Incorporating Burst Power and Slew Rate Tests for Amplifier Reviews

[Cbdb2]

Maybe that was my little experiment:

I didn't see how "edges" could be found in a swept sine wave. So I created a single sample "impulse" and a 10Hz sqare wave and played them through the speakers and recorded the auible result with a UMIK-1

Then, used REW to generate its usual swept sine and had it calculte the impulse and step response.

To my amusement, the "actual" and "calculated" results matched quite closely, much more closely than i would have imagined.



Impulse response calculated from a ten second 10-24kHz sweep tone sent through the speakers in REW:

View attachment 418215

Single Full Scale Bit sent through speakers, playback in room recorded in Audacity:

View attachment 418216

Below:

Step response calculated by REW from a swept sin test tone
And "step" response recorded through the speakers playing a 10hz square wave, zoomed in on the rising "edge" and "flat top" of one cycle of that wave.

View attachment 418217

How the math pulls edges out of a smooth sweep is way over my pay scale, but it obviously does.

And thats in a room. Electronics will be even closer.

 

[Mikig]

Thanks to all for the discussion. The point, regardless of being included in the tests by Amirm is certainly interesting. It is a topic that is often read between the lines and this was a good opportunity to learn some concepts. I think that these data can be interesting but as always taken alone they say and do not say: in the end what matters in my very modest opinion is to understand how these numbers, in the real world, can affect the compatibility between amplification and speakers to improve the listening experience.

 

[Titurel]

So you mean like music is not a stable reference test because it changes from instantaneous value to another that hitting notes is not accurate too due to reference notes used by musicians varies.

Not really. My argument was that a test suite, based on a naive understanding of a single musical parameter (in this case pitch), would probably not yield any new, useful data. But that's not to say that music might not be a source of useful test signals, only that their application and analysis would be non-trivial.

 

[Doodski]

Not really. My argument was that a test suite, based on a naive understanding of a single musical parameter (in this case pitch), would probably not yield any new, useful data. But that's not to say that music might not be a source of useful test signals, only that their application and analysis would be non-trivial.

IC. Makes sense. I figure that music as a reference value is too random and that makes it unusable for tests and reference standard metering. Non-trivial is a understatement. I used music as a final listening test after a soak/burn-in but other than that it was not performant and the music was just to use the component for hours and make sure all was well over a span of time.

 

[Holmz]

I think if you secretly added an allpass filter to an amp you could have flat frequency response and wacky phase response, which could "blur" the sound audibly if it was extreme. I am not aware that this kind of thing happens by accident to any real extent though.

It happens all the time in many crossovers.

Which is very easy to "see". Look at the wave forms. Youll never see a transient thats straight up, they always take at least a few samples. Than look at gun shots and explosions, same thing.

I am not sure that is 100% correct.
The highest frequency would be more “straight up”, then the one 1/2 that would be delayed, etc.
At the lower east frequency (DC) we have the extra gas that is created and there is radial shift outwards of the air.

Even a nuclear blast shows the paint gassing out, before the wind arrives, and then it also snaps backwards after the initial positive pressure.
There is a lot happening.

However most of it has little to do with an amplifier.
Back in the day I had an amp that was advertised as 100 V/us, and the preamp that went with it was limited to 50 V/us.
These number at a lot hight than the 0.63 V/us that was quoted back on page 2 (I think).

 

[kemmler3D]

It happens all the time in many crossovers

Totally, but if the amp doesn't have one I don't know that testing for that would tend to show anything. Maybe bears mentioning that the negative effects of phase distortion per se (ignoring any effects on interference between divers) are often overestimated.

 

[Holmz]

Totally, but if the amp doesn't have one I don't know that testing for that would tend to show anything. Maybe bears mentioning that the negative effects of phase distortion per se (ignoring any effects on interference between divers) are often overestimated.

Yeah - but we almost never look at the phase with FFTs, mostly we just look at the amplitude.

I suspect that there is probably something to be said for quantifying the performance of an amplifier under clipping.
I am not sure how pervasive it is, and with a lot of music compressed to hell and back, then many amps will probably not be clipping - or at least not as much.

But with music that is not compressed, then the amp(s) could be driven into clipping a lot more than one might imagine.
For instance take something with 16dB of crest.
I think that that is 40x above the RMS.
So if we have 3W RMS, then the peaks are at 120W.
(I might have mixed up power and voltage, which I usually do) but if the music is NOT compressed and the system is cranked up, then it gets easy to have the amp(s) running into clipping a lot more than on might imagine.

Compound that with low sensitivity and 2 and 4 ohm speakers, and one can quickly find that a 100W started to sound different with classical than with pop music.

 

[kemmler3D]

Yeah - but we almost never look at the phase with FFTs, mostly we just look at the amplitude.

I suspect that there is probably something to be said for quantifying the performance of an amplifier under clipping.
I am not sure how pervasive it is, and with a lot of music compressed to hell and back, then many amps will probably not be clipping - or at least not as much.

But with music that is not compressed, then the amp(s) could be driven into clipping a lot more than one might imagine.
For instance take something with 16dB of crest.
I think that that is 40x above the RMS.
So if we have 3W RMS, then the peaks are at 120W.
(I might have mixed up power and voltage, which I usually do) but if the music is NOT compressed and the system is cranked up, then it gets easy to have the amp(s) running into clipping a lot more than on might imagine.

Compound that with low sensitivity and 2 and 4 ohm speakers, and one can quickly find that a 100W started to sound different with classical than with pop music.

No disagreement here. I think 100w is perfectly good for most listening, but if you want to sit far away or go really loud (even just for quick peaks) the amp requirements to avoid clipping are probably tougher than most people realize. Most small speakers have their own nonlinearities at high output too.

 

[Holmz]

No disagreement here. I think 100w is perfectly good for most listening, but if you want to sit far away or go really loud (even just for quick peaks) the amp requirements to avoid clipping are probably tougher than most people realize. Most small speakers have their own nonlinearities at high output too.

Yeah both larger for the transients, and much lower for RMS than most people want to believe.
There is a reason why people like those 3-5W SET amps.
At least the distortion profile is nice.

Once one is clipping like mad, then having an amp with the right distortion profile becomes handy.

One either needs it to “not clip”, or clip right… but the testing does not really work in that domain when it is on the rail… or on the corner case..

 

[witwald]

An amplifier with insufficient slew rate can introduce audible smearing, distortion, or loss of detail in fast, dynamic passages, even if it performs well under continuous power tests.

@AdamFrandsen Would you please provide examples of such amplifiers and a suitable test signal that makes the effects of "insufficient slew rate" audible.

 

[antcollinet]

Here's the Farina paper explaining it. Buckle up.

I would have to brush up on my maths a bit lot to follow that.

 

[AdamFrandsen]

@AdamFrandsen Would you please provide examples of such amplifiers and a suitable test signal that makes the effects of "insufficient slew rate" audible.

No

 

[AdamFrandsen]

We care about what reaches your ears, not your eyes. For the latter you, decide what matters to you. Don't ask us to evaluate beauty the same as you...

What's wrong with asking?

 

[Multicore]

A square wave will show it up unless it's pushed a few decades away.

Yes. Or a group delay graph. I guess we don't have those is because they don't often reveal anything interesting because our assumption is good enough, i.e. that with amps, magnitude spectrum is really where it's at.

 

[Multicore]

We care about what reaches your ears, not your eyes. For the latter you, decide what matters to you. Don't ask us to evaluate beauty the same as you...

What's wrong with asking?

Not very much. But the consensus on this forum is pretty clear: we try to identify and characterize the objective and the rest, the subjective, is up to you. Identifying this hand off is one of the goals. Amir does include some personal subjective remarks and evaluations in his reviews but these are clearly separated from the objective measurements.

So if you were to ask us to evaluate beauty the same as you do, the answer will be no. We are each responsible for our own evaluations of beauty. ("You are free, therefore choose...") Our belief in the separation of the scientific image from the artistic (subjectively manifest) image is fundamental. So in a manner of speaking, and depending who is asking whom, their relative sensitivities, the alignment of the stars etc., it might be better to not ask.

 

[AdamFrandsen]

I'm not sure what notes you're talking about. If you mean things representing points on the musical gamut, their actual number (and absolute frequencies) are not knowable, The 124 notes for the Berlin Phil. will not be the same 124 notes for the New York Phil. et al. And, if New York Phil were to play a piano concerto and a violin concerto in the same program, the turning would not be the same after intermission. And which set of 124 notes should stand in for a gamalon or for an ensemble playing Carnatic music? The audible spectrum is not granular and pitch is not fixed.

Standard tuning: I listed all the frequencies already. The point is not to be so precise that you have to include small variations in tuning, but that, in most modern music, there are only 124 notes within the audible spectrum - not thousands - so you could use them to test in a way that is more closely related to real music signals, even if it is not always the precise tuning. Also, lowering the signals to, say, -20 dBFS is doable, no? Finding out when the amp clips due to complex signals is valuable information. Additionally, one could use logarithmic scaling as well as zoomed-in frequency ranges to make it fit visually.

 

[NTK]

What's wrong with asking?

You do realize that you are asking Amir to do more work, do you?

 

[AdamFrandsen]

You do realize that you are asking Amir to do more work, do you?

Yes, that part is clear. Could be fun work though...

 

[antcollinet]

Finding out when the amp clips due to complex signals is valuable information.

Except that is not what happens. The amp is not clipping - why would it unless the peak input waveform x gain was greater than the maximum output voltage.

What must be avoided is the derived waveform from all the individual tones (out of the analyser) clipping (>0dBFS) due to all the individual tones periodically lining up so that they are all at the peak at the same time. So if you have 124 tones, then you have to reduce them all in amplitued by FS/124 (about -42dB) so when they all stack up it comes to less than 0dB FS.

 

[SIY]

Yes. Or a group delay graph. I guess we don't have those is because they don't often reveal anything interesting because our assumption is good enough, i.e. that with amps, magnitude spectrum is really where it's at.

Unless you deliberately add some fairly complex circuitry and do some heroic tuning of the feedback and compensation, an amp will be minimum phase. So indeed, the magnitude curve is sufficient because it will determine the phase curve.