Sound Quality Vs. Measurements

Status
Not open for further replies.
Do you really know that? I think you don't.

If you want to characterize the distortion of an amp, then at what operating point? You'd have to step the DC operating point through the range of output current and output voltage, then step the test signal frequency and amplitude. As I've shown with my simple example, it is easy to build a circuit which has very low THD at the DC operating point (Vout=0V, Iout=0A) but 100x more THD at a small Iout offset from that. And a zero-centered THD test will completely miss it.

I have not measured these specific amps, but usually a 845 SET amp without NFB has high THD, just an example
Newaudiofrontiers >> INTEGRATED AMPLIFIER 845 SE SPECIAL EDITION

While a LM3886 based amp usually has much less THD
Neurochrome LM3886 Done Right - 45 W, 8 Ω, 0.002 % THD+N; 80 W, 4 Ω, 0.004 % THD+N

So I assumed such that gap in measurements.
If that was not true then the measurements match the sound quality.
As I said my mind is open, but unfortunately I can no longer measure the 845 amp since it's gone.
 
Do you really know that?...
His knowing or not means nothing. He has repeatedly mentioned less than zero interest in understanding the reason of reported anecdote. Eduardo de Lima showed that a correctly phased SE amp reduce overall sonic distortion, there was a good chance of that happening with the interaction of an SE 845 with OB11. But I respect Andrea's choice to stay ignorant by giving no attention to any reasonable cause. Ignorance is bliss after all.
 
I have not measured these specific amps, but usually a 845 SET amp without NFB has high THD [...] while a LM3886 based amp usually has much less THD

Yeah, at a single point in the state space which is rarely reached.

I measured the current and voltage on my speakers while playing. While there are nice peaks, especially with music that has not been collateral damage to the loudness wars, the vast majority of the time, it's way below 1W.

"It has more microdetail and ambient cues" the audiophile says, "and you can hear the room reverberation"...

This is about a very small signal, way below 1W, perhaps riding on some larger bass signal, but the 45W measurement says nothing about it...

My point is, if you're talking about "sound quality versus measurements" you should look at measurements done in conditions that replicate the operating point and state space variables that occur while playing real music. And that's mostly below 1W, with non-static signals, and some peaks.

45W THD is the absolute opposite of all this: power is way too high, so THD divided by signal can either be artificially low due to having too much signal compared to real listening, or artificially high due to forcing the tube amp to output a power that it only has to output on peaks.

Here's my theory:

A SE tube, or anything that works in pure class-A without feedback, has one operating mode, with a smooth transfer function, so its THD will fall smoothly with as output level goes down. It looks bad because it gets 5% at 25W, but the measurement was not done where it matters, in the first watt. If it was measured at 0.1W-1W it would get much better numbers. This correlates with my experience of tube amps, which sounded very good at low levels, but sucked when cranked up.

A class-AB output stage, however, behaves differently. There is one crossover if it is biased "blamelessly", and two if it is overbiased. And if it's a CFP or one half of a CFP like LM3886, the crossover is very ugly. But it always has a small region near zero output current where it is very clean.

So, basically, you have two values of output current, more or less symmetric around zero, where distortion rises dramatically, but away from these points it's clean.

When the usual "THD versus output level" is done at high power, THD appears low because the signal is too high, it spends very little time in the crossover, and harmonics, which depend on slew rate, are too high for the measurement bandwidth. When it is done at low power, it stays within the small class-A zone where the output stage is clean, so it is again blind. When it is done at a level where the peaks of the test signal enter the interesting crossover zone, again it is blind because near the peaks of the sine wave is where slew rate is lowest, so the crossover generates lower harmonics, which feedback handles.

That's why I think a much better test would be to replicate a low-amplitude signal (say, room reverberation and microdetail) added to a few amps low frequency signal (ie, bass). Then plot the THD and amplitude/phase variations of the "detail" signal according to the "bass" signal. The offset puts the zero zone of the test signal, where its slew rate is highest, right where it needs to be. I've tried it on a few amps and the results are quite fun to look at.

The boards for the test setup have been shipped, so I'll also do it on a LM3886 in maybe a month.
 
Basically, THD measurements centered on zero are like going in blindfolded and putting your hand on the wall of this house, and concluding it's fine without looking at the whole of it.

495763f88478cc6dbf674f1b6a20902f.jpg
 
Another thing I've done is THD vs time : play a 10kHz sine whose amplitude changes from about 0.1W to 10W every 10 seconds, and plot THD on it versus time. This is very easy to do, just simple math, no FFT. It gets expected results, which is the bias wiggles as the output stage heats and cools, and the THD goes on a rollercoaster.

So if you do a 45W THD test, you'll get a different result depending on where you look from the beginning of the waveform, as heating changes bias. Of course no-one cares, everyone measures with static power signals and lets the amp "settle" so results are meaningless.
 
...Or at least would be honest and say he was doing it for lolz and there is no audible improvement.


That would be untruthful. There is an audible difference.


Regarding another issue raised in this thread, its also true that grille cloths are removed from studio monitors because its presence affects the sound, at least it does in the near field.
 
Last edited:
His knowing or not means nothing. He has repeatedly mentioned less than zero interest in understanding the reason of reported anecdote. Eduardo de Lima showed that a correctly phased SE amp reduce overall sonic distortion, there was a good chance of that happening with the interaction of an SE 845 with OB11. But I respect Andrea's choice to stay ignorant by giving no attention to any reasonable cause. Ignorance is bliss after all.

As I said I have no longer the ability to measure the system because it's gone.

BTW it's true, I don't care about the reasons of the result.

A friend (one of the listening group) did ask me to build an audio system since she likes music.
We have been colleagues for nearly twenty years so I accepted to do.
I have built the system in my own way (fortunately I'm still free to do that although I cannot disagree) and she was very satisfied of what she heard.
The system is gone. Period.

Now I prefer to spend my little free time to work on my current designs rather than understand the reason why the 845 amp was more realistic than the other amp.
But I would be very happy if someone found the reasons.
 
As I've shown with my simple example, it is easy to build a circuit which has very low THD at the DC operating point (Vout=0V, Iout=0A) but 100x more THD at a small Iout offset from that. And a zero-centered THD test will completely miss it.

I would say that is a pathologically bad amplifier. The 845 amp does have more distortion than the 3886 especially the popular composite amp configuration, the benefit of massive feedback you know. Amps with poor thermal design are also bad designs.

You could post an actual plot of a generally accepted as good SS amplifier that exhibits 100X THD with "small" Iout offset. Or maybe you could show me how you measure THD into 8 Ohms at 10W with no deviation of Iout from 0.
 
Another thing I've done is THD vs time : play a 10kHz sine whose amplitude changes from about 0.1W to 10W every 10 seconds, and plot THD on it versus time. This is very easy to do, just simple math, no FFT. It gets expected results, which is the bias wiggles as the output stage heats and cools, and the THD goes on a rollercoaster.

So if you do a 45W THD test, you'll get a different result depending on where you look from the beginning of the waveform, as heating changes bias. Of course no-one cares, everyone measures with static power signals and lets the amp "settle" so results are meaningless.

Do you mean the problem is that the measurements are time invariant?
 
Status
Not open for further replies.