I'm not sure why FFT spectrum analyzers are being brought up? The analyzers in question were analog - no FFT involved. These are heterodyne analyzers (possibly the same thing you are referring to as "Hilbert transform"?) that boast dynamic range of around 140dB (although the spurious signal response is only like -80dB, as I described previously).
>Every signal deviation known to man can be tried for it's audibility in well devised tests, >and I'm afraid FFT meters (with usually pretty narrow-defined implementations) simply >aren't a good signal averaging for measuring most of the errors that could play a role. >They're just popular and everybody uses them, but that's not a proof of fitness. Can you refer us to any tests which indicate that signals similar to those used in Monty's video contain audible distortion? >A simple remark about the noise, and measuring signals: in information theory, there's >a big difference between a tape with (additive and some multiplicative) noise, and fixed >number of bits for a sampled source signal. Only without proper dithering, unless I am misunderstanding what you are trying to say there. Feel free to go into as much depth as necessary: I am an expert on information theory. >I want to progress, not teach half the world of so-so signal processors the >foundations of EE, university level. I think you'll find that the best route to progress is a combination of humility, respect, clarity and careful adherence to repeatable testing. >the errors being made are trivial for a lot of listeners and can very well be qualified and quantified, I'm sure I speak for many participants here when I say that I'd very much like to see any repeatable test results that illustrate whatever point you are trying to make. E On Mon, Feb 17, 2014 at 3:10 PM, Theo Verelst <theo...@theover.org> wrote: > Ethan Duni wrote: > >> " >> >> I recall from the first measurement shown on the analog spectrum analyzer >>> the source sine wave had 2d or 3d harmonic distortion component of -70 >>> dB. >>> My power amps are noticeably better than that. >>> >>> >> But then they don't have the heterodyne circuits you need for an analogue >> frequency analyser either." >> ... >> > > I don't know your background in this matter, but indeed I meant that the > test signal wouldn't be pure enough to test my power amps with, in a > certain (normal) interpretation, so I indeed did not presume an error in > the spectrum analyzer of that level. But I didn't check it's brand/type > number, so that is possible. Spectrum analysis in the sense of a Hilbert > transform with some form of weighing for repetitive signals is an art that > can go very for (say 150 dB) om terms of dynamic range, very, very far > outside the audio band, so I don't think it matters much what the exact SA > was, which was used here. > > Every signal deviation known to man can be tried for it's audibility in > well devised tests, and I'm afraid FFT meters (with usually pretty > narrow-defined implementations) simply aren't a good signal averaging for > measuring most of the errors that could play a role. They're just popular > and everybody uses them, but that's not a proof of fitness. > > A simple remark about the noise, and measuring signals: in information > theory, there's a big difference between a tape with (additive and some > multiplicative) noise, and fixed number of bits for a sampled source > signal. If i don't near any understanding about what this means in terms of > signal self correlation and the ability to find all kinds of information, i > cannot agree the reasoning about the number of bits is up to par with what > I'm talking about. FFT's measure an amount of self-correlation (or create > that when used in the signal path), and this isn't neutral like a decent > Hilbert frequency analysis without ridiculously large bin sizes in relevant > frequency bands. Clear / i hope so, because I'm quitting this discussion, I > want to progress, not teach half the world of so-so signal processors the > foundations of EE, university level. I met to little respect for that, and > believe me, the errors being made are trivial for a lot of listeners and > can very well be qualified and quantified, but apparently there's little > interest for that here, except for rather lame discussion techniques, that > very much bore me. > > T. > > -- > dupswapdrop -- the music-dsp mailing list and website: > subscription info, FAQ, source code archive, list archive, book reviews, > dsp links > http://music.columbia.edu/cmc/music-dsp > http://music.columbia.edu/mailman/listinfo/music-dsp > -- dupswapdrop -- the music-dsp mailing list and website: subscription info, FAQ, source code archive, list archive, book reviews, dsp links http://music.columbia.edu/cmc/music-dsp http://music.columbia.edu/mailman/listinfo/music-dsp
