If these things are a problem for the circuit, it will certainly show up in measurement of IMD and THD as these are essentially noise issues, unless these incursions are only absent during measurement 😉
One would hope for competent circuit design that prevents RFI and EMI becoming an issue! I can see this being a problem in some gear but not others, at any rate not that much to do with feedback, although feedback could clean it up. Are we going to have a discussion about RFI amelioration and proper grounding technique- it seems that is what you're talking about in this post.
It will not necessarily show up as IMD or THD. An FFT takes some time to acquire. If low noise is an issue maybe multiple acquisitions are averaged. For HD or IMD measurements there is an assumption that the input is stationary during the FFT acquisition time. During the time acquisition takes, noise correlated with signal can be changing. In dacs it is measured using a PCM code that gives a digital DC input to the dac for a couple of seconds. An FFT snapshot of the noise is taken at that DC offset level. The measurements are repeated across the input level range of the dac. Does it show up as THD or IMD? No, it shows up as a signal correlated changing noise floor. Even if the noise floor is very low, some (perhaps unpublished) research shows that some people, maybe only a few people (training may help more people to recognize the sound signature), notice something about the sound that isn't right, isn't natural sounding, even if IMD and THD measure fine. Fix the problem and the complaints go away.
If more or less the thing thing happens due to low level RF noise incursion into a feedback loop and is demodulated in an high closed loop gain amplifier input stage which is then effectively modulated by an audio signal, the modulated product is for FFT purpose nothing but noise, but nonetheless it may be strongly correlated with the audio signal. Since the process is not stationary with respect to the FFT acquisition time frame, it may be largely invisible. Yet, some people's brains (according to ESS) are exquisitely sensitive to noticing noise changing in a way correlated with an audio signal. My question for the feedback defenders is how are you measuring today to show the magnitude of such effects? Not at all? Then what do you really know about whether N101 is hearing something real or not? If you don't measure I don't think you can know with certainty.
If more or less the thing thing happens due to low level RF noise incursion into a feedback loop and is demodulated in an high closed loop gain amplifier input stage which is then effectively modulated by an audio signal, the modulated product is for FFT purpose nothing but noise, but nonetheless it may be strongly correlated with the audio signal. Since the process is not stationary with respect to the FFT acquisition time frame, it may be largely invisible. Yet, some people's brains (according to ESS) are exquisitely sensitive to noticing noise changing in a way correlated with an audio signal. My question for the feedback defenders is how are you measuring today to show the magnitude of such effects? Not at all? Then what do you really know about whether N101 is hearing something real or not? If you don't measure I don't think you can know with certainty.
Last edited:
According to ESS marketing material at least. I'm sorry Mark I can't take that ESS reference seriously anymore.
Okay Bill, I guess Bruno Putzeys lied too. According to Terry Demol, Bruno described doing the same tests on dacs.
"Yet, some people's brains (according to ESS) are exquisitely sensitive to noticing noise changing in a way correlated with an audio signal."
Can it be shown that the noise wasn't already part of the signal?
Can it be shown that the noise wasn't already part of the signal?
Sure. It isn't just that there is noise. Its that noise if tracking the audio signal someone is listening to. The noise goes up and down with the audio signal. That doesn't happen in nature. It usually doesn't happen in analog electronics either, at least so far as is commonly understood. However, I personally give some credence to Rob Watts' findings. He understands about noise floor modulation in dacs, its part of his consulting expertise. It makes sense he might think to check for it in an analog preamp when most other people wouldn't (recall it was part of an investigation into how power cords could possibly have an audible effect on sound reproduction, which is to say, how they can act as power line noise filters).
Last edited:
"Sure. It isn't just that there is noise. Its that noise if tracking the audio signal someone is listening to. The noise goes up and down with the audio signal. That doesn't happen in nature."
Do you mean with the overall amplitude? Noise is part of the sound of instruments.
Do you mean with the overall amplitude? Noise is part of the sound of instruments.
Thermal effect in solid state amplifier that cause distortion is rare discuss here. Example of thermal modulation in LTP that cause common mode distortion. Or after transient, the thermal modulation cause changing loop gain. Dr. Arto Kolinummi discuss thermal effect in his book.
This kind of distortion can not show in THD measurement, but IM distortion may be can show it.
This kind of distortion can not show in THD measurement, but IM distortion may be can show it.
Dunno, but the “dynamically modulated noise floor” starts looking like a whack-a-mole game. You whack it in a thread, only to pop its head in the next.
Call me when you’ll succeed building an amplifier with measurable IMD, and THD below the noise floor. You’ll be both famous and rich.
OMG , yer back.
I'm only coming close the noise floor now.
I think only the parts are the limiting factor.
OS
You didn’t get it, do you?.
Yes, I didn't get it. I think you did build it. But I'm not sure you are rich and famous.
Of course, I want to be rich but I do not want to be famous.
Yeah, it took me a few months to acquire an FFT 🙂
Noise as you describe will show up in a THD measurement since noise is always part of that measurement.
With regards to noise and our sensitivity to it, we are adapted to survive in an environment that includes wind. As a result, wind sound (like hiss) is the one exception to the masking rule; we can hear a bit into that kind of noise floor more than any else. That seems to be what you are saying above although I do not want to put words in your mouth!
Now its been known for a long time (see Norman Crowhurst) that feedback can result in IMD at the feedback node (avoided by using an available input means like a grid, base or gate to a differential amplifier at the input of a circuit) and higher ordered harmonic and inharmonic information which then composes the resulting noise floor. It does seem as if the ear cannot penetrate this sort of noise floor as it can when no feedback is used so any detail below it is lost.
However if you use enough feedback the circuit is able to clean up the distortion caused by the application of feedback and this seems to include the noise floor of the circuit.
I used to eschew feedback because as a tube amplifier manufacturer it simply wasn't possible to install enough to get the job done properly without exceeding the phase margin of the circuit. So I relied on the innate linearity of the circuit instead, and on the fact that it expressed a cubic non-linearity and not a quadratic non-linearity. Such circuits tend to have less overall distortion.
But then I realized that class D (actually Bruno Putzeys) was offering a way around this problem. You could put a great deal of feedback in a class D circuit and not worry about exceeding the phase margin- instead use the oscillation as the switching frequency, killing two birds with one stone.
As I stated earlier what is important in the behavior of the amplifier is that the distortion is constant at any audio frequency in the passband. A zero feedback amp can do that, and so can an amp that runs enough feedback. If you're in the middle, that's where you get all the brightness and weird noise floor issues. In a nutshell, inadherence to proper control theory.
One of the bigger problems in audio these days in that people don't know how to design a proper feedback network. And kids should stay off my lawn!
Negative feedback never creates anything "inharmonic". Only time-variant systems (with or without feedback), can create inharmonic artifacts (including distortions), and only if the period of the time-varying system is not harmonically related to the input signal.
An example of time variant system related to audio is an amplifier affected by thermal distortions. This has nothing to do with feedback, and electrical feedback cannot correct anything about.
Second comment, feedback does not lower the noise in any way, shape or form. Feedback may help generating less additive noise in the system it is wrapped around, which is a different story.
Intermodulation distortion is created by the same non-linearities as harmonic distortion.
If feedback creates higher order harmonics, it creates higher order IM products, too.
OTOH, think of a piano. Play high A and high A# at the same time. This is dissonant, and the piano will generate all sorts of harmonics of both notes. It will also sound all sorts of IM tones of the pair. You can bet that a spectrum analysis will show 2nd order oroducts at 52 Hz and 1812 Hz. And all the other IM products, as well.
If you play it through a system that has moderate non-linear distortion, it will still sound like a piano, just one with a different harmonic signature. Maybe even a piano that you like the sound of it better. If you want it to sound like the original Bosendorfer, you have to preserve the original harmonic and intermodulation levels as much as possible.
If feedback creates higher order harmonics, it creates higher order IM products, too.
OTOH, think of a piano. Play high A and high A# at the same time. This is dissonant, and the piano will generate all sorts of harmonics of both notes. It will also sound all sorts of IM tones of the pair. You can bet that a spectrum analysis will show 2nd order oroducts at 52 Hz and 1812 Hz. And all the other IM products, as well.
If you play it through a system that has moderate non-linear distortion, it will still sound like a piano, just one with a different harmonic signature. Maybe even a piano that you like the sound of it better. If you want it to sound like the original Bosendorfer, you have to preserve the original harmonic and intermodulation levels as much as possible.
That's what Mr. Bimo didn't get; one cannot get IMD without HD, and the other way around. With or without feedback.