Wimms,
I didn't imagine anything complicated. Just start with a "good" system (how 'transparent' is "transparent"?), then add a component known to introduce benign distortions, for instance a follower without feedback. Say, at the preamp level, using one of the op amp+buffer combos that measure -80 to -100 dB in THD, and make the feedback switchable - so that the buffer can either operate within the op amp feedback loop (buffer feedback corrected) or outside it (buffer harmonics remain unaltered).
Why did I bring this up at all? Because I keep improving my homebrew dipole speakers, and from time to time, I do get clear improvements in transparency, either by fixing FR aberrations, changing the active X-O, improving filters, tryng different amps or preamps, etc. So, I hear more of the recordings' details. But I can't help but noticing that I now hear distracting details and sometimes identifiable distortions on recordings which used to sound just fine. Sometimes I spend weeks and months chasing perceived distortions that only appear on some recordings - only to conclude that it was on the CD in the first place.
I could have said "Be careful what you wish for" 😉
Francis,
interesting paper indeed, though I don't follow in the math. As you said interpretation would be non trivial. For instance I couldn't quite imagine how to derive possible audible effects from the presented distortion error functions. In other words, I don't quite see how to relate the method to testable hypotheses (both in error generation and in audibility). I guess one would have to measure different topologies with this method to derive typical error functions for each of them, and then test whether psychoacoustic results relate to specific curve types.
I didn't imagine anything complicated. Just start with a "good" system (how 'transparent' is "transparent"?), then add a component known to introduce benign distortions, for instance a follower without feedback. Say, at the preamp level, using one of the op amp+buffer combos that measure -80 to -100 dB in THD, and make the feedback switchable - so that the buffer can either operate within the op amp feedback loop (buffer feedback corrected) or outside it (buffer harmonics remain unaltered).
Why did I bring this up at all? Because I keep improving my homebrew dipole speakers, and from time to time, I do get clear improvements in transparency, either by fixing FR aberrations, changing the active X-O, improving filters, tryng different amps or preamps, etc. So, I hear more of the recordings' details. But I can't help but noticing that I now hear distracting details and sometimes identifiable distortions on recordings which used to sound just fine. Sometimes I spend weeks and months chasing perceived distortions that only appear on some recordings - only to conclude that it was on the CD in the first place.
I could have said "Be careful what you wish for" 😉
Francis,
interesting paper indeed, though I don't follow in the math. As you said interpretation would be non trivial. For instance I couldn't quite imagine how to derive possible audible effects from the presented distortion error functions. In other words, I don't quite see how to relate the method to testable hypotheses (both in error generation and in audibility). I guess one would have to measure different topologies with this method to derive typical error functions for each of them, and then test whether psychoacoustic results relate to specific curve types.
Whacky ideas: I am rather interested in feed-forward and error modelling for cancellation. This started with active speaker design thoughts, since feedback in speakers is an almost intractable problem, but the idea of modelling the major distortion issues in either a hardware analogue, or in software, is not such a difficult idea. Similarly for an amplifier, although it could be tricky to say the least.
Nicefier: In the professional audio world impulse responses for convolving with ones audio source are simply part of the landscape. Of course they are rather more brutal than we might prefer. However the equivalent of an Aphex Aural Exciter might be reasonable. I do tend to ascribe to Douglas Self's point, that any niceness attribute in an amplifier should have knob to control it.
It is sad but true that so many recordings are really badly butchered. The 80's seemed to be peculiarly bad. Maybe far to many new toys in the recording studio. I find that the flavour of a recording is often dominated by the the vagaries of the recording process. Different noise floors, the distinct signature noise spectra of different media, are the least of the issues. Trouble is that many of the more brutal insults to the music are unrecoverable. Close micing of orchestral works, nothing can undo that. Sometimes the studio becomes part of the music however. Early Brian Eno for instance. Dreadful from a pure HiFi point of view - flat, over processed, and yet totally involving and very listenable.
Singapore: I go up there every now and again to teach. Spend four days. Highlight is the pilgrimage to Funan and Sim Lim, to buy toys, and the occasional visit to the Adelphi centre for a really good laugh. I have never seen so much astoundingly overpriced audio gear under one roof.
Nicefier: In the professional audio world impulse responses for convolving with ones audio source are simply part of the landscape. Of course they are rather more brutal than we might prefer. However the equivalent of an Aphex Aural Exciter might be reasonable. I do tend to ascribe to Douglas Self's point, that any niceness attribute in an amplifier should have knob to control it.
It is sad but true that so many recordings are really badly butchered. The 80's seemed to be peculiarly bad. Maybe far to many new toys in the recording studio. I find that the flavour of a recording is often dominated by the the vagaries of the recording process. Different noise floors, the distinct signature noise spectra of different media, are the least of the issues. Trouble is that many of the more brutal insults to the music are unrecoverable. Close micing of orchestral works, nothing can undo that. Sometimes the studio becomes part of the music however. Early Brian Eno for instance. Dreadful from a pure HiFi point of view - flat, over processed, and yet totally involving and very listenable.
Singapore: I go up there every now and again to teach. Spend four days. Highlight is the pilgrimage to Funan and Sim Lim, to buy toys, and the occasional visit to the Adelphi centre for a really good laugh. I have never seen so much astoundingly overpriced audio gear under one roof.
Spot on I think. The paper does pull the punches a bit buy not deriving the full Volterra kernel. But if one took the Volterra kernels one essentially has a ready to roll distortion transfer function that includes time varying effects. So it should be possible to extract, or even synthesise, kernels that relate to different physical issues, both in reproduction, and in hearing. It becomes possible to use the kernels to convolve with test sounds that are used for hearing and hear-ability tests. Thus potentially closing the loop.
The down side is that there is probably far too much information, and it will take quite some time to work out how to use it all.
Wimms,
thinking about your link to stored energy in a crossover
http://www.geocities.com/kreskovs/Stored-energy2.html
does the CSD plot not simply show the group delay? Just wondering. If so, can one call group delay a form of energy storage?
Francis,
re: error correction, you might find PMA's error correction amp interesting (many mentions on these forums if you do a search). PMA, if you are reading this, how do you situate your topology and attached reasoning for using it, in this discussion?
Singapore: I assume you know both Sim Lims, the 'Square' with audio/video, and PC's, and the 'Tower', with electronic components? The Tower is my personal toystore, though always a gamble on the quality you get. Too many indistinguishable knock offs.
Adelphi is great, really. Every once in a while I listen into a demo and always conclude I wouldn't want to buy the stuff even if I did have the money.
thinking about your link to stored energy in a crossover
http://www.geocities.com/kreskovs/Stored-energy2.html
does the CSD plot not simply show the group delay? Just wondering. If so, can one call group delay a form of energy storage?
Francis,
re: error correction, you might find PMA's error correction amp interesting (many mentions on these forums if you do a search). PMA, if you are reading this, how do you situate your topology and attached reasoning for using it, in this discussion?
Singapore: I assume you know both Sim Lims, the 'Square' with audio/video, and PC's, and the 'Tower', with electronic components? The Tower is my personal toystore, though always a gamble on the quality you get. Too many indistinguishable knock offs.
Adelphi is great, really. Every once in a while I listen into a demo and always conclude I wouldn't want to buy the stuff even if I did have the money.
I would recommend to turn your attention to experimental work, rather than neverending "theoretical" analysis.
One idea to combat the info overload from this method:
A noise stimulus by definition contains the maximum amount of information. By choosing the stimulus differently and tailored to an objective, one could directly form and test hypotheses (in conjunction with simulations with stipulated error functions).
Not that I could do anywhere near the math required ...
A noise stimulus by definition contains the maximum amount of information. By choosing the stimulus differently and tailored to an objective, one could directly form and test hypotheses (in conjunction with simulations with stipulated error functions).
Not that I could do anywhere near the math required ...
A reasonable point of view. However I think part of the point of this discussion is to think about how that experimental work might proceed. One thing it should not proceed with right now is with a soldering iron. On the other hand I am developing a design for an audio test system, which is a combination of a few of the ideas that we have seen go past here. The GedLee approach is a good basis for the system design, but it lacks time related mechanisms, Hawksford has a highly computationally expensive mechanism that copes with time.
The resultant system could be very valuable. But will be a lot of work.
Hawksford already includes this. He goes beyond just noise by also allowing for a music source to be convolved with the test system. This addresses criticisms that noise isn't music (some will disagree on both theoretical and aesthetic grounds 🙂 ) but allows one to hone down onto music specific issues - much like we have discussed with the dissonance issues of harmonics - without the clutter of an input that, in principle, contains every possible signal. The music signal could indeed be very carefully chosen, and a good starting point would be some of the aural tests used for dissonance testing.
SimLim Tower. I found it a bit overwhelming. Funny place too - I wonder how some of the shops actually do business. But I think one really needs to know the right places to go for stuff. No-one would sell me the IR fets I was after. But I gave up after a while. (Essentially I was on the lookout for bits for an Aleph-X.) Capacitors would be good.
PMA,
well, this discussion tries to put a name on what many believe to hear but which does not reflect in the usual measurements. And I know that you do have practical experience in this. Both your buffer and your error correction amp, obtain audible (I gather) improvements not properly explained by THD measurements. Therefore, I asked you about your views on the subject, from your practical point of view.
Without a better theoretical / conceptual understanding of, and better metrics for, distortion, we will forever stay in the same sterile "objectivist vs subjectivist" debate of the past 10, 15 years. Say, if your new amp sounds better to you than other amps with similar THD, how do you quantify this improvement? How do you explain it to someone who has not heard it? How do you validate the result? How do you make it replicable? How do you predict performance of possible circuits? etc.
Francis,
Sim Lim Tower is a mess... They possibly have everything ever produced on the planet, if one could only find it. I usually go from shop to shop for special parts, and I have some favorites for common parts. I did manage to find IRFP 140's, quite cheap ones too, and stopped looking for IRF240's, when I built my mini-Aleph.
For caps they have a lot, some shops have surplus parts, but you never know whether they're new or they just pulled them from a 20 year old circuit. For "audiophile" caps and components, I found a handful other, obscure shops elsewhere. Some are in Burlington square opposite the pedestrian place behind Sim Lim Square. Some are in People's Park in Chinatown. I have lived in Singapore for 4 years now and I still discover new places with outlandish parts... It's just so full of mass market stuff that the specialty stores are hard to find in the maze.
well, this discussion tries to put a name on what many believe to hear but which does not reflect in the usual measurements. And I know that you do have practical experience in this. Both your buffer and your error correction amp, obtain audible (I gather) improvements not properly explained by THD measurements. Therefore, I asked you about your views on the subject, from your practical point of view.
Without a better theoretical / conceptual understanding of, and better metrics for, distortion, we will forever stay in the same sterile "objectivist vs subjectivist" debate of the past 10, 15 years. Say, if your new amp sounds better to you than other amps with similar THD, how do you quantify this improvement? How do you explain it to someone who has not heard it? How do you validate the result? How do you make it replicable? How do you predict performance of possible circuits? etc.
Francis,
Sim Lim Tower is a mess... They possibly have everything ever produced on the planet, if one could only find it. I usually go from shop to shop for special parts, and I have some favorites for common parts. I did manage to find IRFP 140's, quite cheap ones too, and stopped looking for IRF240's, when I built my mini-Aleph.
For caps they have a lot, some shops have surplus parts, but you never know whether they're new or they just pulled them from a 20 year old circuit. For "audiophile" caps and components, I found a handful other, obscure shops elsewhere. Some are in Burlington square opposite the pedestrian place behind Sim Lim Square. Some are in People's Park in Chinatown. I have lived in Singapore for 4 years now and I still discover new places with outlandish parts... It's just so full of mass market stuff that the specialty stores are hard to find in the maze.
Brings back memories - I used to scour Sim Lim for exotic parts back in the late 70's! And do Murtabak etc.. in Serangoon Rd.
semiNFB
http://gaydenko.com/um/semiNFB.png
I'm not able to find a place in the thread about trying to use an output stage inside and beyond a global NFB loop (but the place exists! 🙂 The schematics (I and LTspice hope) has interesting feature: changing R13, R14 values you can smoothly change a degree output stage is inside NFB loop. So, "semiNFB".
I'll be glad to get any recomendations. I don't mean technical realization moments (they will be beyond the thread). Rather something like this: what music listening aspects must I try to determine?
As for measuring. I can measure FFT with a sound card (it's own the card distortions):
http://gaydenko.com/preamp/card-own-4KHz-500mV.png
An input may be any (static :-( ) combinations of sines.
P.S. I know, my English is ugly. Sorry...
Thanks!
Andrew
Just now I'm finishing preamp (I think, it will take about 2-3 weeks; BTW, without global NFB) and am going to try a schematics which is a variation of Pavel Macura (Pavel, thanks!) follower. It's here:PMA said:
http://gaydenko.com/um/semiNFB.png
I'm not able to find a place in the thread about trying to use an output stage inside and beyond a global NFB loop (but the place exists! 🙂 The schematics (I and LTspice hope) has interesting feature: changing R13, R14 values you can smoothly change a degree output stage is inside NFB loop. So, "semiNFB".
I'll be glad to get any recomendations. I don't mean technical realization moments (they will be beyond the thread). Rather something like this: what music listening aspects must I try to determine?
As for measuring. I can measure FFT with a sound card (it's own the card distortions):
http://gaydenko.com/preamp/card-own-4KHz-500mV.png
An input may be any (static :-( ) combinations of sines.
P.S. I know, my English is ugly. Sorry...
Thanks!
Andrew
Interesting plot for ur circuit...
what are the spikes at ~7kHz, 15, and 15.5khz. coming from?
wonder if the loopback plot shows them too?
_-_-bear
what are the spikes at ~7kHz, 15, and 15.5khz. coming from?
wonder if the loopback plot shows them too?
_-_-bear
It's computer own addition not correlated with signal under testing.bear said:
The plot represents own (loopback) card's distortions - just for information about my measuring capabilities.
I'd want to clarify: the follower isn't under soldering yet. A prototyping of this schematics is in my near plan (I'm going to begin the prototyping 2-3 weeks later).bear said:
amplifierguru said:
I have Terratec Aureon 7.1 Space. I use Linux. FFT is shown by jaaa (http://users.skynet.be/solaris/linuxaudio/).
(Your plot is strange: I was sure Soundblaster's DAC-ADC loopback test must show much more harmonics rather -105db 🙂 Much better DACs have distortions about -90-95db)
OK, I see. Now I'm surprised with your ADC quality 🙂 (typical ADC of $100-200 sound card is aproximately the same as for DAC or less better): if I understand well, with your external filter only ADC quality does matter.amplifierguru said:
Amplifierguru, I'm afraid our card-related discussion is beyond the thread scope. My email is open for you (and for all, of course): a<at>gaydenko<dot>com
Hi anli,
I'd say we're right on topic and less 'out there' than other posts!
But thanks.
Yes, you're right - it's the DAC side (card output) that seems to be the limitation as I've proved - with a few cheap components, a TL082 and a plugpack!
I'd say we're right on topic and less 'out there' than other posts!
But thanks.
Yes, you're right - it's the DAC side (card output) that seems to be the limitation as I've proved - with a few cheap components, a TL082 and a plugpack!
Hi Amplifierguru
amazing that some things in Singaore seem to have stayed the same over 20 years. Overall, the place changes amazingly fast...
We seem to have the same soundcard, Soundblaster Live 24 bit right? I get the same spikes. Funny that it should be a card issue, I thought it was a PC- specific interference ...
amazing that some things in Singaore seem to have stayed the same over 20 years. Overall, the place changes amazingly fast...
We seem to have the same soundcard, Soundblaster Live 24 bit right? I get the same spikes. Funny that it should be a card issue, I thought it was a PC- specific interference ...
- Status
- Not open for further replies.