Can you hear a difference in Sound between Audio Amplifiers?

[RichB]

There is plenty of data showing that SS amps at very low power are not linear. It is often stated that for a double-blind test (DBT) to be valid, the amps must be level matched to .1 DB. Where that stat comes from I don't know.

Stereophile measurements usually include a 2.83V sweep into a simulated speaker load (reactive) and at 8, 4, and 2 ohms into purely resistive loads.

Here are the Stereophile measurement of the Benchmark AHB2:

http://www.stereophile.com/content/benchmark-media-systems-ahb2-power-amplifier-measurements#q5R6qULWg8zEDz8k.97

The output impedance, including a 6' speaker cable, was a low 0.09 ohm at 20Hz and 1kHz, rising slightly to 0.22 ohm at 20kHz. As a result, the modification of the Benchmark amplifier's frequency response due to the interaction between this impedance and that of our standard simulated loudspeaker was just ±0.1dB (fig.1, gray trace).
Read more at http://www.stereophile.com/content/benchmark-media-systems-ahb2-power-amplifier-measurements#bOrgkp4iTPAW3VbZ.99

+/- .1 DB is certainly not something to be concerned about. However, even at 4 ohms there is a .2 DB drop with a purely resistive load. The majority of amps measured at Stereophile perform similarly, even very expensive amps. There are a some I have seem that are truly load invariant in this measurement.

Increase the power demand, add in a highly reactive and low impedance speaker load, and I suspect the variance from linearity increases. Many SS amps have excellent power reserves. I have not seen an AVR amp put to this test or a similar test. Generally, AVRs are held to different standards in reviews.

So, amps under load do not measure the same. In fact, they can measure different enough to be outside the "recommended" tolerance of a DBT.

- Rich

 

[ski2xblack]

Well, I haven't seen any amps with output impedance of zero, so some interaction is inevitable. That's just the reality of things. But you have to go pretty far off the reservation for output impedance to have audible effects. The graph you posted shows ~1/8 of a db deviation from linearity. Your speakers' acoustic output introduces an order of magnitude greater deviations than that, as does the acoustic environment in which they operate, adding still more. Those inconsequential electrical deviations resulting from amp/speaker interaction are nothing but a fart in a hurricane.

 

[mtrycrafts]

There is plenty of data showing that SS amps at very low power are not linear. It is often stated that for a double-blind test (DBT) to be valid, the amps must be level matched to .1 DB. Where that stat comes from I don't know.

It comes from respected researchers with special test tones, not music, to get threshold detections as some subjects under ideal conditions with the right signals can detect .2dB differences. So, you want to match below that to eliminate level caused audible differences.


Stereophile measurements usually include a 2.83V sweep into a simulated speaker load (reactive) and at 8, 4, and 2 ohms into purely resistive loads.

...

+/- .1 DB is certainly not something to be concerned about. However, even at 4 ohms there is a .2 DB drop with a purely resistive load. The majority of amps measured at Stereophile perform similarly, even very expensive amps. There are a some I have seem that are truly load invariant in this measurement.

That may be so. Then it is not a good match for the difficult load.
But, when you change from the special conditions of test tones to music and your living room, the numbers are much higher to differentiate. Still, one should level match to a very low level, preferably .1dB if possible

Increase the power demand, add in a highly reactive and low impedance speaker load, and I suspect the variance from linearity increases. Many SS amps have excellent power reserves. I have not seen an AVR amp put to this test or a similar test. Generally, AVRs are held to different standards in reviews.

So, amps under load do not measure the same. In fact, they can measure different enough to be outside the "recommended" tolerance of a DBT.

- Rich

Well, be careful here. Amps have their limits. Some can handle difficult loads and others cannot.
Also, you would be level matching the amp/avr to that speaker, don't forget and try to differentiate with your favorite music. The measured differences may still not matter unless you are exceeding the lower
quality amp's capability.

 

[RichB]

Well, be careful here. Amps have their limits. Some can handle difficult loads and others cannot.
Also, you would be level matching the amp/avr to that speaker, don't forget and try to differentiate with your favorite music. The measured differences may still not matter unless you are exceeding the lower
quality amp's capability.

The speaker changes the amps performance. Is that a fart in a storm? Sure, sometimes.

The posted measurements show that under controlled loads, it can see reactive and resistive load changes have an effect on linearity; The speaker can change the amplifier performance. Some more than others. The example is simplistic that does not represent all loaded or source material and does not address audibility with content.

I don't believe the current set of measurements to rule out the possibility that amps can sound differently. Of course, if you compare two amplifiers with similar non-linear and distortion behavior, then they will be indistinguishable

- Rich

 

[ski2xblack]

Any amp with higher than zero output impedance, which is all of them, will exhibit such variability into a load. That's not a surprise at all, merely a trivial fact of life when it comes to amps.

 

[RichB]

Any amp with higher than zero output impedance, which is all of them, will exhibit such variability into a load. That's not a surprise at all. It's what the math tells us and what the measurements confirm.

Perhaps, I am reading this wrong but the ML N33 does not appears to have almost no variability.

The measured output impedance was extremely low at 0.02 ohms at 20Hz and 1kHz, rising slightly to 0.026 ohms at 20kHz. This is an amplifier that won't let itself be bossed around by any loudspeaker out there!
This can be seen in its small-signal frequency-response plot (fig.1). Even driving our dummy loudspeaker load, the change in measured response is negligible.
Read more at http://www.stereophile.com/content/mark-levinson-no33h-monoblock-power-amplifier-measurements#M9eT7rdQEK3kZ54y.99

Most amps also have variability associated with resistance at 2.83 volts. Amps are electrical devices so these measurements are valid.

As you say, amp/speaker interaction between quality amps are likely dwarfed by the room.

These measurements likely may show the maximum deviation.
Therefore, hearing the difference between say a quality amp and AVR even at modest levels is not far fetched.

- Rich

 

[ski2xblack]

Rich, does that second graph include response into a complex load like the first?

 

[RichB]

Rich, does that second graph include response into a complex load like the first?

Unfortunately, they are different scales.

Sterephile states "Even driving our dummy loudspeaker load, the change in measured response is negligible." which is referred to elsewhere as their simulated speaker load.

The graph also includes 8 and 4 ohms measurements with little or no deviation.

- Rich

 

[PENG]

Rich, it is not difficult to do those same plots using a variable complex load to simulate the frequency dependent impedance of real speakers. I suspect labs such as Stereophile might have tried that and perhaps found nothing significant and therefore don't bother doing it all the time.

I don't believe complex load presented by the average/typical speakers would make things worse. The minimum impedance of a passive speaker is typically the d.c. resistance and is therefore the lowest value. Complex load could obviously present phase angles that AVRs may not handle very well but that's more to do with distortions and thermal issues rather than linearity that you are talking about.

Can you write John Atkinson and ask him why don't they use variable complex loads?

 

[3db]

Are we still ??

 

[mtrycrafts]

Are we still ??

...

Sure. Much easier to capture a dead horse than a live one.

 

[jinjuku]

Can't believe this thread is still going...

I've noticed differences between amps, but only in cases where the cardinal rules are violated (low output impedance, low distortion, linear within freq range, operated within limits). Performed level matched a/b test using a Pass Labs (XA30) vs. more modestly priced amps: unable to tell the difference.

I think the transient peaks of the leading edges of notes require more dynamic power, and that clipping occurs far more often, than people realize. I suspect that this accounts for many of the perceived differences when folks go from an AVR to a dedicated amp with more dynamic power reserves and/or improved low impedance drive capability. Unclipped power has a more open, effortless sound, and is more capable of creating a convincing facsimile of the real thing. An AVR may be taxed trying to replicate those leading edge transients, even at modest levels, given the average speaker sensitivity and typical listening distances. It won't come across as gross clipping, because it's not, but if it's happening enough, it will color the sound.

And a lot is due to folks fooling themselves, expectation bias in sighted evaluations. What did Feynman say, the easiest one to fool is oneself? A few measurements, or even calculations based on reasonable input, would dispel the myths.

Folks who explore qualitative differences between amps are missing the boat. Amps are mature technology. Those with low output impedance, low distortion, and linear response are dime a dozen, making power the primary concern.* The folks seeking a qualitative difference need eq, or new speakers, but not amps with different sonic signatures. (Such amps exist, by design, but they largely fail at meeting the basic definition of an amplifier. They are more correctly referred to as processors with voltage gain. See tubes, First Watt amps for examples.)

*The speaker/amp interaction is complex and may account for differences, but such differences would fall under situations where the amp's clean power limits are exceeded.

I have had people try to hammer me for pointing out the amps are commodity products now. The circuits, the buffers, the feed-back designs etc... have solved all the problems needing solved.

I also agree amp manufactures can design an amp to sound different and then market that difference.

There is a reason a series like the Crown XLS DriveCores (I and II) are popular in the home crowd. Some poor bastard over at Polk Audio replaced his Threshold with a $399 Crown to run his modded SDA TL's and they crucified the him.

Stupidity is ignorance left willfully uncorrected.

 

[jp_over]

I have had people try to hammer me for pointing out the amps are commodity products now. The circuits, the buffers, the feed-back designs etc... have solved all the problems needing solved.

I also agree amp manufactures can design an amp to sound different and then market that difference.

There is a reason a series like the Crown XLS DriveCores (I and II) are popular in the home crowd. Some poor bastard over at Polk Audio replaced his Threshold with a $399 Crown to run his modded SDA TL's and they crucified the him.

Stupidity is ignorance left willfully uncorrected.

In short, well said!

It is amusing how often folks will get beat up (like the example you gave) by other "forum experts". Audioholics may have its ills, but it's always my go-to for solid information from folks who eschew popular audio mythology.

My only issue with (most) pro amps is that there isn't a 12v trigger (just saw the XLS has this) and aesthetics. I once had 2 x Yamaha P2500S amps that were dead silent when not playing and sonically identical (subjective, not measured) to my ATI 1803 (another solid performer but heavy as a beast!).

 

[3db]

Unless one overdrives their amp, there will be no sound differences. Yes there are subjective claims that differences exist but these always disappear under proper test condifitons where the tester does not know what amp he's hearing.

 

[Swerd]

There is one and only one way to determine if two different amplifiers make an audible difference while driving the same speakers. That is to test them with human listeners. Anything else is speculation.

I do want to praise RichB for learning how to talk about this without pressing our buttons nearly as much . The lab bench measurements of amplifiers are just that, lab measurements. They cannot substitute for a blind listening test done with enough listeners to achieve statistical significance.

 

[RichB]

Unless one overdrives their amp, there will be no sound differences. Yes there are subjective claims that differences exist but these always disappear under proper test condifitons where the tester does not know what amp he's hearing.

I hooked up a 100 Yamaha to my Salon2's. This was not a good match.
I could not find the Yamaha display showing the clipping and distortion levels.

If you want to hear the effect of limited power on an AVR, switch it into 4 ohm mode. I'll bet you can tell the difference at surprisingly low listening levels.

- Rich

 

[RichB]

There is one and only one way to determine if two different amplifiers make an audible difference while driving the same speakers. That is to test them with human listeners. Anything else is speculation.

I do want to praise RichB for learning how to talk about this without pressing our buttons nearly as much . The lab bench measurements of amplifiers are just that, lab measurements. They cannot substitute for a blind listening test done with enough listeners to achieve statistical significance.

51% of the folks who voted on this thread have heard the difference between amps. So, I am sit comfortably among the unwashed masses.

Even with a properly empanelled trained listeners, you can have either finding distinguishable or not and yet, you would only prove that those listeners with that system in that room on that day found X. An amp may be indistinguishable driving one set of speakers may not be with another.

This is a good interview with Jon Iverson discussion blind testing and the nature of audio testing. Harman tests with trained listeners. Otherwise, you are testing the listeners and not the equipment.

https://twit.tv/shows/home-theater-geeks/episodes/221

- Rich

 

[mtrycrafts]

51% of the folks who voted on this thread have heard the difference between amps. ... Otherwise, you are testing the listeners and not the equipment....
- Rich

Well, 51% could be equivalent to flipping coins unless the trials are huge.

It is always testing the listener. Afterall a listener is the one who will buy audio components. And, I bet those who voted are not trained as the Harman listeners.

 

[mtrycrafts]

...

If you want to hear the effect of limited power on an AVR, switch it into 4 ohm mode. I'll bet you can tell the difference at surprisingly low listening levels.

- Rich

And why would you do that?The power is cut way down or limited.

 

[lovinthehd]

Just because a difference was noticed doesn't mean it was either for the good or the bad in relationship to the specs....just different which in and of itself is rather meaningless unless you can determine just what the difference was caused by.