritical point of view, noise would slightly reduce crossover
distortion. It would move the signal to a more linear portion of the
transfer curve.

No, it would not. You are thinking of using sufficient DC bias so that
only one side of the complementary outputs stays on, always. That is
different than using noise.


I see that you do not expect noise to reduce crossover distortion. It
is very possible that you are right. The difference between you and me
is, I have thought of something new, and I will try it to see if it
works. You on the other hand, reject the idea because it is outside of
your range of experience.


It may be only a small improvment. Even if it caused only a *slight*
reduction in crossover distortion, why get rid of it?


Why, if it does not? All you getting is more noise.


That is a big "if". You are not getting more noise, that noise is
already there on SACD outputs.


Bias moves
the operating curve of tape, to a more linear portion of its' range.


Hysteresis is a memory effect, and it's very different from cross over
distortion.


As the old Wendy's ad said: "parts is parts". Distortion is
distortion, whether it is caused by the non-linearity of magnetic tape
or by the non-linearity of semiconductors. The linearity curve of
unbiased mngnet tape looks very much like the curve of an unbiased
push-pull stage.

Bob Stanton

chung wrote in message news:dfbb6$40369db0$c247604
I would have to verify this with actual measurements,

You do that.

Yes, I will and I will post the results, good or bad.



but just from a
theoritical point of view, noise would slightly reduce crossover
distortion. It would move the signal to a more linear portion of the
transfer curve.

No, it would not. You are thinking of using sufficient DC bias so that
only one side of the complementary outputs stays on, always. That is
different than using noise.


I see that you do not expect noise to reduce crossover distortion. It
is very possible that you are right. The difference between you and me
is, I have thought of something new, and I will try it to see if it
works. You on the other hand, reject the idea because it is outside of
your range of experience.


It may be only a small improvment. Even if it caused only a *slight*
reduction in crossover distortion, why get rid of it?


Why, if it does not? All you getting is more noise.


That is a big "if". You are not getting more noise, that noise is
already there on SACD outputs.


Bias moves
the operating curve of tape, to a more linear portion of its' range.


Hysteresis is a memory effect, and it's very different from cross over
distortion.


As the old Wendy's ad said: "parts is parts". Distortion is
distortion, whether it is caused by the non-linearity of magnetic tape
or by the non-linearity of semiconductors. The linearity curve of
unbiased mngnet tape looks very much like the curve of an unbiased
push-pull stage.

Bob Stanton

Robert Stanton wrote:
chung wrote in message news:dfbb6$40369db0$c247604
I would have to verify this with actual measurements,

You do that.

Yes, I will and I will post the results, good or bad.



but just from a
theoritical point of view, noise would slightly reduce crossover
distortion. It would move the signal to a more linear portion of the
transfer curve.

No, it would not. You are thinking of using sufficient DC bias so that
only one side of the complementary outputs stays on, always. That is
different than using noise.


I see that you do not expect noise to reduce crossover distortion. It
is very possible that you are right. The difference between you and me
is, I have thought of something new, and I will try it to see if it
works. You on the other hand, reject the idea because it is outside of
your range of experience.

The difference is that you have not thought through your idea carefully.



It may be only a small improvment. Even if it caused only a *slight*
reduction in crossover distortion, why get rid of it?


Why, if it does not? All you getting is more noise.


That is a big "if". You are not getting more noise, that noise is
already there on SACD outputs.


Bias moves
the operating curve of tape, to a more linear portion of its' range.


Hysteresis is a memory effect, and it's very different from cross over
distortion.


As the old Wendy's ad said: "parts is parts". Distortion is
distortion, whether it is caused by the non-linearity of magnetic tape
or by the non-linearity of semiconductors.

That's interesting. So you are saying cross-over distortion is the same
as clipping distortion? Or jitter distortion?

The linearity curve of
unbiased mngnet tape looks very much like the curve of an unbiased
push-pull stage.

Which part of "memory effect" do you have trouble understanding?

Robert Stanton wrote:
chung wrote in message news:dfbb6$40369db0$c247604
I would have to verify this with actual measurements,

You do that.

Yes, I will and I will post the results, good or bad.



but just from a
theoritical point of view, noise would slightly reduce crossover
distortion. It would move the signal to a more linear portion of the
transfer curve.

No, it would not. You are thinking of using sufficient DC bias so that
only one side of the complementary outputs stays on, always. That is
different than using noise.


I see that you do not expect noise to reduce crossover distortion. It
is very possible that you are right. The difference between you and me
is, I have thought of something new, and I will try it to see if it
works. You on the other hand, reject the idea because it is outside of
your range of experience.

The difference is that you have not thought through your idea carefully.



It may be only a small improvment. Even if it caused only a *slight*
reduction in crossover distortion, why get rid of it?


Why, if it does not? All you getting is more noise.


That is a big "if". You are not getting more noise, that noise is
already there on SACD outputs.


Bias moves
the operating curve of tape, to a more linear portion of its' range.


Hysteresis is a memory effect, and it's very different from cross over
distortion.


As the old Wendy's ad said: "parts is parts". Distortion is
distortion, whether it is caused by the non-linearity of magnetic tape
or by the non-linearity of semiconductors.

That's interesting. So you are saying cross-over distortion is the same
as clipping distortion? Or jitter distortion?

The linearity curve of
unbiased mngnet tape looks very much like the curve of an unbiased
push-pull stage.

Which part of "memory effect" do you have trouble understanding?

Robert Stanton wrote:
chung wrote in message news:dfbb6$40369db0$c247604
I would have to verify this with actual measurements,

You do that.

Yes, I will and I will post the results, good or bad.



but just from a
theoritical point of view, noise would slightly reduce crossover
distortion. It would move the signal to a more linear portion of the
transfer curve.

No, it would not. You are thinking of using sufficient DC bias so that
only one side of the complementary outputs stays on, always. That is
different than using noise.


I see that you do not expect noise to reduce crossover distortion. It
is very possible that you are right. The difference between you and me
is, I have thought of something new, and I will try it to see if it
works. You on the other hand, reject the idea because it is outside of
your range of experience.

The difference is that you have not thought through your idea carefully.



It may be only a small improvment. Even if it caused only a *slight*
reduction in crossover distortion, why get rid of it?


Why, if it does not? All you getting is more noise.


That is a big "if". You are not getting more noise, that noise is
already there on SACD outputs.


Bias moves
the operating curve of tape, to a more linear portion of its' range.


Hysteresis is a memory effect, and it's very different from cross over
distortion.


As the old Wendy's ad said: "parts is parts". Distortion is
distortion, whether it is caused by the non-linearity of magnetic tape
or by the non-linearity of semiconductors.

That's interesting. So you are saying cross-over distortion is the same
as clipping distortion? Or jitter distortion?

The linearity curve of
unbiased mngnet tape looks very much like the curve of an unbiased
push-pull stage.

Which part of "memory effect" do you have trouble understanding?

Robert Stanton wrote:
chung wrote in message news:dfbb6$40369db0$c247604
I would have to verify this with actual measurements,

You do that.

Yes, I will and I will post the results, good or bad.



but just from a
theoritical point of view, noise would slightly reduce crossover
distortion. It would move the signal to a more linear portion of the
transfer curve.

No, it would not. You are thinking of using sufficient DC bias so that
only one side of the complementary outputs stays on, always. That is
different than using noise.


I see that you do not expect noise to reduce crossover distortion. It
is very possible that you are right. The difference between you and me
is, I have thought of something new, and I will try it to see if it
works. You on the other hand, reject the idea because it is outside of
your range of experience.

The difference is that you have not thought through your idea carefully.



It may be only a small improvment. Even if it caused only a *slight*
reduction in crossover distortion, why get rid of it?


Why, if it does not? All you getting is more noise.


That is a big "if". You are not getting more noise, that noise is
already there on SACD outputs.


Bias moves
the operating curve of tape, to a more linear portion of its' range.


Hysteresis is a memory effect, and it's very different from cross over
distortion.


As the old Wendy's ad said: "parts is parts". Distortion is
distortion, whether it is caused by the non-linearity of magnetic tape
or by the non-linearity of semiconductors.

That's interesting. So you are saying cross-over distortion is the same
as clipping distortion? Or jitter distortion?

The linearity curve of
unbiased mngnet tape looks very much like the curve of an unbiased
push-pull stage.

Which part of "memory effect" do you have trouble understanding?

Robert Stanton wrote:

chung wrote in message news:dd41e$40369df3$c247604

BTW If you know of op-amps with 120 dB linearity, let us know what
they are!! :-)

Sounds like you answered your own question.

It sounds like you are saying there are no op-amps with 120 db
linearity. I would find that easy to believe.

Maybe we should confine discussion to realistic distortion
specificatations.



Maybe you have trouble understanding what I said. For the 4th time, I am
saying that it is not trivial to design active filters with -120dB
THD+N. I did not say there are no op-amps with that kind of linearity,
or that capacitors will be the problem.

Think about the care you need to take to get all spurious noise
contributions to be 120dB down. Think about the contribution of noise
from all the passive parts (like resistors) as well as from the op-amps.
Think about the bandwidth requirement for the op-amps to mainatin a good
loop gain so that the filter works correctly, and that all distortion
effects can be reduced. Think about the difficulty of making
mwasurements at the -120dB level.

As an example, say you have two 10K resistors in your active lowpass
filter, and you have the classical Sallen-Key 2nd order configuration.
That 20K will give you a noise voltage over a 20 KHz bandwidth of 2.5
uV, which is more than -120dB below 2VFS. So you can't have anything
approaching 20K in series with the signal.

I did not say that it is impossible to do. I said it is non-trivial,
meaning it is not a trivial task.

Robert Stanton wrote:

chung wrote in message news:dd41e$40369df3$c247604

BTW If you know of op-amps with 120 dB linearity, let us know what
they are!! :-)

Sounds like you answered your own question.

It sounds like you are saying there are no op-amps with 120 db
linearity. I would find that easy to believe.

Maybe we should confine discussion to realistic distortion
specificatations.



Maybe you have trouble understanding what I said. For the 4th time, I am
saying that it is not trivial to design active filters with -120dB
THD+N. I did not say there are no op-amps with that kind of linearity,
or that capacitors will be the problem.

Think about the care you need to take to get all spurious noise
contributions to be 120dB down. Think about the contribution of noise
from all the passive parts (like resistors) as well as from the op-amps.
Think about the bandwidth requirement for the op-amps to mainatin a good
loop gain so that the filter works correctly, and that all distortion
effects can be reduced. Think about the difficulty of making
mwasurements at the -120dB level.

As an example, say you have two 10K resistors in your active lowpass
filter, and you have the classical Sallen-Key 2nd order configuration.
That 20K will give you a noise voltage over a 20 KHz bandwidth of 2.5
uV, which is more than -120dB below 2VFS. So you can't have anything
approaching 20K in series with the signal.

I did not say that it is impossible to do. I said it is non-trivial,
meaning it is not a trivial task.

Robert Stanton wrote:

chung wrote in message news:dd41e$40369df3$c247604

BTW If you know of op-amps with 120 dB linearity, let us know what
they are!! :-)

Sounds like you answered your own question.

It sounds like you are saying there are no op-amps with 120 db
linearity. I would find that easy to believe.

Maybe we should confine discussion to realistic distortion
specificatations.



Maybe you have trouble understanding what I said. For the 4th time, I am
saying that it is not trivial to design active filters with -120dB
THD+N. I did not say there are no op-amps with that kind of linearity,
or that capacitors will be the problem.

Think about the care you need to take to get all spurious noise
contributions to be 120dB down. Think about the contribution of noise
from all the passive parts (like resistors) as well as from the op-amps.
Think about the bandwidth requirement for the op-amps to mainatin a good
loop gain so that the filter works correctly, and that all distortion
effects can be reduced. Think about the difficulty of making
mwasurements at the -120dB level.

As an example, say you have two 10K resistors in your active lowpass
filter, and you have the classical Sallen-Key 2nd order configuration.
That 20K will give you a noise voltage over a 20 KHz bandwidth of 2.5
uV, which is more than -120dB below 2VFS. So you can't have anything
approaching 20K in series with the signal.

I did not say that it is impossible to do. I said it is non-trivial,
meaning it is not a trivial task.

Robert Stanton wrote:

chung wrote in message news:dd41e$40369df3$c247604

BTW If you know of op-amps with 120 dB linearity, let us know what
they are!! :-)

Sounds like you answered your own question.

It sounds like you are saying there are no op-amps with 120 db
linearity. I would find that easy to believe.

Maybe we should confine discussion to realistic distortion
specificatations.



Maybe you have trouble understanding what I said. For the 4th time, I am
saying that it is not trivial to design active filters with -120dB
THD+N. I did not say there are no op-amps with that kind of linearity,
or that capacitors will be the problem.

Think about the care you need to take to get all spurious noise
contributions to be 120dB down. Think about the contribution of noise
from all the passive parts (like resistors) as well as from the op-amps.
Think about the bandwidth requirement for the op-amps to mainatin a good
loop gain so that the filter works correctly, and that all distortion
effects can be reduced. Think about the difficulty of making
mwasurements at the -120dB level.

As an example, say you have two 10K resistors in your active lowpass
filter, and you have the classical Sallen-Key 2nd order configuration.
That 20K will give you a noise voltage over a 20 KHz bandwidth of 2.5
uV, which is more than -120dB below 2VFS. So you can't have anything
approaching 20K in series with the signal.

I did not say that it is impossible to do. I said it is non-trivial,
meaning it is not a trivial task.