Extend the S/N measurement bandwidth to (say) 1MHz. Disallow external filtering. That will prevent hiding errors as noise above 20KHz.
Ed
Back to back power mosfets with a photovoltaic optoisolator gate driver can make a nearly ideal switch without any contact pitting or welding issues
But with more distortion than a good electromechanical relay, like those Amplimo relays with a tungsten contact bridged by a golden contact - although you could maybe put the main feedback loop around the electronic switch to solve that again, if it needs solving.
In any case, I don't see why the type of switch has to be on the list at all.
In any case, I don't see why the type of switch has to be on the list at all.
One wonders how one might measure bass slam, or bass articulation, or soundstage height? Soundstage depth? How about air? Presence? Musicality?
So much to cover, so little time!
I wonder how my 'mythical' amplifier would fare?
-Power seems to be counted at least 2 or 3 times: things like PSU rejection or mains hum could be highly dependent on power level, as would distortion. I can see designers being pressured to game the system by advertising their amplifier in a lower power category. Or caving to some made-up requirement, like sizing the power supply for high power test tones rather than for actual music or movies.
What happens if you have something like 20W with the upper 15W having progressive soft clipping? IMO the whole 20W would be a lot more usable than a design which only has hard clipping, but the latter could easily get more points. #10 could probably be reworked for a more user-friendly and realistic assessment of overdrive. Rail sticking is basically an "unhandled exception" in the hardware, not a musical understanding of overdrive.
#3 — Surely at least half the effort, bulk and complexity goes towards making things stupid-proof. +1 for "active speakers", which is kind-off annoying if it means things are limited to plate amps welded to the back of their speaker.
#5 THD+N — +1 for noise. But unweighted THD would just add its own useless noise (pun intended) to the rating. It needs to be weighted, both as a function of absolute frequency (a la Fletcher Munsen curves and co.) and masking, prioritizing the audibility of high harmonics over low ones.
Harmonic distortion could even be subdivided into flavours, to keep the various H2 and H3 camps happy. Personally, I haven't heard enough real-world designs to have a strong opinion, other than a general suspicion that, for some people, 1% of pure H2 (or H3?) is less obtrusive than <0.01% HD where every sine wave has a faint but "harmonious" (chuckle chuckle) sawtooth wave buzzing in the background.
#7 "full power". As above. For a 100W amplifier, it would be reasonable to think of the 1st watt as most important for audio quality, and the rest is just 20dB of crumple zone so that "NCIS" gunshots don't clip. Although I can kind-of see how the quality of the gunshot effects are the whole point of some TV shows.
I wonder how my 'mythical' amplifier would fare?
-Power seems to be counted at least 2 or 3 times: things like PSU rejection or mains hum could be highly dependent on power level, as would distortion. I can see designers being pressured to game the system by advertising their amplifier in a lower power category. Or caving to some made-up requirement, like sizing the power supply for high power test tones rather than for actual music or movies.
What happens if you have something like 20W with the upper 15W having progressive soft clipping? IMO the whole 20W would be a lot more usable than a design which only has hard clipping, but the latter could easily get more points. #10 could probably be reworked for a more user-friendly and realistic assessment of overdrive. Rail sticking is basically an "unhandled exception" in the hardware, not a musical understanding of overdrive.
#3 — Surely at least half the effort, bulk and complexity goes towards making things stupid-proof. +1 for "active speakers", which is kind-off annoying if it means things are limited to plate amps welded to the back of their speaker.
#5 THD+N — +1 for noise. But unweighted THD would just add its own useless noise (pun intended) to the rating. It needs to be weighted, both as a function of absolute frequency (a la Fletcher Munsen curves and co.) and masking, prioritizing the audibility of high harmonics over low ones.
Harmonic distortion could even be subdivided into flavours, to keep the various H2 and H3 camps happy. Personally, I haven't heard enough real-world designs to have a strong opinion, other than a general suspicion that, for some people, 1% of pure H2 (or H3?) is less obtrusive than <0.01% HD where every sine wave has a faint but "harmonious" (chuckle chuckle) sawtooth wave buzzing in the background.
#7 "full power". As above. For a 100W amplifier, it would be reasonable to think of the 1st watt as most important for audio quality, and the rest is just 20dB of crumple zone so that "NCIS" gunshots don't clip. Although I can kind-of see how the quality of the gunshot effects are the whole point of some TV shows.
Might be of your interest as well:
https://www.listeninc.com/wp/media/2023/06/paper_51_AES_practical_implementation_of_PRB.pdf
Just to reiterate, what is being attempted here is a set of measurements that recognize the performance of an amplifier is not just how low the distortion is. You want a system that in most cases will tell you if it has been well-designed and engineered. It should be a set of criteria that anyone with a good understanding of the art and some measurement gear (scope, audio card) can assess.
Designers could game the system, but at best they score highly in one category. In the envisaged tests, they would have to game all areas to get a very high score, in which case they would have surely come up with a good amplifier?
Marcel, the only relay I know of specifically designed for audio is the Amplimo. In general, electromechanical relays are not a good idea for switching DC fault currents, as when one of the output devices goes short to rail. solid state relays using high power mosftets have been measured delivering a range of powers and distortion is effectively zero. There is a long thread about EMR's here https://www.diyaudio.com/community/threads/output-relays.191449/ and I believe the vast majority of builders here have gone over to this type of speaker switching. I note that commercial amplifier builders are also moving to SSR's - see Accuphase for example.
As far as I know, switching off DC is mainly a problem when the voltage exceeds about 30 V. My amplifier has +/- 21 V rails. In general, you shouldn't use relays to switch DC voltages they were not designed for.
I think there are two possibilities under single-fault conditions in my amplifier:
1. An output MOSFET fails, putting the full 21 V on the output. The loudspeaker's DC resistance limits the current, so the relay that has to switch off the output works within its normal ratings.
2. The output gets shorted. The amplifier then switches off its output stage (electronically, not electromechanically) when the current gets too large.
I have tested case 2, not case 1.
I think there are two possibilities under single-fault conditions in my amplifier:
1. An output MOSFET fails, putting the full 21 V on the output. The loudspeaker's DC resistance limits the current, so the relay that has to switch off the output works within its normal ratings.
2. The output gets shorted. The amplifier then switches off its output stage (electronically, not electromechanically) when the current gets too large.
I have tested case 2, not case 1.
Last edited:
Then you have an amplifier that works well up to 5 W, while it could have worked well up to 20 W if you hadn't included the soft clip feature.
The ultrasonic noise is hundreds of millivolts. An oscilloscope can measure that. My purpose is to close the loophole of not counting ultrasonic noise as noise.
Ed
Noise above 85 kHz is neither audible to you, nor to your dog, nor to your cat. As long as it is not so bad that it fries tweeters or causes bats to get disoriented, why do you care?
The standard is not proposing how the protection is implemented (other than SSR output switching) and just awards points for protection.
For Class D, it does appear that it is being treated in a special way wrt HF noise with the filters. If a linear amp produced artifacts at HF, it would be deemed unacceptable by most engineers. I suspect this is being driven by industry bodies with a stake in that market and certain companies.
One of the tests proposed is into a difficult load to certain SOAR capability. If an amp has aggressive VI limi, it will manifest in the associated distortion. That probably should be made more explicit in the test requirement.
So, this is an N+1 attempt to create a weighted one-dimensional representation of a multidimensional system quality estimation that claims to be objective.
Unfortunately, this is impossible even theoretically.
And in the end, it doesn’t matter at all what weights of what parameters someone assigns because that’s what someone wants, only one thing is important - the one who listens to music amplified by this amplifier is satisfied.
But yes, it is a fact that the closeness of a particular amplifier to an ideal linear system is a good design criterion.
Unfortunately, this is impossible even theoretically.
And in the end, it doesn’t matter at all what weights of what parameters someone assigns because that’s what someone wants, only one thing is important - the one who listens to music amplified by this amplifier is satisfied.
But yes, it is a fact that the closeness of a particular amplifier to an ideal linear system is a good design criterion.
For a class-A, -B or -AB amplifier, it would probably indicate some sort of malfunction. For class-D amplifiers, sigma-delta DACs and analogue tape recorders, far ultrasonic rubbish indicates that everything works as designed. I suspect that is the difference.
This is probably not something everyone will agree on. We could of course insist that all artifacts above 20 kHz in a linear amp be filtered out and then we’d have great 20 kHz harmonic distortion figures. If HF artifacts were not a concern in cladd D, why demand test regimes that filter this out?