I am actually grateful to the good people of ASR, believe it or not. I like measurements and data, and enjoy thinking about the technical aspects of audio kit, as well as actually listening to it. I see no conflict here, I enjoy both. I have said before that I love the concept of ASR, but I am a little troubled by some apparent biases and some of the posting there is a little thoughtless. But this is the internet, and I do note that in between some of the more gleeful Devialet bashing, there is some interesting, technical and considered posting going on. OK, I wish Amir could be a bit more considered in his approach, but I have enjoyed a bit of forum watching, and I genuinely appreciate how the ASR article has pushed some Devialet related technical issues into the spotlight. If nothing else, it has introduced me the wonderful world of Paul Miller’s AVtech measurement site. So maybe I would prefer a slightly different ASR, but in the absence of that I would rather we have the ASR that we have in the world rather than none at all.
There are some that consider measurements to be definitive. Subjective observations are terrible things, inconsistent, unreliable. But a measurement is definitive, repeatable. If product X measures better than product Y, then product X is better. Oh, and if you think product Y sounds good, that it has excellent sound quality, well good for you, but you probably just like the sound of the distortions. You could even throw in an insult for good measure. (accidental pun) Oh, you think product Y sounds good? Did your wife tell you that when she walked in from the kitchen?
Some typical “objectivist” words and thinking above, but one thing that has fascinated me recently is how the language surrounding the measurements seems to have far more power than the objective measurements themselves. This can be seen everywhere surrounding ASR review, the ASR posting and posts here.
Consider the first part of the ASR review. It starts with “Distortion products are quite low at < -110 dB. However, the noise floor is high in low frequencies causing our SINAD to suffer a lot, ending with a metric of 92 dB. This is well above average compared to all the amplifiers we have measured but way short of state-of-the-art:” later it states: “There are some lofty numbers like -130 dB noise floor is thrown around by Devialet. I am not sure how that is computed. Using analog Line-in for consistency with other amplifier measurements we get rather ordinary numbers:”
Interestingly, the numbers are very similar to that made by Paul Miller for the HFN review, although Paul’s amp did not fail during testing and he proceeded to measuring via the digital input, which offered even less noise and generally better measurements elsewhere. The comments in the HFN measurements side bar stated “The vast majority of users will, I presume, engage with the amp digitally either via USB or wired/Wi-Fi network, in which case a 0dBFs input realises exactly 100W/8ohm and 200W/4ohm at the ‘0.0dB’ position (both at 0.0007% THD). Via the analogue input this increases to 120W/8ohm and 240W/4ohm, the near perfect ‘stiffness’ of the power supply reflected in the 115W, 230W and 450W delivered under dynamic conditions into 8, 4 and 2ohm loads. Note also how distortion barely increases with reducing load impedance [see Graph 1]. Couple this with the infinitesimally low 0.003ohm (3mohm) output impedance and, provided you don’t use scrawny cables, the 170 will rule your loudspeakers with the proverbial rod of iron. Through bass and mid frequencies, ‘digital’ distortion is half that for the analogue input, the former achieving 0.0005% as opposed to ~0.001% for the same 1kHz/10W/8ohm output. The opposite occurs at very high frequencies where the line input reaches 0.006%/20kHz and the digital inputs, regardless of sample rate, increase to 0.017% (10W/8ohm). The A-wtd S/N ratio is fabulously wide at 93dB and 98dB re. 0dBW (analogue/ digital) or 118dB re. 100W/8ohm while the response(s) are fl at to +0.14dB/20kHz, –0.6dB/45kHz and –1.6dB/85kHz (48kHz, 96kHz and 192kHz media).”
The same measurements are there, but ASR thrusts the bad at you, the HFN text is so bland that nothing really grabs your attention, but similar measurements are there. It is perhaps also worth noting that the HFN article ends with a “sound quality” score of 91%. I have no idea how this number is established, but as a regular reader of HFN I know that scores of over 90% are VERY rare. ASR state that using the analog input provides consistency with tests of other amplifiers, HFN state that it is assumes that most will use the amp via digital input, both valid points, but placing the emphasis in different directions.
Same numbers, different language, and we being human react accordingly.
It is the same story with the frequency sweep test. Despite the fact that the amp failed during this test, ASR are happy to state “Turns out this amplifier cannot reproduce high frequencies reliably at 5 watts!”.
Thanks to this, a number of us have started thinking about this, posting, asking questions and speculating as to why a Devialet Expert happens to have restricted power at high frequencies. Yet, this information has been in the public domain for well over a decade, with Stereophile reporting “By contrast, the D-Premier, given a mix of 19 and 20kHz tones at a peak level of 20W into 4 ohms (the highest level the amplifier would deliver with this signal without the power supply collapsing), performed well on this test (fig.9).”
This is a classic case of hiding in plain sight. I have read the Stereophile review, I have read the above comment, and thought nothing of it. Again, a bland statement that goes unnoticed, but when ASR thrust it in your face, we start thinking… Think about this, that statement has been out there for maybe 12 years now, and nobody has cared. I guess if nothing else this demonstrates that it genuinely does not matter. I have never heard of anyone complaining that their D-Premier kept shutting down, but it is fascinating how the words, not the measurements get our attention.
Thinking about how words can influence something as objective as a measurement, I thought I would have a go at it myself. In terms of amplifiers, the ASR measurement king at the moment is the Benchmark AHB2. Stereophile have also measured this amp, and no doubt it displays incredibly good performance, truly state of the art, for a product that is not overly expensive by “high end” audio standards.
https://www.stereophile.com/content/benc...-amplifier
Some ASR contributors have noted that you could buy an AHB2 based system for a lot less money than a Devialet. Seems a fair point to me, state of the art measurements, good price, why not?
Then we have that pesky subjective thing. Some of the subjective comments for the AHB2 have been a bit mixed. OK – objectively you could say that maybe it is just that some week minded folk who subjectively do not like the sound of the objectively measured superiority of the AHB2. They actually like the distortions other amps make. This might even be true. To be honest, I cannot judge, I haven’t heard one myself. (I would love to listen to the AHB2, BTW – If someone wants to lend me a couple)
But maybe it is not true, maybe there is an objective reason a Devialet could sound better? Here goes, true objective measurements coupled with made up by me in your face language.
It is no wonder that the AHB2 sounds dreadful in comparison to a Devialet, OK the AHB2 has fractionally lower noise and distortion, but at the kind of levels we are talking about here any notion that you could “subjectively” hear this as “better” is for the fairies. What really matters in terms of amplifier performance is the low output impedance, this any idiot could clearly hear in terms of accuracy and transparency in terms of how your amp is controlling your speakers. And look at the numbers, these don’t lie. The Devialet has a measured output impedance of 0.001–0.004ohm, this is state of the art. The Benchamark measures 0.034–0.17ohm, this is an order of magnitude worse! No wonder the Devialet crushes it for performance.
OK – The above text is something I have made up for a bit of fun. As I said, I have not listened to an AHB2. Who knows, if I did I might imediately sell my Devialet and swap. Or maybe not. The measurements in the text above are real, taken from Paul Miller’s AVtech, and the Devialet is the better of the two. I was chatting to someone once who works for a speaker manufacturer, the topic was that I subjectively thought that the Devialet was just about the best amp I have heard in terms of bass depth and resolution. The speaker guy suggested that low output impedance is the key here, it gives the amp better control over the speaker drivers. And what about other things, slew rates, input impedance, transient response?
So there might be something in my words above. There is a recent thread on Audiophile Style:
https://audiophilestyle.com/forums/topic.../#comments
Some interesting posts in the many pages, but it strikes me that although the industry, magazines etc. are very good at taking measurements, we do not really have much of a clue what they mean for sound quality. As an example, we can look at the noise level of a Devialet via analogue versus digital input, and the differences are clear, and yet run an external DAC into a Devialet via the digital inputs, it sounds dead quiet. But how is it affecting the sound? I know mine sounds a lot sharper, clearer and more detailed running digitally rather than fed via my AVR. But what about slew rate, jitter, input and output impedance? I love measurements, data and all the rest of it, but I would need to listen as well to have any degree of confidence as to what it means. But one thing has become very clear to me, the language and words surrounding these little measurement numbers have huge power and influence over many of us, and I would have to include myself in that.
One irony in all this was that some people on this forum we a little disappointed the read that Devialet’s Mathieu Pernot uses measurements during product development, rather than listening. I recall him saying something along the lines of “the measurements are the same first thing Monday morning as last thing Friday night”, implying that subjective hearing will vary, measurements not. It is hard to disagree with this notion. This makes me wonder just how well Mathieu might understand what measurements actually matter, in terms of actual delivered sound quality. I have no idea, but my subjective view of the Devialet hints that he might actually be rather good at this. In terms of R&D, there must be compromises between aspects of design, do you optimise what really matters to sound quality or what gives great headline numbers? A fascinating topic, but one I can only guess at.
So what of Devialet? Now we have the calls for them to explain why an Expert can only deliver 20W or whatever at high frequencies. I might be wrong, but I suspect Thumb5 is along the right lines when he suggests this could be an inherent limitation of how ADH works, and that it does not matter one iota in terms of actual performance delivering music.
Do I expect Devialet to respond and clearly explain all this? OK – If Devialet respond directly stating something like “yes, it is an inherent aspect of ADH, they can only produce 20W at 20kHz”, then I would be so surprised I would gladly eat my remote. If they do respond, I would expect something more along the lines of “The ADH system is not like conventional amplifiers and it includes sophisticated protection mechanisms that can react to the input signal both dynamically and in relation to frequency response, for this reason the amplifier does place some restrictions on frequency sweep type measurements that are of zero significance in normal use.” As for the other measurements that Devialet quote, you can find more or less equivalent numbers in the Paul Miller tests, but you have to be selective, and the Devialet quoted numbers are typically absent of the associated qualifying factors, so they can relate to “selective” measurements. A manufacturer selectively quoting numbers to make the product look as good as possible? It has been seen before.
One thing to keep in mind, the numbers measured independently by Paul Miller are good – and hey – it scores 91% for sound quality. We should all try to keep in mind that we should accept the measurements for what they are and what they mean, and if we are interested maybe we should try to understand better what they mean. Reacting to possibly inflammatory language that may surround these measurements can lead us astray, and remember, these words can have real power.
Anyway, I have surprised myself by how much I have written here. Time to listen to my little Devialet for a while, whilst we all wait for Devialet to respond to all this.
There are some that consider measurements to be definitive. Subjective observations are terrible things, inconsistent, unreliable. But a measurement is definitive, repeatable. If product X measures better than product Y, then product X is better. Oh, and if you think product Y sounds good, that it has excellent sound quality, well good for you, but you probably just like the sound of the distortions. You could even throw in an insult for good measure. (accidental pun) Oh, you think product Y sounds good? Did your wife tell you that when she walked in from the kitchen?
Some typical “objectivist” words and thinking above, but one thing that has fascinated me recently is how the language surrounding the measurements seems to have far more power than the objective measurements themselves. This can be seen everywhere surrounding ASR review, the ASR posting and posts here.
Consider the first part of the ASR review. It starts with “Distortion products are quite low at < -110 dB. However, the noise floor is high in low frequencies causing our SINAD to suffer a lot, ending with a metric of 92 dB. This is well above average compared to all the amplifiers we have measured but way short of state-of-the-art:” later it states: “There are some lofty numbers like -130 dB noise floor is thrown around by Devialet. I am not sure how that is computed. Using analog Line-in for consistency with other amplifier measurements we get rather ordinary numbers:”
Interestingly, the numbers are very similar to that made by Paul Miller for the HFN review, although Paul’s amp did not fail during testing and he proceeded to measuring via the digital input, which offered even less noise and generally better measurements elsewhere. The comments in the HFN measurements side bar stated “The vast majority of users will, I presume, engage with the amp digitally either via USB or wired/Wi-Fi network, in which case a 0dBFs input realises exactly 100W/8ohm and 200W/4ohm at the ‘0.0dB’ position (both at 0.0007% THD). Via the analogue input this increases to 120W/8ohm and 240W/4ohm, the near perfect ‘stiffness’ of the power supply reflected in the 115W, 230W and 450W delivered under dynamic conditions into 8, 4 and 2ohm loads. Note also how distortion barely increases with reducing load impedance [see Graph 1]. Couple this with the infinitesimally low 0.003ohm (3mohm) output impedance and, provided you don’t use scrawny cables, the 170 will rule your loudspeakers with the proverbial rod of iron. Through bass and mid frequencies, ‘digital’ distortion is half that for the analogue input, the former achieving 0.0005% as opposed to ~0.001% for the same 1kHz/10W/8ohm output. The opposite occurs at very high frequencies where the line input reaches 0.006%/20kHz and the digital inputs, regardless of sample rate, increase to 0.017% (10W/8ohm). The A-wtd S/N ratio is fabulously wide at 93dB and 98dB re. 0dBW (analogue/ digital) or 118dB re. 100W/8ohm while the response(s) are fl at to +0.14dB/20kHz, –0.6dB/45kHz and –1.6dB/85kHz (48kHz, 96kHz and 192kHz media).”
The same measurements are there, but ASR thrusts the bad at you, the HFN text is so bland that nothing really grabs your attention, but similar measurements are there. It is perhaps also worth noting that the HFN article ends with a “sound quality” score of 91%. I have no idea how this number is established, but as a regular reader of HFN I know that scores of over 90% are VERY rare. ASR state that using the analog input provides consistency with tests of other amplifiers, HFN state that it is assumes that most will use the amp via digital input, both valid points, but placing the emphasis in different directions.
Same numbers, different language, and we being human react accordingly.
It is the same story with the frequency sweep test. Despite the fact that the amp failed during this test, ASR are happy to state “Turns out this amplifier cannot reproduce high frequencies reliably at 5 watts!”.
Thanks to this, a number of us have started thinking about this, posting, asking questions and speculating as to why a Devialet Expert happens to have restricted power at high frequencies. Yet, this information has been in the public domain for well over a decade, with Stereophile reporting “By contrast, the D-Premier, given a mix of 19 and 20kHz tones at a peak level of 20W into 4 ohms (the highest level the amplifier would deliver with this signal without the power supply collapsing), performed well on this test (fig.9).”
This is a classic case of hiding in plain sight. I have read the Stereophile review, I have read the above comment, and thought nothing of it. Again, a bland statement that goes unnoticed, but when ASR thrust it in your face, we start thinking… Think about this, that statement has been out there for maybe 12 years now, and nobody has cared. I guess if nothing else this demonstrates that it genuinely does not matter. I have never heard of anyone complaining that their D-Premier kept shutting down, but it is fascinating how the words, not the measurements get our attention.
Thinking about how words can influence something as objective as a measurement, I thought I would have a go at it myself. In terms of amplifiers, the ASR measurement king at the moment is the Benchmark AHB2. Stereophile have also measured this amp, and no doubt it displays incredibly good performance, truly state of the art, for a product that is not overly expensive by “high end” audio standards.
https://www.stereophile.com/content/benc...-amplifier
Some ASR contributors have noted that you could buy an AHB2 based system for a lot less money than a Devialet. Seems a fair point to me, state of the art measurements, good price, why not?
Then we have that pesky subjective thing. Some of the subjective comments for the AHB2 have been a bit mixed. OK – objectively you could say that maybe it is just that some week minded folk who subjectively do not like the sound of the objectively measured superiority of the AHB2. They actually like the distortions other amps make. This might even be true. To be honest, I cannot judge, I haven’t heard one myself. (I would love to listen to the AHB2, BTW – If someone wants to lend me a couple)
But maybe it is not true, maybe there is an objective reason a Devialet could sound better? Here goes, true objective measurements coupled with made up by me in your face language.
It is no wonder that the AHB2 sounds dreadful in comparison to a Devialet, OK the AHB2 has fractionally lower noise and distortion, but at the kind of levels we are talking about here any notion that you could “subjectively” hear this as “better” is for the fairies. What really matters in terms of amplifier performance is the low output impedance, this any idiot could clearly hear in terms of accuracy and transparency in terms of how your amp is controlling your speakers. And look at the numbers, these don’t lie. The Devialet has a measured output impedance of 0.001–0.004ohm, this is state of the art. The Benchamark measures 0.034–0.17ohm, this is an order of magnitude worse! No wonder the Devialet crushes it for performance.
OK – The above text is something I have made up for a bit of fun. As I said, I have not listened to an AHB2. Who knows, if I did I might imediately sell my Devialet and swap. Or maybe not. The measurements in the text above are real, taken from Paul Miller’s AVtech, and the Devialet is the better of the two. I was chatting to someone once who works for a speaker manufacturer, the topic was that I subjectively thought that the Devialet was just about the best amp I have heard in terms of bass depth and resolution. The speaker guy suggested that low output impedance is the key here, it gives the amp better control over the speaker drivers. And what about other things, slew rates, input impedance, transient response?
So there might be something in my words above. There is a recent thread on Audiophile Style:
https://audiophilestyle.com/forums/topic.../#comments
Some interesting posts in the many pages, but it strikes me that although the industry, magazines etc. are very good at taking measurements, we do not really have much of a clue what they mean for sound quality. As an example, we can look at the noise level of a Devialet via analogue versus digital input, and the differences are clear, and yet run an external DAC into a Devialet via the digital inputs, it sounds dead quiet. But how is it affecting the sound? I know mine sounds a lot sharper, clearer and more detailed running digitally rather than fed via my AVR. But what about slew rate, jitter, input and output impedance? I love measurements, data and all the rest of it, but I would need to listen as well to have any degree of confidence as to what it means. But one thing has become very clear to me, the language and words surrounding these little measurement numbers have huge power and influence over many of us, and I would have to include myself in that.
One irony in all this was that some people on this forum we a little disappointed the read that Devialet’s Mathieu Pernot uses measurements during product development, rather than listening. I recall him saying something along the lines of “the measurements are the same first thing Monday morning as last thing Friday night”, implying that subjective hearing will vary, measurements not. It is hard to disagree with this notion. This makes me wonder just how well Mathieu might understand what measurements actually matter, in terms of actual delivered sound quality. I have no idea, but my subjective view of the Devialet hints that he might actually be rather good at this. In terms of R&D, there must be compromises between aspects of design, do you optimise what really matters to sound quality or what gives great headline numbers? A fascinating topic, but one I can only guess at.
So what of Devialet? Now we have the calls for them to explain why an Expert can only deliver 20W or whatever at high frequencies. I might be wrong, but I suspect Thumb5 is along the right lines when he suggests this could be an inherent limitation of how ADH works, and that it does not matter one iota in terms of actual performance delivering music.
Do I expect Devialet to respond and clearly explain all this? OK – If Devialet respond directly stating something like “yes, it is an inherent aspect of ADH, they can only produce 20W at 20kHz”, then I would be so surprised I would gladly eat my remote. If they do respond, I would expect something more along the lines of “The ADH system is not like conventional amplifiers and it includes sophisticated protection mechanisms that can react to the input signal both dynamically and in relation to frequency response, for this reason the amplifier does place some restrictions on frequency sweep type measurements that are of zero significance in normal use.” As for the other measurements that Devialet quote, you can find more or less equivalent numbers in the Paul Miller tests, but you have to be selective, and the Devialet quoted numbers are typically absent of the associated qualifying factors, so they can relate to “selective” measurements. A manufacturer selectively quoting numbers to make the product look as good as possible? It has been seen before.
One thing to keep in mind, the numbers measured independently by Paul Miller are good – and hey – it scores 91% for sound quality. We should all try to keep in mind that we should accept the measurements for what they are and what they mean, and if we are interested maybe we should try to understand better what they mean. Reacting to possibly inflammatory language that may surround these measurements can lead us astray, and remember, these words can have real power.
Anyway, I have surprised myself by how much I have written here. Time to listen to my little Devialet for a while, whilst we all wait for Devialet to respond to all this.
1000 Pro - KEF Blade - iFi Zen Stream - Mutec REF10 - MC3+USB - Pro-Ject Signature 12
