I can hear it pretty easily when I switch while music is playing, my problem is that I can tell which, 0 or 180 is the more correct.
Charles Hansen said:
Charles, I am aware of this. The hypothetical question might be which of the two DC "injection" points should be used. I assume there are pros and cons in both cases. What would be less perturbing, to shift FB point, or common gate point?
BTW, I tried to decrease servo gain in "your" connection.
Self-testing.
Take ie the AES/EBU SQAM source material, feed it though an allpass-type multiband compressor plugin with zero gain change settings (say, good old Waves C4), adjust slopes and xo freqs and monitor on your system or preferable maybe, excellent headphones. If you want delay/allpass effects in the opposite direction, reverse the audio during processing.
I utterly failed, exept for the most extremes cases (cascaded C4's with low and steep xo makes the bass really come "late"), also I failed with any absolute polarity switching under blind test conditions.
Further, only a subset of acoustical instruments (and voice, of course) have an identifiable "correct" phase (most drum types don't, and in general all sources with some dipole content in the way the sound source is constructed, say a vibraphone etc). Electrical instruments of any sort surely don't have a correct phase.
Key point is, of course nothing is wrong to set the polarity to your liking. Grey has mentioned that kick drums sounded "better" in one setting than in the other, and that could be declared as the "correct" one.
But to conlude that this is the correct polarity per se is wrong IMHO, especially with a kick drum which clearly has no correct acoustical polarity to begin with. It well may have one that sounds better from a mic's view and that is related to how it is best miked -- also a practical issue for the sound engineer (leakage etc). Now take a floor tom and everthing is vice versa...
Enough of that...
Take ie the AES/EBU SQAM source material, feed it though an allpass-type multiband compressor plugin with zero gain change settings (say, good old Waves C4), adjust slopes and xo freqs and monitor on your system or preferable maybe, excellent headphones. If you want delay/allpass effects in the opposite direction, reverse the audio during processing.
I utterly failed, exept for the most extremes cases (cascaded C4's with low and steep xo makes the bass really come "late"), also I failed with any absolute polarity switching under blind test conditions.
Further, only a subset of acoustical instruments (and voice, of course) have an identifiable "correct" phase (most drum types don't, and in general all sources with some dipole content in the way the sound source is constructed, say a vibraphone etc). Electrical instruments of any sort surely don't have a correct phase.
Key point is, of course nothing is wrong to set the polarity to your liking. Grey has mentioned that kick drums sounded "better" in one setting than in the other, and that could be declared as the "correct" one.
But to conlude that this is the correct polarity per se is wrong IMHO, especially with a kick drum which clearly has no correct acoustical polarity to begin with. It well may have one that sounds better from a mic's view and that is related to how it is best miked -- also a practical issue for the sound engineer (leakage etc). Now take a floor tom and everthing is vice versa...
Enough of that...
Allpass does NOT count. This was shown in the Bell Labs research by Manfred Schroeder. It has to do with the change in nonlinearities generated by the inner ear.
This so called 'allpass' test is often used to disprove hearing 'absolute polarity'. However, MANY other kinds of phase shift exist, check it out.
This so called 'allpass' test is often used to disprove hearing 'absolute polarity'. However, MANY other kinds of phase shift exist, check it out.
First let me apologize for letting a few comments get “under my skin”. I had come with what I thought were a couple of interesting circuit “thought experiments” that I went ahead and tried out. So with a more simplified circuit let me try again. The question was, “ how do you make a low gain, low noise line stage, in a conventional sense i.e. with feedback?” By low noise I mean preserving 1-2nV even at a gain of -1. I came across this while experimenting with some old ideas involving charge amps and microphones. Drawn simplified I hope it’s clear that I’m talking about putting the transducer into a virtual ground and the bias resistor can be large enough so that its low frequency noise is less than the op-amp’s 1/f. Peter Baxandall wrote an article about using loudspeakers as microphones where he discussed reciprocity at length. Well Q=C*V and reciprocity applies just as well to capacitive transducers.
I’ve build several of these and they work fine. Short of taking them to our MEMS mike lab and putting them in a B&K test chamber I can find no difference in performance from the normal JFET circuit except that the distortion holds up rail to rail. I’ve compared frequency response with a spark gap excitation and made some test recordings.
The real point is that A could be any gain cell even John’s 6 device one. Biased up for more open-loop gain so it acts a little like an op-amp and using capacitors in the 100’s of pF range it should work fine. I have tried it with my FET/op-amp combo. Who need’s a 2nV line stage? I don’t know, but people do make line stages with variable transformers at great expense. The input load is no worse than a couple of meters of cable. The amp can easily be made cap load insensitive at high frequencies
Just a little thinking outside the box. BTW except for the 'feedback' cap (which can be as high a quality as desired, it's small) there are no caps in the signal path, no source resitor bypasses none.
I’ve build several of these and they work fine. Short of taking them to our MEMS mike lab and putting them in a B&K test chamber I can find no difference in performance from the normal JFET circuit except that the distortion holds up rail to rail. I’ve compared frequency response with a spark gap excitation and made some test recordings.
The real point is that A could be any gain cell even John’s 6 device one. Biased up for more open-loop gain so it acts a little like an op-amp and using capacitors in the 100’s of pF range it should work fine. I have tried it with my FET/op-amp combo. Who need’s a 2nV line stage? I don’t know, but people do make line stages with variable transformers at great expense. The input load is no worse than a couple of meters of cable. The amp can easily be made cap load insensitive at high frequencies
Just a little thinking outside the box. BTW except for the 'feedback' cap (which can be as high a quality as desired, it's small) there are no caps in the signal path, no source resitor bypasses none.
Attachments
Scott, your model still implies a linear source cap. If not, then what? Please read Erling Fredricksen's paper on cap loading of microphone capsules. IF your setup actually bypasses the problems with cap loading, then OK. If not, maybe you are generating some amount of negative capacitance as an active load for the mike. Just the right amount would do amazing things, but it could not be changed, just to adjust the gain, then. Please read the exact paper in the B&K Technical Review on the B&K website, aka 1996, by Erling Fredreksen so that we get on the same track.
Personally, I like your approach, but it should be problematic, according to Erling.
Also, I do not believe that it is any quieter than any other approach, because the resistor is still the fundamental limiting component at low frequencies, and the capsule damping at mid frequencies for low noise performance. This is also brought out in the B&K Technical Review
Personally, I like your approach, but it should be problematic, according to Erling.
Also, I do not believe that it is any quieter than any other approach, because the resistor is still the fundamental limiting component at low frequencies, and the capsule damping at mid frequencies for low noise performance. This is also brought out in the B&K Technical Review
I was of the impression (perhaps wrongly, school was a long time ago) that a 50 Ohm resistor and a 50 Ohm reactive impedance produced the same amount of noise, with the limitation that the noise of the impedance must be measured at the frequency where the impedance (in this case) equals 50 Ohms.
In other words the universe forces Kelvin noise on us proportional to bandwidth and impedance regardless of type. Am I incorrect?
In other words the universe forces Kelvin noise on us proportional to bandwidth and impedance regardless of type. Am I incorrect?
Caps and ideal inductors do not make noise, BUT they create conditions where noise can be seen that is intrinsically in resistors or any active devices.
In regards to polarity:
The best first reference is: "Models of Hearing" by Manfred R. Schroeder 'Proceedings of the IEEE' Vol 63, No. 9, September 1975 pp. 1132-1350.
In regards to polarity:
The best first reference is: "Models of Hearing" by Manfred R. Schroeder 'Proceedings of the IEEE' Vol 63, No. 9, September 1975 pp. 1132-1350.
Pavel,
Thanks for showing a different way of using a servo configuration which works. Would it not be more effective to use two different kind of servo's:
One servo which compares each output and adjusts them to be close to each other and another servo tries to keep the average of the 2 outputs near Ground (words of wisdom 😉 )
Maybe a combination of your take and the one Charles has suggestged might do the trick?
BTW: how do I get the DC component of Voltage sources in MC8 to supply different voltages. If I change the DC component of one Voltage source, the other changes with it. Quite strange.
As ever, thanks.
Thanks for showing a different way of using a servo configuration which works. Would it not be more effective to use two different kind of servo's:
One servo which compares each output and adjusts them to be close to each other and another servo tries to keep the average of the 2 outputs near Ground (words of wisdom 😉 )
Maybe a combination of your take and the one Charles has suggestged might do the trick?
BTW: how do I get the DC component of Voltage sources in MC8 to supply different voltages. If I change the DC component of one Voltage source, the other changes with it. Quite strange.
As ever, thanks.
Herman, no, only the real part of impedance contributes to Johnson noise, because that's the only dissipative part. There's a fantastic explanation of this in vol 1 of the Feynman Lectures.
"Sigh" John I've read all the papers, they are fundamentally about open loop capacitive transducers this is a short circuited transducer. The paper you mention is a research paper, there is no evidence that it is used in any product certainly not in any audio related one. Most of their products predate 1996 by many years anyway. The cancellation is found by solving an elaborate mathematical model and taking careful measurements of capsule deflection. The cancellation is a very sharp suckout not unlke neutralization. And again, 160dB is where you measure 'knock' in jet engines.
Actually the 'A' weighted contribution of the resistor noise is much less that one thinks. I never said the circuit had less noise just there are no additional sources.
And just to be pedantic, I quote a little physics (couldn't help it).
"The voltage variance at a capacitance C with a noisy resistance in parallel is kT/C, even if the resistance is infinite."
Actually the 'A' weighted contribution of the resistor noise is much less that one thinks. I never said the circuit had less noise just there are no additional sources.
And just to be pedantic, I quote a little physics (couldn't help it).
"The voltage variance at a capacitance C with a noisy resistance in parallel is kT/C, even if the resistance is infinite."
Scott, you haven't shown me much, unfortunately. I, for one, would avoid your approach until I was sure. Still, it should be OK for birdcalls and such. I hope that Calrec does not adopt it, without further testing.
courage said:
Put a "battery" in series with "sine source", for example. Voltage of any battery is independent, you can set it for each of them.
Also, you might like to try pulse/square source, with long pulse duration, like one minute, and observe DC settling. Again, you may put sine in series. Do not forget to make sine source output impedance zero.
Regarding servos, John has already replied.
SY, you and your 'fantastic explanations' . The comments in Vol. 1 on Johnson noise are so dense that I am not at all surprised that most students left his lectures in droves. Are you thinking about the 'drunken sailor' approach?
Hi,Bob Cordell said:
back a few decades ago, I went to Syn-Aud-Con for some training.
There, polarity audibility was demonstrated on live speech.
Reversing the polarity was immediately obvious.
Can I notice with A-B-X on music? not readily
So, like many other things in audio, it may only matter to purists.
As for speakers,
Reversing the midrange polarity is usually done to clean up even worse design ills.
R. Heyser taught me that often such things were unnoticeable until pointed out in a demo.
No. It's derived as an analogy to Brownian motion, and in his derivation, he shows that reactive terms do not generate noise, only shunt it.
- Status
- Not open for further replies.