Review and Measurements of Benchmark AHB2 Amp

[John_Siau]

The ABH2 works with European voltage, 220v, or a non-US version is needed?.

The AHB2 has an auto-sensing power supply and the amplifier is certified for sale in Europe and many other parts of the world. Unlike most amplifiers, the AHB2 has a regulated power supply. This means that the maximum output power is not reduced when the line voltage sags. It also means that you get full power when running from 100 V in Japan.

 

[John_Siau]

I admire this amplifier, but I have two questions:
1 Will it handle electrostats, and my Quad 2805s in particular?
2 Will the impressive measured performance translate into audible superioity compared to, say, a Quad QSP?

Yes we have customers that love the way the AHB2 works with the Quads.

At trade shows, Martin Logan demonstrates their speakers with AHB2 amplifiers running in mono mode.

Paul Seydor uses the AHB2 to drive his Quad 2805s. Here is what he had to say:

"What about the big stuff? Well, I don’t know much bigger stuff than the lowest organ notes on Kei Kioto’s Bach recital, and they’re not transients, either, but deep, long, sustained notes. No problem on the Quads or the Super HL5plus." https://www.theabsolutesound.com/articles/benchmark-ahb2-amplifier/?page=2

 

[John_Siau]

Other day, I extracted the following from a wikipedia article

The improved transient response, as well as the more linear phase response of DMM improve the overall stability and depth-of-field in the stereo image. In addition, disturbing adjacent groove print-through sounds (groove echoes) are reduced in DMM.

@John Siau

Do you agree?

Stereo depth is very dependent upon a precise phase matching between the right and left channels at all frequencies in the audio band. Differential (left/right) phase errors will blur the location left to right and especially front to back. Left to right positioning can be overridden by left to right amplitude differences. Depth can only be conveyed by accurately preserving the phase response in both channels all the way from a stereo microphone pair to the pair of loudspeakers. You also must be seated in the sweet spot to hear depth.

Many very fine studio recordings have absolutely no depth because the left/right stereo image was entirely created with pan pots. The left/right positions of the musical voices are entirely determined by the left/right amplitude differences. Some fine examples are all of the tracks on Steely Dan's "Two Against Nature". This CD is a great recording but it has absolutely no stereo depth when played on a system with an accurate phase response. Systems with a poor phase response may give these tracks a false sense of depth. This false sense of depth is easily identifiable because each musical voice will fill the entire depth instead of locating to a specific distance behind the plane of the speakers.

If you want a good explanation of depth and a good demonstration play tracks 4 and 5 of "The Ultimate Demonstration Disk" from Chesky Classics. In fact, the entire disk contains tracks that have wonderful stereo depth with pinpoint focus of location. These recordings were made with stereo microphone arrays and the stereo images were simply captured (not created with pan pots).

It is now possible to artificially create stereo recordings that have depth. DSP techniques can be used to place a musical voice anywhere in a 3-D space. This 3-D space can then be rendered by a pair of stereo speakers or by an array of speakers (Atmos). 3-D position is determined by phase and amplitude. Left/right phase errors will expand the size of each musical voice (blurring its location).

Whenever a musical voice is mixed to the phantom center of the stereo image, this image is easily blurred by differential phase errors. If the high frequencies seem to come from the speakers instead of from the phantom center image, this is an indication that there is a left/right differential phase error at high frequencies. I have some recordings that are essentially mono in stereo. In other words, the entire mix is dead center. Phase errors immediately destroy this phantom center image by moving some frequencies away from the center and toward the speakers. Some of the tracks on Eric Clapton's "Old Sock" CD are useful for this test.

 

[D700]

The left/right positions of the musical voices are entirely determined by the left/right amplitude differences. Some fine examples are all of the tracks on Steely Dan's "Two Against Nature". This CD is a great recording but it has absolutely no stereo depth when played on a system with an accurate phase response.

Well don't tell them that, Walter Becker will rise from the grave to record the whole damn thing another 27 times to get it right.

 

[John_Siau]

Martin Logan Neolith has 0.43 ohm at 20khz. In this case the damping factor will be 0.43/0.235=1.8 The capacitance of speakers cables is irrelevant indeed but what about speaker capacitance ? How AHB2 behave with speaker capacitance like 1uF or 2uF ?

Martin Logan uses the AHB2 to drive their speakers at trade shows. They work well together. Our use of feed-forward correction makes it much easier to drive capacitive loads. Capacitance can cause stability problems in a feedback system, but capacitive loading is not a problem for a feed-forward system.

 

[John_Siau]

Well don't tell them that, Walter Becker will rise from the grave to record the whole damn thing another 27 times to get it right.

He can rest in peace. It is a great CD.

 

[GrimSurfer]

It would be great to capture and consolidate John's posts into a wiki on differences in sound between amps. Bits of this may be commonly known, but the breadth of what John lays out here helps debunk the myth that all amps of the same power, when level matched, are indistinguishable. It also speaks to the importance of source material.

 

[RayDunzl]

It would be great to capture and consolidate John's posts into a wiki on differences in sound between amps.

Something like this?

https://benchmarkmedia.com/blogs/application_notes

 

[GrimSurfer]

Exactly, Ray, but posted on an independent site which people are more likely to trust/accept. (I worked my way thru those bench notes some time ago and found them to be excellent, even though some of the concepts were a little over my head.)

 

[Music1969]

Hi @John_Siau

I've seen a couple Stereophile measurements of GoldenEar Triton speakers (Triton One and Triton Reference) that say:

"As tends to be the case with a design using a passive high-pass filter with a fairly low corner frequency, the electrical phase angle becomes increasingly capacitive below that frequency; although the impedance magnitude rapidly increases below 100Hz, mitigating the effect of that phase angle, there is still a combination of 4.1 ohms and –50° at 100Hz"

and

"The magnitude does dip below 6 ohms in the midrange and high treble, with a minimum value of 3.4 ohms at 320Hz and a combination of 4 ohms and –45° phase angle at 73Hz."

Just using these speakers as an example only - I've seen other speaker measurements will similar comments.

My question isn't about the 4-ohms but this combination with these phase angles at low frequencies.

Do these phase angles of –50° at 100Hz and –45° phase angle at 73Hz at 4-ohms (just examples only), present a challenge for the AHB2?

If not, can you share info about how it's not an issue for the AHB2.

Thanks in advance

 

[Tks]

Just wanted to say thank you for the many insights @John_Siau ! Real big fan of what Benchmark has been doing and continues to do. One of the very few companies based on outward behavior I can throw my trust behind (scores points with having started off on the right foot with not indulging things like multi-thousand dollar "sonic enhancing" audiophile cables, and also being heavily measurement and science driven are two such facets that I appreciated when I got to know the company).

Also the blog on the Benchmark website is great for people like me learning about audio and science on a slightly more detailed, but easy to digest beginner level stuff.

This is definitely the power amp I want to put in my living room, the black looks great

 

[dkfan9]

Thanks. Your analysis, although extracted from disparate sources, fascinates me because it provides a potential explanation for what I hear as a difference between two excellent (but unequal) amplifiers with a single speaker. I'd be really happy if you based the re-analysis on the F228Be because that is where I heard it most clearly.

Here are the comparison charts for the F228Be. These are a little more precise. Since output impedance is not constant across frequency, I modelled the amplifiers in VirtuixCAD as muted speaker drivers in series with the F228Be. I traced the damping factor charts published by Soundstage Network here and here, exported as text, transformed the DF to impedance in LibreOffice Calc, re-exported as text, and uploaded the impedance response to drivers representing the A31 and AHB2. To represent the AHB2 in monoblock mode, I put two of them in series prior to the driver. I gathered one set of data points using the F228Be's measured frequency response, and another assuming a flat amplitude, both using its measured impedance. Both were placed around 80dB because that's the program's default driver SPL. The F228Be's FR and impedance measurements were taken from the Stereophile review. Here's what the model looked like:

Since this can be a sensitive subject, a few disclaimers before I begin: I don't claim these changes are or are not audible. They are, objectively, very small in level relative to many other pieces of an audio setup, particularly the room and speakers. If perceptible, they may still be subjectively outweighed by the amplifier's distortion and noise. On the other hand, they may also amplify or counteract issues in the room or speakers. If nothing else, they're interesting and seem to be in the spirit of this forum. The scales vary between some of the graphs, not to mislead, but to examine differences in maximum detail. I can't claim the sourced measurements are flawless, but they're the best I found, and the two amps are measured with the same methodology. I certainly can't claim my generated numbers are flawless. I know I'm short phase information, and to quote a guy from the last gasp of the age of irony, "there are also unknown unknowns." Finally, throughout this post, when I refer to the amplifiers, specifically or generically, I'm using it as shorthand for "the amplifier's output impedance".

Now, jumping in: The first two graphs show the "flat" F228Be's amplitude response through the amplifiers, with the second graph level matched at 1000Hz. Top to bottom here is roughly .6dB. REW was used for all display and most manipulation.

The A31 is louder across the audio band, but below the last two octaves, the difference between the A31 and monoblocked AHB2 is less than .2dB (ignoring any level increases brought by D+N), so perfect level matching might be a challenge. By 10kHz the difference is over .25 dB, reaching over .3 at 20k. Below 5k, even at this scale differences between the AHB2 in stereo and the A31 are minor, but the AHB2 in monoblock configuration does diverge a little more. In the upper midrange/lower treble region where the four lines run together on the bottom graph, the Revel is at its highest impedance above port tuning and the Benchmark's output impedance is still relatively low. At high frequencies the Benchmark's impedance rises quite a bit, and at low frequencies the Revel's drops below 4 ohms, so the amps diverge. Whether or not this is what he was hearing, these graphs do seem to match up with Kal's impressions on bass performance in his F228Be review.

Ok, next up are the differences when modelling using the speaker's measured response. The following three graphs show purely differences, comparing each amp to the others (including AHB2 in stereo vs. mono configuration), and are scaled even tighter, at ~.4dB between the top and bottom of the graph. 0dB on the graph is set to 1000Hz to better see differences in frequency balance, since the differences in level are pretty clearly seen above (and again below).

These graphs don't provide any info that can't be extracted from the first two graphs, but they do highlight two things: differences in output impedance are primarily significant when accompanied by swings in speaker impedance (as shown by John Atkinson in nearly every set of tube amp measurements into his mock speaker load), and the AHB2's frequency response in stereo is closer to the A31's than to itself in a monoblock configuration.

And for the final graph, we have the F228Be's measured FR on its own and attenuated by each amp. These are the same numbers that got us the three previous graphs. Around 6dB from top to bottom of the graph (compared to .4-.6 for the graphs not showing speaker response). Oh, and a bigger picture to help see the details.

And here we see why it's easy to discount an amp's significance to frequency response and overall sound quality. The largest deviation in any amp above was below half a dB. Here, we see the speaker's response swing by a couple of dB in the treble multiple times, and more than that lower in frequency (though with the limited far field measurement resolution and compromises of near field measurements leading to the upper bass bump, we can't be too sure what's real and what's a measurement artifact). And this is a great speaker.

This graph still has an interesting story to tell in the treble. Where the speaker starts shelving up, above 4k, the monoblock AHB2's higher output impedance flattens the speaker a bit, acting like an EQ circuit (pretty good THD+N for an EQ circuit). The same output impedance that causes almost half a dB in variance when "driving" the "flat" F228Be's in the first graph. To bring it full circle to Kal's review, the "slight bit of softening" he sensed above the bass in the A31 relative to the AHB2 may have come from the Parasound shelving down 1.5-4kHz relative to the treble a little more than the Benchmark. Or maybe the increased noise and distortion in the A31 caused the impression. And maybe it was the increased distortion and noise that gave the A31 more body down low. Maybe he imagined all of it. I don't know.

In the end, I'm pretty confident either of these amplifiers would outperform my Denon X2100W. I don't know about the audibility of low level distortion and small FR variances, but I do know I can hear noise, and in my former apartment home theater it was easy to hear hiss on silence from the couch, 6 feet from the TV. Not always, but any time I caught a glimpse, it would be days before I would forget and be able to tune it out again. But hey, I don't even have pre-outs, let alone professional balanced ones, so I guess I wouldn't put these amps to good use anyway...

 

[maty]

...In the end, I'm pretty confident either of these amplifiers would outperform my Denon X2100W. I don't know about the audibility of low level distortion and small FR variances, but I do know I can hear noise, and in my former apartment home theater it was easy to hear hiss on silence from the couch, 6 feet from the TV. Not always, but any time I caught a glimpse, it would be days before I would forget and be able to tune it out again. But hey, I don't even have pre-outs, let alone professional balanced ones, so I guess I wouldn't put these amps to good use anyway...

[Polish] https://audio.com.pl/testy/kino-domowe/amplitunery-av/2141-denon-avr-x2100w

to English: https://translate.google.com/transl...o-domowe/amplitunery-av/2141-denon-avr-x2100w

 

[LTig]

The amplifier will stay clean when driving very low impedances. The published rating is the long-term continuous output power and this is a function of heat sink area. Driving 3-Ohm, 2-Ohm or 1-Ohm impedance dips are not a problem. What is unique about the AHB2 is that is stays clean when driving these low impedances. There is virtually no rise in THD as load impedance drops. No other amplifier does this so well. Most amplifiers show a significant increase in THD when driving low impedances. In contrast, the feed-forward correction system in the AHB2 keeps the output distortion free.

This is unbelievable performance, never seen before - at least not by me.

Yes it is OK to test at 3, 2 and 1-Ohm loads. You cannot do damage to the AHB2 amplifier by very low load impedances or short circuits. It is fully protected and it will shut itself down before you can cause damage. You can even short it out while it is playing at full output.

And this is what pro equipment is all about: It can handle all kinds of use and abuse without breaking. Delivering such performance in a small and light weight enclosure is a first, I'd say.

 

[svart-hvitt]

TIME FOR SOME HUMOUR

(From The Audio Critic, number 23, 1995, page 5)

 

[LTig]

Stereo depth is very dependent upon a precise phase matching between the right and left channels at all frequencies in the audio band. Differential (left/right) phase errors will blur the location left to right and especially front to back. Left to right positioning can be overridden by left to right amplitude differences. Depth can only be conveyed by accurately preserving the phase response in both channels all the way from a stereo microphone pair to the pair of loudspeakers. You also must be seated in the sweet spot to hear depth.

This may explain why there are folks who prefer vinyl over CD. I once heard that vinyl playback does not guarantee stable phase matching between the stereo channels. With the explanation above you can expect a somewhat broader and/or deeper soundstage due to bad phase matching. Many people may prefer this even if it is not accurate.

Many very fine studio recordings have absolutely no depth because the left/right stereo image was entirely created with pan pots. The left/right positions of the musical voices are entirely determined by the left/right amplitude differences. Some fine examples are all of the tracks on Steely Dan's "Two Against Nature". This CD is a great recording but it has absolutely no stereo depth when played on a system with an accurate phase response. Systems with a poor phase response may give these tracks a false sense of depth. This false sense of depth is easily identifiable because each musical voice will fill the entire depth instead of locating to a specific distance behind the plane of the speakers.

This can easily demonstrated with monaural recordings. In a good playback system the sound must be pinpointed exactly in the middle of the line drawn between the speakers.

[..]Whenever a musical voice is mixed to the phantom center of the stereo image, this image is easily blurred by differential phase errors. If the high frequencies seem to come from the speakers instead of from the phantom center image, this is an indication that there is a left/right differential phase error at high frequencies. I have some recordings that are essentially mono in stereo. In other words, the entire mix is dead center. Phase errors immediately destroy this phantom center image by moving some frequencies away from the center and toward the speakers. Some of the tracks on Eric Clapton's "Old Sock" CD are useful for this test.

I once compared Mackie HR824 (the first version) with K&H O300D. Both are active studio monitors with wave guides. An old mono recording sounded wonderful on the Mackies, it filled the room and didn't sound much like mono. Switching to the O300D destroyed this impression immediately and the music seemed to come from a very narrow and smalll strip between the speakers.

I'm not sure whether the Mackies sounding bigger/broader/deeper results from phase differences between the internal amplification or from response differences between the drivers or both. When I heard these Mackies the first time they were positioned on a side board in a high and very lively room. They filled this room with enveloping sound and it took me a minute or two to realize that one speaker wasn't working at all. So even one single speaker is able to fill this kind room.

 

[DonH56]

Generally (very generally!) speaking large phase angles, and particularly negative (capacitive) angles, are more a problem for class-D amplifiers due to the required output filter (which is also needed by some class A/AB designs as well, natch). The main issue is stability, then thermal management to dissipate the heat whilst handling (high) power into a low-impedance load. An amp that handles sub-2-ohm ESLs is not likely to have an issue with 4 ohms and -45 degrees. The class-H design largely mitigates thermal issues.

Why do you think it might be an issue? I suspect John is not prepared to spend hours defending every possible speaker load, but what he has already provided is pretty substantial evidence that the AHB2 would not suffer from such a load.

 

[John_Siau]

Hi @John_Siau

I've seen a couple Stereophile measurements of GoldenEar Triton speakers (Triton One and Triton Reference) that say:

"As tends to be the case with a design using a passive high-pass filter with a fairly low corner frequency, the electrical phase angle becomes increasingly capacitive below that frequency; although the impedance magnitude rapidly increases below 100Hz, mitigating the effect of that phase angle, there is still a combination of 4.1 ohms and –50° at 100Hz"

and

"The magnitude does dip below 6 ohms in the midrange and high treble, with a minimum value of 3.4 ohms at 320Hz and a combination of 4 ohms and –45° phase angle at 73Hz."

Just using these speakers as an example only - I've seen other speaker measurements will similar comments.

My question isn't about the 4-ohms but this combination with these phase angles at low frequencies.

Do these phase angles of –50° at 100Hz and –45° phase angle at 73Hz at 4-ohms (just examples only), present a challenge for the AHB2?

If not, can you share info about how it's not an issue for the AHB2.

Thanks in advance

No this is not a problem at all. Capacitive or inductive speaker loads will not cause stability issues. The feed-forward system is inherently stable. The only limitation is the over-current protection circuit. The amplifier will shut down if repetitive peak currents exceed 29 Amps or if the RMS current exceeds 20.5 Amps for more than a few seconds.

The TEMP lights will flash when peak currents exceed 29 Amps. This is a warning that the protection will kick in if the condition persists. The CLIP lights will flash whenever the amplifier is driven into voltage clipping causing the THD to exceed about 0.5%. If the CLIP and TEMP lights flash together, it indicates that you have simultaneously reached the voltage and current limits of the AHB2. Each channel has its own set of lights and has its own protection monitoring. If protection is triggered, both channel will mute. To protect your tweeters, the distortion monitoring will not allow sustained operation if the amplifier is driven into voltage clipping resulting in a sustained THD exceeding 1%.

The feed-forward correction keeps the AHB2 distortion free when driving very low impedances and difficult phase angles. The protection system in the AHB2 monitors output current, output voltage, distortion, temperature and other critical parameters. The AHB2 can cleanly drive a 1.4-Ohm resistive load to full output voltage on both channels for several seconds without triggering the protection circuits. This is much longer than any musical peaks. Load impedances below 1.4 Ohm (stereo mode) or 2.8 Ohms (mono mode) may flash the TEMP lights indicating that peak currents exceed 29 Amps. If the overload condition is severe enough, and persists for a long enough time, the over-current protection may trip. The amplifier remains distortion free until the protection system shuts it down.

Unlike most amplifiers, the protection circuits are not in the audio path. For this reason, the protection circuits have no impact on the audio quality until they activate and mute the amplifier to protect the amplifier and speakers from damage. The speed at which the protection circuits react is a function of the severity of the overload. You can place a short circuit across the output of the amplifier while it is delivering full output. This will immediately activate the protection and the amplifier will be fully protected from this short-circuit event. In contrast, we allow peak currents to exceed 29 Amps without immediately shutting down. The time interval is determined by the severity of the overloads. This is all controlled by digital signal processing (DSP) in a Xilinx FPGA.

From the amplifiers perspective, -45 degrees at 4 Ohms is equivalent to driving a 2.8-Ohm resistive load. This can be driven to full output voltage in stereo mode or in mono mode.

-50 degrees at 4.1 Ohms is equivalent to driving a 2.6-Ohm resistive load. This can be driven to full output voltage in stereo mode and can be driven to within 0.1 dB of full output in mono mode without flashing the TEMP lights. Even in mono mode, it would be virtually impossible to trip the protection with music although it could be done with a 100 Hz sinusoidal test tone if it was played at, or slightly above, full power for a few seconds.

One unique characteristic of the AHB2 is that the power supply rails are tightly regulated. In contrast, virtually all other amplifiers use unregulated power supplies. This means that the DC rails sag in a conventional amplifier when the amplifier is driving a difficult load. With the AHB2 the rails maintain their voltage and the amplifier continues to deliver nearly the same rail-to-rail voltage swing. This can be seen from the fact that the AHB2 delivers almost exactly twice the power into 4 Ohms as into 8 Ohms. It is also why we get a near 4:1 increase in power when running in bridged mono.

 

[dkfan9]

[Polish] https://audio.com.pl/testy/kino-domowe/amplitunery-av/2141-denon-avr-x2100w

to English: https://translate.google.com/translate?sl=auto&tl=en&u=https://audio.com.pl/testy/kino-domowe/amplitunery-av/2141-denon-avr-x2100w

That site is a wonderful resource. I've checked out that review before. I'm almost a little disappointed the 1W SNR is on the above average side for AVRs... little room for within category upgrade in the main area I have performance complaints. Though now that i look through the tests again, the 4 ohm performance looks above par for AVRs, both in pure power and in matching THD+N to 8ohm voltage up to 15V, and exceeding it at low output). That's a nice verification of my experience. I've never felt the need for more power, or felt bass performance being held back, despite having 4 speakers that dip below 4 ohms at low frequencies (90+db sensitivity but still).

 

[John_Siau]

Here are the comparison charts for the F228Be. These are a little more precise. Since output impedance is not constant across frequency, I modelled the amplifiers in VirtuixCAD as muted speaker drivers in series with the F228Be. I traced the damping factor charts published by Soundstage Network here and here, exported as text, transformed the DF to impedance in LibreOffice Calc, re-exported as text, and uploaded the impedance response to drivers representing the A31 and AHB2. To represent the AHB2 in monoblock mode, I put two of them in series prior to the driver. I gathered one set of data points using the F228Be's measured frequency response, and another assuming a flat amplitude, both using its measured impedance. Both were placed around 80dB because that's the program's default driver SPL. The F228Be's FR and impedance measurements were taken from the Stereophile review. Here's what the model looked like:
View attachment 26933

Since this can be a sensitive subject, a few disclaimers before I begin: I don't claim these changes are or are not audible. They are, objectively, very small in level relative to many other pieces of an audio setup, particularly the room and speakers. If perceptible, they may still be subjectively outweighed by the amplifier's distortion and noise. On the other hand, they may also amplify or counteract issues in the room or speakers. If nothing else, they're interesting and seem to be in the spirit of this forum. The scales vary between some of the graphs, not to mislead, but to examine differences in maximum detail. I can't claim the sourced measurements are flawless, but they're the best I found, and the two amps are measured with the same methodology. I certainly can't claim my generated numbers are flawless. I know I'm short phase information, and to quote a guy from the last gasp of the age of irony, "there are also unknown unknowns." Finally, throughout this post, when I refer to the amplifiers, specifically or generically, I'm using it as shorthand for "the amplifier's output impedance".

Now, jumping in: The first two graphs show the "flat" F228Be's amplitude response through the amplifiers, with the second graph level matched at 1000Hz. Top to bottom here is roughly .6dB. REW was used for all display and most manipulation.

View attachment 26916

View attachment 26917

The A31 is louder across the audio band, but below the last two octaves, the difference between the A31 and monoblocked AHB2 is less than .2dB (ignoring any level increases brought by D+N), so perfect level matching might be a challenge. By 10kHz the difference is over .25 dB, reaching over .3 at 20k. Below 5k, even at this scale differences between the AHB2 in stereo and the A31 are minor, but the AHB2 in monoblock configuration does diverge a little more. In the upper midrange/lower treble region where the four lines run together on the bottom graph, the Revel is at its highest impedance above port tuning and the Benchmark's output impedance is still relatively low. At high frequencies the Benchmark's impedance rises quite a bit, and at low frequencies the Revel's drops below 4 ohms, so the amps diverge. Whether or not this is what he was hearing, these graphs do seem to match up with Kal's impressions on bass performance in his F228Be review.

Ok, next up are the differences when modelling using the speaker's measured response. The following three graphs show purely differences, comparing each amp to the others (including AHB2 in stereo vs. mono configuration), and are scaled even tighter, at ~.4dB between the top and bottom of the graph. 0dB on the graph is set to 1000Hz to better see differences in frequency balance, since the differences in level are pretty clearly seen above (and again below).

View attachment 26918View attachment 26922
View attachment 26920

These graphs don't provide any info that can't be extracted from the first two graphs, but they do highlight two things: differences in output impedance are primarily significant when accompanied by swings in speaker impedance (as shown by John Atkinson in nearly every set of tube amp measurements into his mock speaker load), and the AHB2's frequency response in stereo is closer to the A31's than to itself in a monoblock configuration.

And for the final graph, we have the F228Be's measured FR on its own and attenuated by each amp. These are the same numbers that got us the three previous graphs. Around 6dB from top to bottom of the graph (compared to .4-.6 for the graphs not showing speaker response). Oh, and a bigger picture to help see the details.
View attachment 26931
And here we see why it's easy to discount an amp's significance to frequency response and overall sound quality. The largest deviation in any amp above was below half a dB. Here, we see the speaker's response swing by a couple of dB in the treble multiple times, and more than that lower in frequency (though with the limited far field measurement resolution and compromises of near field measurements leading to the upper bass bump, we can't be too sure what's real and what's a measurement artifact). And this is a great speaker.

This graph still has an interesting story to tell in the treble. Where the speaker starts shelving up, above 4k, the monoblock AHB2's higher output impedance flattens the speaker a bit, acting like an EQ circuit (pretty good THD+N for an EQ circuit). The same output impedance that causes almost half a dB in variance when "driving" the "flat" F228Be's in the first graph. To bring it full circle to Kal's review, the "slight bit of softening" he sensed above the bass in the A31 relative to the AHB2 may have come from the Parasound shelving down 1.5-4kHz relative to the treble a little more than the Benchmark. Or maybe the increased noise and distortion in the A31 caused the impression. And maybe it was the increased distortion and noise that gave the A31 more body down low. Maybe he imagined all of it. I don't know.

In the end, I'm pretty confident either of these amplifiers would outperform my Denon X2100W. I don't know about the audibility of low level distortion and small FR variances, but I do know I can hear noise, and in my former apartment home theater it was easy to hear hiss on silence from the couch, 6 feet from the TV. Not always, but any time I caught a glimpse, it would be days before I would forget and be able to tune it out again. But hey, I don't even have pre-outs, let alone professional balanced ones, so I guess I wouldn't put these amps to good use anyway...

As you have suggested, the frequency response variations are much too small to be significant. You would discover much larger variations between the left and right speakers and between speaker samples. You would also discover larger variations by moving you listening location by an inch or two. I would be more inclined to look at THD differences and phase response differences.

I will say that I like the numeric calculations that you have made to combine the speaker data with the amplifier data to predict the system frequency response - nice work!