This helps explain why sometimes people listening to the same amp in different sytems hear it differently. Since few of us have TEF equipment, can you show some examples of worse case add-subtract with different amps/speaker?? That info helps a lot in opening the discusssion on the divergent experiences. Thx-RNM
There is a case where we can cancel lot of power amps distortions, using active multi-way speakers with some added passive filtering.
Reason why i always feel a kind of incoherence in their reproduction, compared with passive filtering, incoherency that i used to tribute to dynamic behavior (PSUs) ?
distortion too low to hear
I wouldnt expect to hear amp harmonics much below .01%. And, I would also agree that something else would be at work if you did hear differences. It is the whole system that needs to be measured. If the whole system is well below .01 then I am not going to worry - I'd be a very happy camper. Then, on to speakers - the weakest link by far.
Input Z variations and source z variations can have infruence at the interfacing that doesnt show up on individual thd bench tests of the equipment. Zo of some designs vs freq vs load z etc etc etc. Other interfacing issues than test bench thd would show up when testing the entire system for thd. How Your thd adds/subtracts will be reveiled.
For the record: The very best IC opamps are now a non-issue as far as thd under its normal operating conditions are concerned or within their design range of operation. But, they are not now used from end to end [record/playback] in the listening chain. In many places they cant be used alone. And, are too expensive for the average Joe to ever experience in his equipment. Until that day happens we have interfacing issues and the like. Esp in need of help are power amps... not anywhere near as good as the best opamps...descrete or IC... and with the most severe interfacing and design condition issues. [ performance of the best opamps at 250-500W into low reactive Z please.. someone... anyone?]
Thx- RNMarsh
I wouldnt expect to hear amp harmonics much below .01%. And, I would also agree that something else would be at work if you did hear differences. It is the whole system that needs to be measured. If the whole system is well below .01 then I am not going to worry - I'd be a very happy camper. Then, on to speakers - the weakest link by far.
Input Z variations and source z variations can have infruence at the interfacing that doesnt show up on individual thd bench tests of the equipment. Zo of some designs vs freq vs load z etc etc etc. Other interfacing issues than test bench thd would show up when testing the entire system for thd. How Your thd adds/subtracts will be reveiled.
For the record: The very best IC opamps are now a non-issue as far as thd under its normal operating conditions are concerned or within their design range of operation. But, they are not now used from end to end [record/playback] in the listening chain. In many places they cant be used alone. And, are too expensive for the average Joe to ever experience in his equipment. Until that day happens we have interfacing issues and the like. Esp in need of help are power amps... not anywhere near as good as the best opamps...descrete or IC... and with the most severe interfacing and design condition issues. [ performance of the best opamps at 250-500W into low reactive Z please.. someone... anyone?]
Thx- RNMarsh
Last edited:
And it is the whole system that is measured .... but at acoustical side FFT measurements show no difference.
IMO, it is the nature of FFT (averaging, longterm measurement) that would not give us answers. We rather need analysis of "transient spectrum" with enough resolution. I also suspect that it is impossible to split linear and non-linear distortions, we are not able to blame sound differences to non-linear distortion in isolation, examined by steady-state, averaging methods.
groan - not again...
the fft contains exactly the same information, and all of the same information, is a mathematical Dual of the time series - any information captured in the time series can be recovered 100% from the fft - just do the inverse transform
the reason to use the fft view is for human readability - some patterns our eyes can pick out better in one format (fft for tonally structured test signals, errors that pile up at discrete frequencies) vs the other
"averaging" implies loss of information - thats not happening with fft
you can average either the input time series or the fft over multiple runs - but that is a entirely independent choice
the fft contains exactly the same information, and all of the same information, is a mathematical Dual of the time series - any information captured in the time series can be recovered 100% from the fft - just do the inverse transform
the reason to use the fft view is for human readability - some patterns our eyes can pick out better in one format (fft for tonally structured test signals, errors that pile up at discrete frequencies) vs the other
"averaging" implies loss of information - thats not happening with fft
you can average either the input time series or the fft over multiple runs - but that is a entirely independent choice
More randomizing for listening tests?
The rhythm or cycling seems to be so that the brain isnt blocking other input or something new/important.
Last edited:
jcx, 64K samples of 48kS/s takes 1.36 second. Even if averaging is set to 1, the FFT averages information during these 1.36 seconds. Non-periodic phenomena are suppressed in resulting amplitude. If you go faster (shorter FFT length), like 8K ... 1K, whatever, you get the view of shorter event, but loosing resolution (bin width). So, we may go fast with poor resolution, or slow with high resolution, but averaged result valid only for periodic signal components.
I know that FFT is a frequency domain interpretation of time domain signals (but we need not only amplitude spectrum but phase spectrum as well to get all the information), but for transients FFT is quite useless for the reason of poor resolution. Time domain is much better then. I agree, give me a 20-bit system at more than 10MS/s and then we can analyze transient signals properly. But not with 100kS/s.
I know that FFT is a frequency domain interpretation of time domain signals (but we need not only amplitude spectrum but phase spectrum as well to get all the information), but for transients FFT is quite useless for the reason of poor resolution. Time domain is much better then. I agree, give me a 20-bit system at more than 10MS/s and then we can analyze transient signals properly. But not with 100kS/s.
Last edited:
No, you only need to concentrate to other effects than a mere non-linear distortion.
If you look at the build quality of some amplifiers you will see a lot spent on RFI shielding, proper routing of wires, high quality power supplies, etc. and often a fairly simple circuit topology. The quest for lower THD goes nowhere, there aleady are things like Halcro.
Someone has mentioned a short impulse. Well, the problem of short impulse analysis (Fourier) is that one would need very high impulse to get enough energy and enough resolution. We may get basic amplitude and phase responses from impulse response and that's all. No new insight and not enough resolution, not enough dynamic range.
Hello Scott
So are you saying that the way forward is to concentrate your efforts on reducing RFI and amount of ripple on the power supplies.
I would think seconds depend on top/bottom of the wave effects so for each device the phase is a coin toss. I was not really addressing identical stages at identicle operating points connected in series.
There are lots of circuits with harmonic null properties.
- Status
- Not open for further replies.