AB(x) testing of Power Amplifiers

[Matias]

So where was any of this demonstrated? Cite?

I am no expert in this but see my reasoning below.

First we see that output impedance (damping factor) changes frequency response in real loads (actual speakers). Here are 2 extreme opposite examples demonstrating this:

SUMMER MUSINGS: No... Not all amplifiers sound the same (but many do! ;-).

A blog for audiophiles about more objective topics. Measurements of audio gear. Reasonable, realistic, no snakeoil assessment of sound, and equipment.

archimago.blogspot.com

Then we know that different amplifiers have different damping factors, intentional by design ("tuned to taste") or not. And I mean not only the highest number, but also the curve, in which frequencies it rises and falls.

Examples:

Mola Mola Kaluga

Parasound Halo A23

Hegel H590

Manley Mahi

My conclusions:
1. Amplifiers do sound different, and this is both measurable and noticeable.

2. These differences in how they sound are many times chosen ("voiced") by the designer through output impedance and/or distortion profiles.

3. If you mix different speakers (and their different loads) and different amplifiers (and their damping factors) the results are all over the place. Which explains why subjective audiophiles, who know nothing of measurements, do "mix and match" their speakers and amps to taste.

Are there flaws in my reasoning? Please let me know.

 

[SIY]

I am no expert in this but see my reasoning below.

First we see that output impedance (damping factor) changes frequency response in real loads (actual speakers). Here are 2 extreme opposite examples demonstrating this:

SUMMER MUSINGS: No... Not all amplifiers sound the same (but many do! ;-).

A blog for audiophiles about more objective topics. Measurements of audio gear. Reasonable, realistic, no snakeoil assessment of sound, and equipment.

archimago.blogspot.com

Then we know that different amplifiers have different damping factors, intentional by design ("tuned to taste") or not. And I mean not only the highest number, but also the curve, in which frequencies it rises and falls.

My conclusions:
1. Amplifiers do sound different, and this is both measurable and noticeable.

2. These differences in how they sound are many times chosen ("voiced") by the designer through output impedance and/or distortion profiles.

3. If you mix different speakers (and their different loads) and different amplifiers (and their damping factors) the results are all over the place. Which explains why subjective audiophiles, who know nothing of measurements, do "mix and match" their speakers and amps to taste.

Are there flaws in my reasoning? Please let me know.

1. No one said anything contrary to that. If you have a high source impedance, there will be deviations in the amp-speaker frequency response. Of course, that is not characteristic of the amp alone, since the deviations will be different for every different speaker. However, it takes a relatively high output impedance before it becomes significant. In the examples you cited, all but the tube amp showed no particular deviation with different loads (why don't they use a reactive speaker load like, cough, cough, some of us do?) beyond a few tenths of a dB of rolloff in the extreme treble.
2. Since the "voice" will be different with different speakers, this is dubious at best, other than marketing claims. And attributing it to brand names as the other poster did is ludicrous. Ditto distortion- it takes a LOT before it's audible, humans are far more sensitive to frequency response changes..
3. More likely that hobbyist "audiophiles" are doing this randomly, based on crap they read in magazines or the internet rather than actual controlled validation.

 

[Matias]

1. In the examples you cited, all but the tube amp showed no particular deviation with different loads (why don't they use a reactive speaker load like, cough, cough, some of us do?) beyond a few tenths of a dB of rolloff in the extreme treble.

I agree, which is why evaluating frequency response in fixed resistance loads is useless. The changes appear only with reactive loads, like speakers. And as per the graph in Archimago's graph, there is a 5 dB difference in 100 Hz bass region between both amps... this is huge!

I am genuinely curious how the difference would be between less extreme cases, say 2 solid state amplifiers. @pma already showed 2 amps here, but would be interesting to see other combinations of amps and speakers.

2. Since the "voice" will be different with different speakers, this is dubious at best, other than marketing claims. And attributing it to brand names as the other poster did is ludicrous.

Which is why you see some companies teaming up on audio shows and show rooms, either in house (B&W + Classé, Sonus Faber + Audio Research) or different companies (KEF + Hegel). It's like they are saying "Hear this combo! This we have tested and approved" and influences the customers.

 

[dominikz]

I agree, which is why evaluating frequency response in fixed resistance loads is useless. The changes appear only with reactive loads, like speakers. And as per the graph in Archimago's graph, there is a 5 dB difference in 100 Hz bass region between both amps... this is huge!

I am genuinely curious how the difference would be between less extreme cases, say 2 solid state amplifiers. @pma already showed 2 amps here, but would be interesting to see other combinations of amps and speakers.

But these (FR) deviations are in my understanding completely described by the voltage division caused by the amplifier's output impedance and the loudspeaker's complex impedance. I.e., if we have the amplifier's output impedance plotted vs frequency, we can predict the amp FR for any known loudspeaker load (i.e. known loudspeaker impedance vs frequency plot).

So testing amplifiers with pure resistive loads is IMHO not an issue as long as we also measure the output impedance vs frequency diagram for the amplifier (I'd personally love to see that plot added to @amirm's amplifier reviews!) - that would explain in details how sensitive the amp is to load impedance variance (at least as far as FR is concerned).

A huge benefit of resistive loads when testing amps is that it makes it easy to compare measurements done by different sources. Using complex loads would require some standard load (which is unlikely to happen), and we could no longer compare to older measurements that were typically done with 4 and/or 8 ohm resistive loads.

In practice however, even in pretty extreme cases (see example here) the impact of amplifier's output impedance IMHO seems relatively insignificant if one has EQ capability - magnitude of variations even in bad cases is often low, and most of it would usually be in low frequencies (where anyway room EQ is normally needed and applied). In some cases there are audible deviations in the high frequencies as well, but seems EQ shelving should be able to address those easily.

 

[ZolaIII]

When you put a kid in front of the new shiny toy you can't expect it won't play with it nor will you try (I hope) to stop it. Similarly when you put someone who likes to listen in front of the new audio system no matter what you say to him (pure facts or epic fantasy stories) he will try to adapt it to his likings (absolutely doesn't matter will he be successful in that or not). Simply that's how we are. One doesn't fit all no matter if it's objective or subjective. If we can more precise determine what influences subjective we can make a better correlation to objective (to extend of future classification).

 

[ahofer]

Just found this device on Amazon, which should provide some visual excitement to your AB testing:

https://www.google.com/url?sa=t&rct=j&q=&esrc=s&source=web&cd=&ved=2ahUKEwjf3szj3qn1AhUxkYkEHe8BAvAQFnoECAQQAQ&url=https%3A%2F%2Fwww.amazon.com%2FAmplifier-Speaker-Audio-Switcher-Display%2Fdp%2FB096KLDMCV&usg=AOvVaw2dalu5LgIq96iTp3l-ln2e

 

[bethslave]

I appreciate all of the explanations and discussion here, I think it has solved a mystery for me. I've been a hifi enthusiast for a very long time now. I grew up reading Julian Hirsch's reviews and appreciating his methodology . I have always believed that all competently designed and manufactured amps, when operated within their design parameters, should sound about the same. Most stereo amplifiers/receivers I've owned seemed to sound about the same. I also freely admit the equipment I'm about to list is lower end to many of you, but still should be competent. The quick version is I still have vinyl that isn't available in digital formats and so I still use a turntable. For my bedroom system, the phono preamp stopped working on a Denon DRA-345R I was using. I considered specs and reviews and purchased an Onkyo A-9110 to replace the Denon. I was surprised when I heard a difference and switched back and forth a few times to try and understand what I thought I was hearing. It all came down to bass which is noticeably improved with the Onkyo [no eq, tone controls on both in 'flat' position]. I believe I now understand the following to be true:

1] The Denon is not operating within its design parameters as is evidenced by the phono preamp failure
2] The Onkyo is better at properly driving my 6-Ohm speakers, which are Advent Legacy IIIs

My experience was similar when my Onkyo failed [within warranty] and I put an older HK AVR-146 in its place while it was out for repair. I don't put much faith in audio memory but the HK can't seem to drive the Advents on the lower end like the Onkyo can, though in every other way it sounded fine. The quick summary is to agree that speaker loads do seem to react differently with different amps and that this can be both audible and measured. Interesting stuff, I believe I've learned something.

 

[Chr1]

The more I read about this, the more I am starting to think that mild clipping is actually fairly common... and possibly that the additional complexity within AVRs may be causing a degree of signal interference/distortion.