Remarkable article about amp distortion

[SpursGator]

This is my first post in years - I'm back just to share this.

I bought a Benchmark DAC1 the year it was released, long before Benchmark knew they were an audiophile company (Dec. 2002). I've been on their mailing list ever since and I never paid much attention. Today they sent me this. It obviously favours their amp, but this article is a remarkable read for any audiophile:

How Loud is the Distortion from Your Power Amplifier?

We took Stereophile's THD+N plots for 7 top-rated amplifiers and replotted the data to show the loudness of the distortion and noise at the listening position.

benchmarkmedia.com

 

[RJW232]

I haven't read the article in detail, but if my interpretation of the graphs is correct, distortion will be at washing machine level if the sound pressure is >115dB. Even rock concerts are limited to 103 dB SPL. At living room SPL's (80...90 dB), the noise+distortion is ~20 dB, probably lower than ambient noise in the average room.

 

[Jasonovich]

This is my first post in years - I'm back just to share this.

I bought a Benchmark DAC1 the year it was released, long before Benchmark knew they were an audio company (Dec. 2002). I've been on their mailing list ever since and I never paid much attention. Today they sent me this. It obviously favours their amp, but this article is a remarkable read for any audiophile:

How Loud is the Distortion from Your Power Amplifier?

We took Stereophile's THD+N plots for 7 top-rated amplifiers and replotted the data to show the loudness of the distortion and noise at the listening position.

benchmarkmedia.com

It's a great read, unfortunately my brain started to vegetate halfway through, you can call it grey cells distortion.
Distortion is all around, when you put your needle on the record, when you listen to your tube amp but not all of it is welcoming.

 

[SpursGator]

I haven't read the article in detail, but if my interpretation of the graphs is correct, distortion will be at washing machine level if the sound pressure is >115dB. Even rock concerts are limited to 103 dB SPL. At living room SPL's (80...90 dB), the noise+distortion is ~20 dB, probably lower than ambient noise in the average room.

What I thought was so interesting was how much more useful these graph plots were - especially looking at the different profiles of how Class D and AB amps deal with 4 ohms vs. 8 ohms. It would be so helpful to see plots of this nature for every amp. It answers so many more questions about how the amp performs than the data we are given as consumers - even though it's the same data! Just replotted.

The fact that Benchmark's own amp performs so well is practically a footnote. The AHB2 is already widely regarded as one of the quietest amps around.

RJW, you should look again more closely! 105 dB is a typical volume for movies, for which an awful lot of people use their amps. At that volume, distortion will clearly be audible from 99% of the amps on the market, especially in certain amp/speaker combos that happen to be very common (e.g., a class D 7-channel home theatre amp driving inefficient small speakers).

On the one hand, this shows how misleading the "0.00x% THD" stat is. On the other hand, it goes a long way towards explaining why different amps sound so different from each other despite appearing to have similarly microscopic levels of distortion on paper. Which I found quite enlightening.

 

[nopiano]

I get these too, and they are usually worth a read. Obviously this topic shows their kit in a good light, but they skate over the masking aspect, which I’ve read before is highly relevant in practice. Obviously I don’t have the tools to prove that, but I find it plausible.

Secondly, they seem to ignore the ambient noise in most rooms. Maybe the dB scale or weighting isn’t consistent, but 30dB is a quiet room, so I don’t think even a luxury washer is only 41dB, and there’s no way anyone would hear 8dB except in a test chamber.

Interestingly, a few designs of amp actually encourage harmless ‘white noise’ because it allegedly improves the sound subjectively. Go figure!

 

[SpursGator]

Interestingly, a few designs of amp actually encourage harmless ‘white noise’ because it allegedly improves the sound subjectively. Go figure!

I think these are commonly referred to as "valve amps." 🙄

 

[abacus]

What people forget is that most amp usage is around 1w (Which produces between about 85-90db from the speaker at 1m) , with the higher output being used just to make sure peaks are not compressed by the amp.
Also the louder the music peaks the more it will hide the distortion from the listener. (Effectively this is how Dolby noise reduction works).
Good article overall, but as yet nobody else seems to have taken up the idea, so it will be interesting to see if it becomes more mainstream, or just goes the way of the many new theories that have have appeared over time, as they don't give as much clarity as initially seems.

Bill

 

[Angle Poise-Lamp]

I don't see the point of obsessing over this. If your system sounds good to your ears, just get on with listening to music.

 

[ianrjones]

For me, THD is the most important measure, in making an "informed" decision.
Am I now being told that it is all "reletive"?

 

[nopiano]

I think these are commonly referred to as "valve amps." 🙄

No, I mean solid state. Models from d’Agostino for example, as measured in Hifi News.

 

[nopiano]

For me, THD is the most important measure, in making an "informed" decision.
Am I now being told that it is all "reletive"?

For anyone that swears valve amps are more ‘musical’ or records more accurate, then THD isn’t an issue!

 

[RoA]

For me, THD is the most important measure, in making an "informed" decision.
Am I now being told that it is all "reletive"?

If that is important to you then that is fine. I would argue that output impedance is just as important in an amplifier if buying 'blind'. Otherwise you have no idea how it will interact with your speaker loads.

 

[nopiano]

Just came across a slightly oblique connection to distortion in a social media post this morning. Sean Olive, the famous guy who researched speaker design in the USA, mentioned a publication from 2006 about perception of distortion in transducers (that is, speakers, in this case). A contributor went on to explain about his own tests on DACs where they matched the output to the millivolt. But they found listeners perceived the more distorting DAC as louder, so they used hearing to match perceived volumes.
Then they found guitar and voice sounded preferable on the distorting DAC to 70% of listeners. More ‘spectrally rich’ metal and rock was much preferred on the cleaner DAC, the previous favourite being called ‘unlistenable’.

For me this easily explains why we see polarised views on different DACs, and possibly amps and speakers too. Much depends on the levels and the musical choices.

I don’t think it’s a coincidence that hifi demos much more often feature guitar and voice, over say choral or opera. Not because everyone dislikes the latter, but because it’s so much more challenging to reproduce well.

 

[Jasonovich]

Just came across a slightly oblique connection to distortion in a social media post this morning. Sean Olive, the famous guy who researched speaker design in the USA, mentioned a publication from 2006 about perception of distortion in transducers (that is, speakers, in this case). A contributor went on to explain about his own tests on DACs where they matched the output to the millivolt. But they found listeners perceived the more distorting DAC as louder, so they used hearing to match perceived volumes.
Then they found guitar and voice sounded preferable on the distorting DAC to 70% of listeners. More ‘spectrally rich’ metal and rock was much preferred on the cleaner DAC, the previous favourite being called ‘unlistenable’.

For me this easily explains why we see polarised views on different DACs, and possibly amps and speakers too! Much depends on the levels and the musical choices.

I don’t think it’s a coincidence that hifi demos much more often feature guitar and voice, over say choral or opera. Not because everyone dislikes the latter, but because it’s so much more challenging to reproduce well.

Interesting read, here we are, thinking 1010101s and not appreciating that DACs also generate their fair bit (excuse the pun!) of distortion.

I wonder, the distortion that was measured, was it derived from erroneous sampling or did it emanate from the analogue components supporting the chipset or both?

 

[nopiano]

Interesting read, here we are, thinking 1010101s and not appreciating that DACs also generate their fair bit (excuse the pun!) of distortion.

I wonder, the distortion that was measured, was it derived from erroneous sampling or did it emanate from the analogue components supporting the chipset or both?

The details re the DACs that they quoted:- “…where one was orders of magnitude more THD-ful than the other (RME ADI-2 Pro and Dacmagic 100). Both were voltage matched to a microvolt.”

The AES Convention Paper 6910 was about the poor correlation between speaker distortion measurements and listening preferences.

We can measure gear better than ever, but it seems there’s quite a gap, still, in understanding or predicting listener preference from the data.

 

[Jasonovich]

The details re the DACs that they quoted:- “…where one was orders of magnitude more THD-ful than the other (RME ADI-2 Pro and Dacmagic 100). Both were voltage matched to a microvolt.”

The AES Convention Paper 6910 was about the poor correlation between speaker distortion measurements and listening preferences.

We can measure gear better than ever, but it seems there’s quite a gap, still, in understanding or predicting listener preference from the data.

My cognition appears to be on the wane 😏 Thanks for explaining.

 

[Gez]

I haven't read the article in detail, but if my interpretation of the graphs is correct, distortion will be at washing machine level if the sound pressure is >115dB. Even rock concerts are limited to 103 dB SPL. At living room SPL's (80...90 dB), the noise+distortion is ~20 dB, probably lower than ambient noise in the average room.

It's just a long winded advertisement for Benchmark amplifiers.

 

[Gez]

105 dB is a typical volume for movies

Anyone that believes that has never been subjected to music at an average SPL of 105dBA. Maybe, just maybe it might be a peak level when measured by a very fast SPL meter, but it'll be a peak of milliseconds duration max.

This is what the HSE (UK body) says about concert volume levels:

"Noise limits

Unlike workers, there is no specific legislation setting noise limits for the audience exposure to noise. However, HSE strongly recommends that the A-weighted equivalent continuous sound level over the duration of the event (Event LAeq) in any part of the audience area should not exceed 107 dB"

Noise - HSE

Information to help those organising events control or reduce exposure to noise.

www.hse.gov.uk

Now I don't know about anyone else here, but I've never been to any cinema ever, where the film soundtrack was anywhere near as loud as some music concerts I've been to (e.g. Iron Maiden).

 

[nopiano]

Now I don't know about anyone else here, but I've never been to any cinema ever, where the film soundtrack was anywhere near as loud as some music concerts I've been to (e.g. Iron Maiden).

No, I agree that soundtracks are generally lower level. Obviously films contain effects like explosions, or helicopters landing nearby, but they are brief peaks. ‘Rock’ music events often seem continuously loud and compressed, or at least they did when I last attended.

Last year we saw Hamilton performed in Cardiff and that was over-amplified for my taste. Luckily I had my earplugs* handy so I used them throughout.

* bought for watching Formula 1 racing!

 

[AJM1981]

Just came across a slightly oblique connection to distortion in a social media post this morning. Sean Olive, the famous guy who researched speaker design in the USA, mentioned a publication from 2006 about perception of distortion in transducers (that is, speakers, in this case). A contributor went on to explain about his own tests on DACs where they matched the output to the millivolt. But they found listeners perceived the more distorting DAC as louder, so they used hearing to match perceived volumes.
Then they found guitar and voice sounded preferable on the distorting DAC to 70% of listeners. More ‘spectrally rich’ metal and rock was much preferred on the cleaner DAC, the previous favourite being called ‘unlistenable’.

For me this easily explains why we see polarised views on different DACs, and possibly amps and speakers too. Much depends on the levels and the musical choices.

I don’t think it’s a coincidence that hifi demos much more often feature guitar and voice, over say choral or opera. Not because everyone dislikes the latter, but because it’s so much more challenging to reproduce well.

Point for sure. A little to counter; Hi-fi demonstrations aimed for selling gear to the largest average group of consumers would not do really well if they would mainly throw in Opera or Choral. Unless it uplifts the aura around their flagship model or something. Quite arbitrary.

Something a brand demonstrates is usually based on the average profile of their fanbase and what those groups expect of its image (or what a manufacturer wants to push it into). Also what fits to its kind of speaker design.

A brand like Fleetwood sound demonstrated fine live recorded Tango on their deville loudspeakers with loads of hall reverb and dynamics. It is not that other loudspeakers can't reproduce that, it is that that particular music has become synonym to that particular loudspeaker for their live or online audience. It contains woodwinds and has a certain texture, people see a wooden horn and get that direct click to the music mentally as well. It 'belongs' to that type of music. Same as with that woofer grille that almost resembles the decorations of a medieval lute. The 'art meets art' connection brings a desire to own or, at least, take note of that demonstrated loudspeaker.

The hi-fi world is data, voicing, room conditions, branding and a heavily opinionated audience. Some amps like the vintage Sansuis produce more bass and bloom on neutral, as this was to compensate for the lack of bass extension of their own brand bookshelf loudspeakers in the late 60s and 70s. This is why anywhere online people mention that it does not measure flat (true, and.. "accepted", because it has an enjoyable voicing signature) and praise models like the au505 'also' for low volume listening.

I can play opera once a year, but it's not enough to look for the best loudspeakers for Opera (maybe true horns are the big deal here, don't know, didn't dive into that) .

Big cases (not the tall floorstanders) of loudspeakers are usually best for types of classical music with a lot of "slam" and presence. The smallest of bookshelves are often great for that 'crisp' in accoustic guitar strumming (especially rhythm guitar tracks) and equally enjoyable for other treats. Forgot to mention the dozens of other models. It is all not in the 'perfect' sense, there are always minor exchanges.