(since I've taken a vow not to post again in the sound vs measurements thread which I find to be an utter waste of server space).
I've made measurements of this type before but never bothered to investigate too much or post them.
here's the thing. 2 amps with pretty different sound. the measurement is made at the level I normally listen at, that is ~9V RMS with full-scale sine input (4 ohm, 83 dB W/m speakers, impedance equalized, 4x5 m room, listening position at ~2.something meters away from the speakers plane). I've used a song which is especially revealing wrt sonic differences.
one of the amps has average harmonic distortion (3rd order harmonic @-50dB, the rest vanishing pretty rapidly) but I've tried to simulate something similar in software and it doesn't seem to explain the differences (quick math confirms it, at 20W in 4 ohms THD is ~0.2%). the other one has considerably lower distortion.
oh, and the load was the actual speaker playing music.
so, here are the 3 time-domain pics. amp input vs outputs of both amps. the outputs were low-pass filtered at 250Hz to eliminate phase-shift issues at high frequencies due to different corner frequencies.
if it's not already obvious, both amps are flat in that band (<250Hz).
first interesting question is, which of the two is the better sounding one?
(try to view the files with IrfanView or a similar viewer, and switch from one to another in order to see the differences more easily)
I've made measurements of this type before but never bothered to investigate too much or post them.
here's the thing. 2 amps with pretty different sound. the measurement is made at the level I normally listen at, that is ~9V RMS with full-scale sine input (4 ohm, 83 dB W/m speakers, impedance equalized, 4x5 m room, listening position at ~2.something meters away from the speakers plane). I've used a song which is especially revealing wrt sonic differences.
one of the amps has average harmonic distortion (3rd order harmonic @-50dB, the rest vanishing pretty rapidly) but I've tried to simulate something similar in software and it doesn't seem to explain the differences (quick math confirms it, at 20W in 4 ohms THD is ~0.2%). the other one has considerably lower distortion.
oh, and the load was the actual speaker playing music.
so, here are the 3 time-domain pics. amp input vs outputs of both amps. the outputs were low-pass filtered at 250Hz to eliminate phase-shift issues at high frequencies due to different corner frequencies.
if it's not already obvious, both amps are flat in that band (<250Hz).
first interesting question is, which of the two is the better sounding one?
(try to view the files with IrfanView or a similar viewer, and switch from one to another in order to see the differences more easily)
Attachments
Last edited:
sorry, but I don't see how that is possible. I agree, maybe I'm missing something but at this point I don't know what that may be. "slight" (how slight?) FR variations translating to huge time-domain variations? are you sure? have any quantitative example? 0.5dB means a 5% difference on a linear scale.
first, I would say the time-domain differences are huge, I didn't need a looking glass
second, there are no significant freq. response variations (mentioned in the original post).
and, speaking of which, guess which of the 2 amps has the FR that most deviates from flat?
are you sure you switched from one pic to another like I described? (amp in vs A, amp in vs B) using the forum pic viewer is useless unless you're very patient.
PS: IrfanVIew is free
later edit: the level difference seen in the first few negative spikes is ~1.5dB, I think that would be pretty oticeable in a FR plot
Last edited:
It can be done in windoze if you save the pics.
Anyway, it's an interesting approach, I just don't know how to interpret the results.
Is there some software that would allow an FFT of that difference signal? Basically, I'd like to see the spectrum of the I/O difference with a musical signal. I know that's not all there is to it, but I would be a help to get started.
Anyway, it's an interesting approach, I just don't know how to interpret the results.
Is there some software that would allow an FFT of that difference signal? Basically, I'd like to see the spectrum of the I/O difference with a musical signal. I know that's not all there is to it, but I would be a help to get started.No, I didn't compare the pics as I don't see the point. It is easy to produce a big time domain difference which is almost inaudible - just change the phase a little. On the other hand, some small time domain differences can be quite audible - perhaps a bit of crossover or peak clipping but then smoothed by a low pass filter.
Time domain pics are good for debugging/repairing faults but poor for determining sound quality.
Time domain pics are good for debugging/repairing faults but poor for determining sound quality.
the problem with difference signal (null testing) is that you need to make sure there's no phase shift involved. which is practically never true and even pretty severe with some amps.
I'll give you a hint: one of the amps is bass shy and pretty uninvolving
DF96, don't get me wrong, I'm sure what is seen in the pics above is pretty simple to explain, but I dare you to come up with a frequency response that is +/- 0.05 dB 10Hz-3kHz but gives time-domain variations of ~1.5dB. of course, non-linearities aside, since it's obviously the only thing that can explain those differences, otherwise we're into twilight zone
I'll give you a hint: one of the amps is bass shy and pretty uninvolving
DF96, don't get me wrong, I'm sure what is seen in the pics above is pretty simple to explain, but I dare you to come up with a frequency response that is +/- 0.05 dB 10Hz-3kHz but gives time-domain variations of ~1.5dB. of course, non-linearities aside, since it's obviously the only thing that can explain those differences, otherwise we're into twilight zone
Are you saying that the amps have the same measured frequency response into the same real speaker load? Or the same response into resistors?
'Bass shy' means what, given the same frequency response?
You can get phase changes without frequency changes if the circuit includes an all-pass filter or some other non-minimum phase filter (typically, any circuit where there are two or more parallel signal paths feeding the output - can be deliberate or accidental).
'Bass shy' means what, given the same frequency response?
You can get phase changes without frequency changes if the circuit includes an all-pass filter or some other non-minimum phase filter (typically, any circuit where there are two or more parallel signal paths feeding the output - can be deliberate or accidental).
who said there's any phase shift involved? a time domain difference being almost inaudible has nothing to do with my original post, not sure what's your point.
I'll summarize again:
both amps have flat FR responses, actually the worse time-domain plot is of the amp which is the flattest (+/- 0.05 dB).
I'll do some frequency response measurements tomorrow. at least the ones published by a well-known magazine would not explain it.
maybe it's exactly something along those lines that this topic aims at?
do I generally strike you as the type who expects that there is some "obscure" or undiscovered part of audio engineering? actually, I had suspected that the amp becomes highly non-linear above a certain level (and that is intentional). which is not visible is the publised plots because the measurement is done ~2W in 4 ohms.
meanwhile I did FFTs of the pictured time-domain data. in the one displaying larger spikes, there is actually a pretty consistent difference as compared to the input signal and the other amp, as expected. but the difference is pretty localized at ~120 Hz and is ~4dB high (!). I'll do FR measurements tomorrow with the real (speaker) load. I trust my drivers will withstand short sweeps of ~20W
at any rate, there's something interesting to be learned. maybe I have a defective (but nevertheless good sounding) unit, maybe the published data is intentionally misleading... we'll see.
IMO, if the amps are flat they probably don't have phase differences if the signal is well in the passband. Few amps include all-pass filters! Have you done any square wave testing into speaker loads? They might well have different stability or zobels or something that explains the differences, even though steady state signals look identical. I'd also look for stray RF from the PS or stability related as I've seen RF oscillations on the peaks of waveforms change the average path of the waveform. This is the sort of test I like because it goes after real world differences. OTOH, I've never heard two amplifiers that sounded different unless something was wrong with one or both. Correctly designed and functioning power amps operated at moderate levels are and should be indistinguishable. Any amp that's different from the average is suspect.
I too believe that at least major differences are easily observable, given the right measurements.
you wouldn't believe though the amount of people that are sure the same major differences are unmeasurable
I'll do square-wave measurements too.
the last thing I would expect from a solid-state amp is FR deviations larger than 0.5dB. what I saw in the FFTs I just did contradicts all my expectations and is surely not confirmed by the independent measurements.
ok, I'll admit one more thing to me, one of the most baffling things in audio has been the anecdotal "speaker driving (in)ability". which I have experienced myself in the past. I've always believed (and still do) that it is a very basic phenomenon that only audiosnobs believe to be unmeasurable. I keep reading stories about gigavolts/femtosecond or hundreds of amps being required for short durations but so far I haven't encountered step-like signals in music and the usual speakers are never pure capacitors from an electrical point of view
and since what I'm experiencing with this amp is partially what some would describe as "solid bass", this thread came to life.
PS: in case you're wondering and to prevent any further discussion on this, the amp has never been repaired, although I'm not the first owner. it's built in such a way that even the most basic repair would require the "special" output wires to be bent and/or cut. they look untouched.
you wouldn't believe though the amount of people that are sure the same major differences are unmeasurable
I'll do square-wave measurements too.
I believe the same, that's why I'm investigating these major differences that aren't visible in the published independent measurements.
the last thing I would expect from a solid-state amp is FR deviations larger than 0.5dB. what I saw in the FFTs I just did contradicts all my expectations and is surely not confirmed by the independent measurements.
ok, I'll admit one more thing
to me, one of the most baffling things in audio has been the anecdotal "speaker driving (in)ability". which I have experienced myself in the past. I've always believed (and still do) that it is a very basic phenomenon that only audiosnobs believe to be unmeasurable. I keep reading stories about gigavolts/femtosecond or hundreds of amps being required for short durations but so far I haven't encountered step-like signals in music and the usual speakers are never pure capacitors from an electrical point of view and since what I'm experiencing with this amp is partially what some would describe as "solid bass", this thread came to life.
PS: in case you're wondering and to prevent any further discussion on this, the amp has never been repaired, although I'm not the first owner. it's built in such a way that even the most basic repair would require the "special" output wires to be bent and/or cut. they look untouched.
Last edited:
Using a differencing approach to finding out meaningful variations, or to rate how things rank is fraught with difficulties. Just last night I tried comparing the original track with the nominally next best, copper link, from the latest set in the Interconnect Test thread, doing this -- and, the main conclusion is that they're different! The magic number always seems to be 20dB, the difference spikes to this level, over and over again. But if you zoom in to that point the difference spike seems to mean very little, it's largely an echo of the original signal which didn't quite null, perhaps because there was a momentary slight variation in circuitry gain.
And there are phase issues - one could go nuts trying to align different parts of the spectrum perfectly, and the alignment also seems to 'wobble'. Only very sophisticated algorithms would have a chance of working out precisely what's happening, impossible to do by eyes alone, at least for me ...
And there are phase issues - one could go nuts trying to align different parts of the spectrum perfectly, and the alignment also seems to 'wobble'
. Only very sophisticated algorithms would have a chance of working out precisely what's happening, impossible to do by eyes alone, at least for me ...I've found differential tests to be quite useful. I put an adjustable divider on the output of the amp and use a differential scope amp to compare input and output. Rarely does an amp have enough phase difference to be a big problem, though it's certainly visible. I think that's faster and easier than trying to record the outputs of different amps and comparing them. There have been various amps made with the same circuit and LEDs to detect distortion. I think my Crown PS-200 does that, as did some Japanese amps. I don't know the specific levels but remember reading that the LEDs would trip at quite low THD levels.
Speakers aren't as difficult a load as some believe, but it would be interesting to do the exact same differential test with one probe at the speaker terminal, at the end of the wire, rather than the amp output. My guess is pretty big differences could be seen just because the ratio of speaker impedance and wire impedance isn't a huge number unless you're using welding cable. (I've made current measurements with musical signals and have never seen the big numbers sometimes cited, but I also don't have any exotic speakers that might present very low ohm loads.)
Speakers aren't as difficult a load as some believe, but it would be interesting to do the exact same differential test with one probe at the speaker terminal, at the end of the wire, rather than the amp output. My guess is pretty big differences could be seen just because the ratio of speaker impedance and wire impedance isn't a huge number unless you're using welding cable. (I've made current measurements with musical signals and have never seen the big numbers sometimes cited, but I also don't have any exotic speakers that might present very low ohm loads.)
Last edited:
Yes, performance of high impedance amplifier (or current feedback) v low impedance voltage amplifier easily will have significant differences below <250Hz driving reactive load, that are completely invisible with purely resistive load or no load.
High pass filtering at amplifier input also may have impact. Single pole with input cap needs to be set really low. Higher order input filters twist the phase even more.
ok, looks like we have a dud.
it took a while because I wanted to redo the measurements properly.
I think it was the measurement setup that caused the huge unexplained time-domain difference.
now the time-domain plots look pretty much identical, as expected.
it was worthwhile though, I did some frequency-sweep measurements and it's really intersting to slide the cursor in CoolEdit and see how the distortion spectrum varies with frequency and how it rides above mains-related peaks.
it differs quite a lot between the two amps.
it took a while because I wanted to redo the measurements properly.
I think it was the measurement setup that caused the huge unexplained time-domain difference.
now the time-domain plots look pretty much identical, as expected.
it was worthwhile though, I did some frequency-sweep measurements and it's really intersting to slide the cursor in CoolEdit and see how the distortion spectrum varies with frequency and how it rides above mains-related peaks.
it differs quite a lot between the two amps.
FWIW
here are 2 videos of a 20Hz-20kHz sweep with a resistive load of ~4ohms, one channel working only, at a power of about 20W.
it's interesting to see what happens with the distortion spectum as frequency varies.
amp1 - YouTube
amp2 - YouTube
so we have:
amp 1 - pretty high distortion but practically frequency-independent and mains-related components
amp 2 - lower distortion, no mains-related components but distortion varies with frequency (harmonics distribution is different too) and rises at the top end
to me and everyone else who listened, amp 1 sounds clearly better.
here are 2 videos of a 20Hz-20kHz sweep with a resistive load of ~4ohms, one channel working only, at a power of about 20W.
it's interesting to see what happens with the distortion spectum as frequency varies.
amp1 - YouTube
amp2 - YouTube
so we have:
amp 1 - pretty high distortion but practically frequency-independent and mains-related components
amp 2 - lower distortion, no mains-related components but distortion varies with frequency (harmonics distribution is different too) and rises at the top end
to me and everyone else who listened, amp 1 sounds clearly better.
Last edited:
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.