Review and Measurements of Benchmark AHB2 Amp

[RayDunzl]

The conventional measures, the ones in product specifications and reviews, are typically only a subset of significant measurements, so can only describe some aspects of the sound.

No offense, but humans can come up with some remarkably florid descriptions of "sound".

I expecct to continue to pursue the "cleaner is better" idea for the in-room reproduction of recorded sound.

I have only crude measures as verification of any steps I might take toward that fuzzy goal, but, I'm happy with the results so far.

 

[GrimSurfer]

Me, too.

Me three!

 

[tktran303]

@JohnPM

Re:

Sean E. Olive said:
Pref. = 12.69 − 2.49 * NBD_ON − 2.99 * NBD_PIR − 4.31 * LFX + 2.32 * SM_PIR

(The paper defines NBD_ON as "On-axis frequency response Average Narrow Band Deviation (dB) in each ½-octave band from 100 Hz-12 kHz". NBD_PIR is the same for the predicted in-room response (which itself is derived from on-axis and off-axis measurements). LFX is "Low frequency extension (Hz) based on -6 dB frequency point transformed to log 10", and SM_PIR is "Smoothness (r²) of predicted in-room response based on a linear regression line through 100 Hz-6 kHz")

Have you looked into this?
Is this something that could be incorporated into REW?

 

[edechamps]

Is this something that could be incorporated into REW?

In order to measure the metrics that are used in that equation, you need a measurement setup that is similar to the one described in ANSI/CTA-2034 - which basically means you need an anechoic chamber to measure the speaker in. That's a much bigger obstacle than having these calculations integrated in REW!

 

[tktran303]

The measurement setup is probably not within the realms of the average amateur.

But it IS possible to get reasonably accurate measurements without a full anechoic chamber, down to about 10Hz using a combination of near field, far field and “blending” as described here-

Reference:
http://audio.claub.net/software/FRD...curate In-Room Frequency Response to 10Hz.pdf

Or could one use simulated measurements?
The linear regression model “cares” to the half octave, not unsmoothed or even 1/48 or 1/24 octave smoothed that loudspeaker designers like to work with.

Example- is there enough information in the graph below to generate the preference value?

Reference:
https://pkaudio.webnode.cz/

graphs generated by VirtuixCAD by Kimmo Saunisto

 

[GrimSurfer]

I used to believe that, but I've learned over the years that the issues are a bit more complex. For example, I referred earlier to the masking of offensive higher order harmonics by the purposeful addition of lower order harmonics. The more colored tube amps would be an example of this.

Another consideration is *what* you measure. Studies suggest that harmonic distortion is inaudible on a sine wave at the levels in any good amplifier. And when you look at DAC's, the measured steady state distortion is typically so low that it is almost certainly inaudible. And yet it has also been demonstrated that there are other forms of distortion that are very audible in small amounts. Crossover notch distortion, which as I mentioned is audible in very low amounts, is an example. A couple of years John Siau told me that he'd successfully ABX'd crossover notch distortion at a remarkably low level, I forget what it was but it was impressive. And what is inaudible on a sine wave is *not* necessarily inaudible on music, even though it is composed of sine waves. But that's at a single frequency. Play those sine waves together, and they will intermodulate. And intermodulation distortion sounds awful.

The conventional measures, the ones in product specifications and reviews, are typically only a subset of significant measurements, so can only describe some aspects of the sound.

I understand how masking is an audible (or, more correctly, inaudible) phenomenon. I don't understand how masking applies when measuring output electronically. More to the point, I don't think it's possible to hide (for example) an unwanted uneven order harmonic from the scope without employing additional electronic means.

Of course, some (possibly most or all) Class D amp manufacturers employ various means to screen out ultrasonic harmonics. But this isn't masking as much as it is filtering.

I believe that whether harmonics are audible or not depend on (1) their frequency, (2) their intensity, and (3) proximity to the fundamental. I'm open to the view that odd order harmonics may be more discordant to the ear than even ones of diminishing intensity though... but I've seen research material in the past that challenges this. I suppose this becomes immaterial once harmonics or any kind fall below -120dB (which would suggest that I'm not a believer in psychoacoustics!).

So for the time being, I'm inclined to accept that an amp that limits harmonics from exceeding natural norms (natural norms being: Even order, diminishing with frequency, and of diminishing orders of magnitude lower than the fundamental... which is the same as one might experience directly with most musical instruments) perform better than those which don't.

With regard to measurements, I totally agree that not everything which may be important is measured (or, even, reported) by manufacturers. But certain basic measurements, when verified, give an indication that a manufacturer hasn't screwed something basic up. Frequency range, THD+N, IMD, linearity, slew rate, output in RMS or reasonable equivalent, etc. are a few of the things that immediately come to mind.

Once a reasonable threshold for these measures has been reached, one may more reliably have faith in what their ears are telling them... and even that can be affected significantly by hearing acuity, loudspeaker performance and room effects.

 

[edechamps]

The linear regression model “cares” to the half octave, not unsmoothed or even 1/48 or 1/24 octave smoothed that loudspeaker designers like to work with.

That's not true for SM_PIR, which is computed from high-resolution data (1/20 octave, IIRC). The paper even explicitly states "Models based on 1/3-octave measurements produce less accurate predictions of sound quality."

It might be possible to come up with something usable in non-anechoic conditions through gating/windowing, but that's fraught with peril and won't work for low frequencies anyway.

 

[tktran303]

Hi Ed,

I should really read the actual AES publication before commenting further...

BRB.

 

[Xulonn]

Interesting discussion. Subjectivists usually say that it is impossible for live music to be reproduced faithfully.

One of the most frustrating things for me in audio listening discussions is people who listen primarily to amplified instruments and performances repeat ad nauseum and want their super fantastic audio systems to sound like live performances. Really? You expect a Benchmark/Focal system to sound just like a stack of Marshalls and a Peavey PA system in a night club? Get real!

Most music recordings are collaborative creations between musicians, producers, recording engineers and mastering engineers - and have no reality as a live event to be recreated. They are studio creations. In my view, any quest for extreme "accuracy" in an audio system can logically only be to reproduce the sound of the mastering studio - and/or the information the engineers encoded into the recording.

Classical music in a good acoustic venue - from solo instruments to full orchestras, folk music or jazz with only acoustic instruments and no microphones and PA system, again in a good acoustic environment, are the only types of music where there is any logic in wanting to reproduce a live event. As a high school student in Chicago, I was an usher for the Chicago Symphony Orchestra near the end of the Fritz Reiner era - and right in the middle of the period of his magnificent series of pioneering early stereo recordings for RCA. Over the many years, I have attended a fair number of acoustic music events, and such events are the basis for my desire to reproduce live music events, which are a small portion of my collection of recorded music.

OTOH, I have a fair collection of electronic (synthesizer) music - which has only electronic equipment and speakers as sound sources. If you directly record a guitar pickup or synthesizer digital output, what is the "live" acoustic reference you want recreate? Oops - it doesn't exist - care to restate your audio reproduction goals?

Many audiophiles who only listen to studio recordings and live, PA amplified non-classical concert music "say that it is impossible for live music to be reproduced faithfully" when such a phrase is essentially a non-sequitur.

In spite of my profound disagreement with contributor Maty here on many issues the "how" and "why" of things, I agree whole-heartedly that what the music does for us emotionally as individuals should be the core of any audio system quest. It was interesting for me to learn here today that recording and mastering engineers "dry" (clincal, detailed, analytical?) acoustic systems and environments for making recordings so they can more easily detect flaws and imperfections, but prefer "wetter" (not clincal, detailed or analytical) for listening at home.

The people that I respect most in audio - including contributors at forums like this - are the ones who pursue listening goals that are real and approachable, even if those goals are never quite completely attainable. I do not respect those seeking to attain what is not possible - reproduction of something that does not even exist except as strange fantasies. (The technical aspects of audio system is another interest that is mostly separate from my enjoyment of music, but interconnected with it. I am a "techhie" at heart.)

Maybe I will start using my old Les McCann line when people start babbling about making studio-created and manipulated recordings sound like live and ask which live events they are trying to reproduce.

View attachment 26683

p.s. I still would love to own a Benchmark DAC and amp driving a pair of Goldenear Technology Reference or similar quality loudspeakers

 

[GrimSurfer]

One of the most frustrating things for me in audio listening discussions is people who listen primarily to amplified instruments and performances repeat ad nauseum and want their super fantastic audio systems to sound like live performances. Really? You expect a Benchmark/Focal system to sound just like a stack of Marshalls and a Peavey PA system in a night club? Get real!

Most music recordings are collaborative creations between musicians, producers, recording engineers and mastering engineers - and have no reality as a live event to be recreated. They are studio creations. In my view, any quest for extreme "accuracy" in an audio system can logically only be to reproduce the sound of the mastering studio - and/or the information the engineers encoded into the recording.

Classical music in a good acoustic venue - from solo instruments to full orchestras, folk music or jazz with only acoustic instruments and no microphones and PA system, again in a good acoustic environment, are the only types of music where there is any logic in wanting to reproduce a live event. As a high school student in Chicago, I was an usher for the Chicago Symphony Orchestra near the end of the Fritz Reiner era - and right in the middle of the period of his magnificent series of pioneering early stereo recordings for RCA. Over the many years, I have attended a fair number of acoustic music events, and such events are the basis for my desire to reproduce live music events, which are a small portion of my collection of recorded music.

OTOH, I have a fair collection of electronic (synthesizer) music - which has only electronic equipment and speakers as sound sources. If you directly record a guitar pickup or synthesizer digital output, what is the "live" acoustic reference you want recreate? Oops - it doesn't exist - care to restate your audio reproduction goals?

Many audiophiles who only listen to studio recordings and live, PA amplified non-classical concert music "say that it is impossible for live music to be reproduced faithfully" when such a phrase is essentially a non-sequitur.

In spite of my profound disagreement with contributor Maty here on many issues the "how" and "why" of things, I agree whole-heartedly that what the music does for us emotionally as individuals should be the core of any audio system quest. It was interesting for me to learn here today that recording and mastering engineers "dry" (clincal, detailed, analytical?) acoustic systems and environments for making recordings so they can more easily detect flaws and imperfections, but prefer "wetter" (not clincal, detailed or analytical) for listening at home.

The people that I respect most in audio - including contributors at forums like this - are the ones who pursue listening goals that are real and approachable, even if those goals are never quite completely attainable. I do not respect those seeking to attain what is not possible - reproduction of something that does not even exist except as strange fantasies. (The technical aspects of audio system is another interest that is mostly separate from my enjoyment of music, but interconnected with it. I am a "techhie" at heart.)

Maybe I will start using my old Les McCann line when people start babbling about making studio-created and manipulated recordings sound like live and ask which live events they are trying to reproduce.

View attachment 26683

p.s. I still would love to own a Benchmark DAC and amp driving a pair of Goldenear Technology Reference of similar quality loudspeakers

So true...

 

[phoenixdogfan]

Are you referring to that paper, written by Sean Olive under Harman? If so, it is publicly accessible, though sadly behind a paywall, so I'll paraphrase it. Here's the equation:



The paper defines NBD_ON as "On-axis frequency response Average Narrow Band Deviation (dB) in each ½-octave band from 100 Hz-12 kHz". NBD_PIR is the same for the predicted in-room response (which itself is derived from on-axis and off-axis measurements). LFX is "Low frequency extension (Hz) based on -6 dB frequency point transformed to log 10", and SM_PIR is "Smoothness (r²) of predicted in-room response based on a linear regression line through 100 Hz-6 kHz".

The above equation can predict the subjective preference score that human listeners gave while blind testing the 70 loudspeakers in the study with a coefficient of determination of 89%. The nitty-gritty details of these metrics and how to compute them are described in the paper.

I would find it highly valuable if people doing loudspeaker measurements also provided the result of this equation, as it condenses most of what matters into a single, easily comparable number (similar to how @amirm likes to use SINAD to score electronics).



I never heard of that one. I'd be interested in learning about studies on that topic - accurately quantifying the audible impact of non-linear distortion is still a somewhat unsolved problem, as far as I know. 15 years ago Geddes and Lee proposed a new metric that seems better than THD but it doesn't look like it caught on.

If the NRC measurements included this calculation, it would eliminate the need for Stereophile altogether.

 

[josh358]

No offense, but humans can come up with some remarkably florid descriptions of "sound".

I expecct to continue to pursue the "cleaner is better" idea for the in-room reproduction of recorded sound.

I have only crude measures as verification of any steps I might take toward that fuzzy goal, but, I'm happy with the results so far.

You mean as in "woofer" and "tweeter"? We engineers do it too.

I actually find those terms easy to understand for the most part, because they're simple analogies and I'm familiar enough with the audible phenomena they represent that I usually know what they mean.

The recording engineer's "wet" or "dry" acoustic would be an example. Obviously, acoustics don't have humidity; the technical equivalent would be something like RT60. But if you've heard spaces with more or less reverberation, you know what they're getsing at with the analogy -- the reverberation is like a certain "wetness." Similarly when we talk about a room that's "live" or "dead." And many other metaphoric expressions that have become more or less standard through usage.

The key I think is having heard the phenomenon -- once you have, it's pretty easy to associate the expression with it. If you haven't, I can see that it would be puzzling.

 

[josh358]

Are you referring to that paper, written by Sean Olive under Harman? If so, it is publicly accessible, though sadly behind a paywall, so I'll paraphrase it. Here's the equation:



The paper defines NBD_ON as "On-axis frequency response Average Narrow Band Deviation (dB) in each ½-octave band from 100 Hz-12 kHz". NBD_PIR is the same for the predicted in-room response (which itself is derived from on-axis and off-axis measurements). LFX is "Low frequency extension (Hz) based on -6 dB frequency point transformed to log 10", and SM_PIR is "Smoothness (r²) of predicted in-room response based on a linear regression line through 100 Hz-6 kHz".

The above equation can predict the subjective preference score that human listeners gave while blind testing the 70 loudspeakers in the study with a coefficient of determination of 89%. The nitty-gritty details of these metrics and how to compute them are described in the paper.

I would find it highly valuable if people doing loudspeaker measurements also provided the result of this equation, as it condenses most of what matters into a single, easily comparable number (similar to how @amirm likes to use SINAD to score electronics).



I never heard of that one. I'd be interested in learning about studies on that topic - accurately quantifying the audible impact of non-linear distortion is still a somewhat unsolved problem, as far as I know. 15 years ago Geddes and Lee proposed a new metric that seems better than THD but it doesn't look like it caught on.

Excellent, thanks. I'd missed the paper and wasn't aware that they had published the formula.

 

[March Audio]

I also agree, to a degree, with these comments based on comparison with the similar Parasound A31. It is no so much that the Benchmark robs the music of excitement but that the Parasound and Benchmark provide different types of excitement.

Hi Kal,

You do know the question that's coming

Did you compare blind and accurately level matched?

Second question. I have no idea what you mean by "excitement". Can you expand on this?

 

[pozz]

It's worth keeping in mind that this amp was created not only to reduce noise and distortion but also to avoid the euphonic qualities that would make for inaccurate monitoring in studios.

A really "good" sounding amp like the First Watt SIT3 (besides clipping because of its low output) would be totally unsuitable. Same goes for the euphonic qualities of certain DACs—but not for the rest of the signal chain. The Lavry AD122 96 has a strong reputation for the saturation processing it uses on clipping signals during recording (kind of like tape), for example, and then there are loads of things said about compressors, EQs, filters, mixers, etc.

 

[Kal Rubinson]

Did you compare blind and accurately level matched?

Of course not.

Second question. I have no idea what you mean by "excitement". Can you expand on this?

Actually, I was just riffing on the use of the word in the quote I was responding to.

 

[josh358]

I listen to every genre except country music. My system is listed here: https://www.audiosciencereview.com/...measurements-of-lyngdorf-roomperfect-eq.6799/

Nice system, but by revealing I mean something like stats or ribbons. Though I'm sure it would show up the difference between something like the e38 and the Gungnir MB.

 

[josh358]

Not really when they all parrot the same thing. Oh there was more micro dynamics. The background was blacker. The soundstage wider. I could go on. All the reviews read like each other.

I have tested products that are supposedly euphonic but they absolutely are not. They just distort the sound, making it brighter. Audiophiles are lucky that they don't hear the impairments and instead, substitute improper testing to arrive at positives.

I don't use generic descriptions or clichés as verification. Rather, I'm interested in specifics, like the description of the "ESS glare" that I quoted above.

Of course, there are differences in soundstaging. The Gungnir for example has a much deeper soundstage on my system than the e38. Schiit, rightly or not, attributes that to the closed form reconstruction filter. The fact that low distortion electronics have different soundstaging is something that has always puzzled me from an engineering perspective. We have a good handle now on why loudspeakers vary in their spatial representation, but I've never seen a satisfying explanation of why electronics do, other than Nelson Pass's interesting observation about third harmonic (which would essentially apply to tube gear).

 

[amirm]

I don't use generic descriptions or clichés as verification. Rather, I'm interested in specifics, like the description of the "ESS glare" that I quoted above.

That is as cliche as it gets. You are not joking with us, are you?

 

[maxxevv]

'That all said I concur that there is a fairly narrow band where the contrast sounds excessive (to me) on Sabre DAC products. My question today is "What, in your opinion, is that frequency range where the Sabre DACs sound too dynamic?"

'Now if you disagree with the premise, and like the Sabre DAC sound as is, then don't muddy the waters. But if you agree that Sabres sound bright, then what's your best guess as to the frequency span?'

That is over generalising in my experience and opinion.

And there are way too many factors involved to say for certain which might be the dominant one.

I have listened to different generations of the ESS chips, albeit all with different makes, and output stage designs.

ES9018, ES9028Q2M, ES9038Q2m, ES9038PRO.

Subjectively, with blind volume equalisation (since I do not have the right equipment to equalise them), the ES9018 sounds somewhat different from the other chips on the list. And if anyone wishes to talk about "glare", the only one on the list that sounds like it does have the so called "glare" is the 9018. So, even from a completely subjective point of view, this "ESS Glare" thing is completely outdated. Those forum parrots need to update their listening experiences.

Now, going back to a little more logic but still subjective experience, the intrinsic output design stage of the different DAC's can play a part too.

The ES9018 DAC I used had a separate RCA output option. Direct from its headphone output, it sounded "airy" and "well separated" and a little thin in its bass presentation and was very obviously tonality wise, different from the other ESS chips on the list. (I have actually spent a full 3 hour session comparing all of them back to back. ) But then I plug the RCA output to the JDS Atom and listened to it, the differences between it and the ES9038Q2M and the ES9038PRO were much , much less discernible. For that matter, the ES9038Q2M was from a Khadas Tone Board.

So, whatever people say about the idea of "ESS Glare", its at best outdated. At worst, its their inherent bias as a result of listening to outputs with poor treble acuity prior to it or worse, using systems that have poor implementations high on distortion, which our ears are much more sensitive to from the treble range onwards than the bass frequencies.

Edit and Addendum:

All the above comparisons were done using A HD800, HD6XX and HE4XX, with all 3 headphones being used repeatedly between identical tracks for comparison.