True. However spectral views are typically power spectral density plots. That's why scalloping is down by 3dB for frequencies exactly between bins. Otherwise it would be 6dB, wouldn't it?
Yeah, and how much time and money does it cost to prove every little thing that is audible? If you want to pay for the cost of a study, I can arrange to repeat the experiment.
Anyway, of course current noise is an intermodulation with an audio signal, since it is a multiplicative process. Its like you have a noise source and a VGA. You feed noise into the VGA, then you use an audio signal to drive the VGA gain, then add the VGA output with the audio signal. How is that not producing a modulated noise effect? The noise level goes up and down with the audio. If the noise is going up and down then there must be sidebands (which is well known by people who do the electronics) because the noise amplitude is not composed of steady-state sine waves any more. There have to be few frequencies created that correspond with the gain modulation.
Moreover, the people who manufacture resistors (such as Susumu) and make special resistors for audio say that the audio resistors are optimized for low noise. Even those guys know current noise can be audible.
Last edited:
Which is a clear reason why one would use a steady state signal because then this won't happen.
Dito
I have used cross correlation methods extensively myself (classical signal processing.) I don't use them anymore because other methods are far superior.
Then don't make the mistake that you described.
It has been shown many many times that measurements done correctly always show the exact same things regardless of how they are performed - sweep sines, noise, whatever, all yield identical results.
Sorry, but sigma delta dacs can do just that. The effects are called chirps.
Also, what about when RF gets into an amplifier? According to Bill Whitlock it often happens through the AC power line. According to him the sound effect can be "veiled" or "grainy." Have you ever measured the audibility of veiled or grainy?
I mean, you are saying "don't do that," for real problems that plague people's systems.
But that doesn't matter, the phase is still there in the transfer function regardless of how it is displayed.
This system is nonlinear. Multiplicative processes in the time domain are not linear. Time domain signals in a linear system can only add.
Then they are correlated and that's not common noise. We usually think of noise as being uncorrelated to the signal. Perhaps there is some strange process in electronics that allows this (it is nonlinear,) but I don't think that it is common.
I'm not going to justify crappy designs.
Again, it's not my area, but your claims seem unlikely to me. RF shielding is not a new subject - fix it.
It does matter exactly because a human is interpreting the visual plot and concluding with absolute certainty that AM and FM can't sound the same because the what they person is looking at doesn't look different.
Of course not. But you have to understand you are justifying Ethan's ignorance when it comes to misinterpreting what he is looking at.
He designs something with RF getting into it, with correlated noise, can't see any of that crap on his computer display, then makes up really poorly designed tests with all kinds of veiled, grainy, correlated masking noise in them to prove to people they can't hear crap when they are actually hearing real problems in their systems. And those same poor people don't know how to troubleshoot because they are using the same stupid spectral views because that's what everybody does.
Last edited:
@Markw4 Can you provide some measurements of the behavior that is described above? Some plots of results and explanations would hopefully help me understand the nature of the problem much better than I do at present.
Isn't that just an example that a human is simply looking at the wrong type of plot for the problem at hand? Transfer function analysis is a readily available tool.
Sometimes the measurements are incorrectly referred to a phase noise (PN) skirts. Sometimes they have also been presented at ASR that way (as a measure of jitter more sensitive than J-test). However the skirts also contain information about intermodulated (correlated) amplitude noise (AN).
Some measurements and an overall thread worth reading is at: https://www.diyaudio.com/community/threads/phase-noise-in-ds-dacs.387862/post-7063038
Regarding the relationship between AN and PN skirts, a diagram may be helpful:
Also, a recent paper on separating AN from PN in dacs is attached.
Please note that the effect of current noise is described in another thread as an edit at the end of main post: https://www.diyaudio.com/community/...-for-measuring-amplifiers.363385/post-8046825
Last edited:
Hopefully the noise skirts discussion will help make clear its not a matter of transfer function either.
As an aside, as shown in one of the other threads, PN of analog oscillators can be a problem too. I believe the second image in the post at: https://www.diyaudio.com/community/...-for-measuring-amplifiers.363385/post-8046873 ...is of an analog oscillator. One solution to that problem is to use dacs as test signal sources combined with an LP filter to suppress HD in the dac output. PN of test tones produced that way can be superior to analog oscillators.
Last edited:
But isn't RFI a well-known problem? There is plenty of sensitive electronic gear out there that would be significantly and adversely affected by RFI, not just audio gear where so-called "veiling" and "graininess" might surface. So, don't all audio equipment designers ostensibly guard against RFI during the equipment design process?
Last edited: