Yes, the tests of class D amplifiers with AES17 low pass filters makes them look better than they are and such measurements are incomparable with 90kHz BW distortion measurements of linear amplifiers. It is nothing but business and the industry insiders will keep repeating that measurements with BW>22kHz are unimportant and that 18.5+19.5kHz IMD is telling the whole story. It is not and the explanations are nothing but excuses.
All these discussions about how an amplifier should clip miss the point as far as I'm concerned: an amplifier for domestic music listening simply shouldn't clip at all. If it does, you have to reduce volume or build yourself a higher-powered amplifier. That's why a clipping indicator is a useful feature: if it never turns on during normal use, you know you have enough power, if it does turn on, you turn down the volume.
Not everyone is an audiophile ... they DO clip their amps ! High dynamic HT sources will (sometimes) clip , even as most of the program is
"normal".
OS
"normal".
OS
Most power in music is located at LF.
So measuring power should be at LF, with eye on whats happening in area where we are most sensitive around 3khz. (1-5khz)
Flatcher Munson at 60-80 Phone is where we are, when listening music at home.
If we have 89dB/2.83V speaker and we looking at 60 phon curve at Flatcher Munson,
for equal loudness at 50Hz and 3 kHz we have 2,83V at 50Hz and only 22mV for 3khz.
Maybe some kind of multitone test with respect to equal loudness.....
So measuring power should be at LF, with eye on whats happening in area where we are most sensitive around 3khz. (1-5khz)
Flatcher Munson at 60-80 Phone is where we are, when listening music at home.
If we have 89dB/2.83V speaker and we looking at 60 phon curve at Flatcher Munson,
for equal loudness at 50Hz and 3 kHz we have 2,83V at 50Hz and only 22mV for 3khz.
Maybe some kind of multitone test with respect to equal loudness.....
For me the ability to convert a (chip) amp to current drive for use with fullrange drivers is of such high sonic importance that absolute (chip) quality is overridden by the fact that this driving mode has its advantages.
So an old contender like TDA2003 / TDA2009 does the job of making the loudspeaker actively sound better than with a better spec voltage amplifier.
Can this be taken into consideration in the evaluation list?
"easy convertability to current drive"?
So an old contender like TDA2003 / TDA2009 does the job of making the loudspeaker actively sound better than with a better spec voltage amplifier.
Can this be taken into consideration in the evaluation list?
"easy convertability to current drive"?
Attachments
I tried to show my view and to introduce testing according IEC 60268-3 standard
https://www.audiosciencereview.com/...iers-according-to-iec-60268-3-standard.42488/
I think it covers everything what is needed to describe the amplifier. My attempt to test is here:
https://pmacura.cz/DIY_250W_4ohm_amplifier.html
I would be careful to emphasize a limited set of parameters and would never use a single one as a measure to compare various amplifiers.
Isn’t it more important that it recovers quickly and gracefully? High feedback amps will always clip ‘hard’ - surely the challenge is they exit clipping without ringing, oscillation or rail sticking?
OK, but even then it will only sound as good as running out of digital headroom. Unless it's managed at the source with a digital volume control, compressed programme material, and AGC — all that good stuff — then it's unpredictable and a matter of statistical uncertainty. A failure to adjust the volume ahead of time is precisely because you don't always know what will come out of the speakers in the next moment.
Maybe my example was a bit under-powered. If a clean 100W is needed, then 300W of buffer is not that much. It's only 6dB. I suppose for any given power level it could be argued: why not just do the same thing but without the soft-clipping? Because with the same resources, it's an open question which will bring better results, say:
A) 400W standard amplifier
or
B) 50W with a 'clean' 10W + 40W soft clipping.
With option A, it may be a lot of effort for a tight class-B design that manages a low idle current and low distortion at the same time, and lots of padding in case someone decides to run continuous 400W signals through it for testing purposes.
With option B, the whole thing becomes smaller and lighter, and you can have class-A with a high bias. And if anyone complains about subtle crossover distortion past 10W, well it's already clipping, isn't it?
Maybe my example was a bit under-powered. If a clean 100W is needed, then 300W of buffer is not that much. It's only 6dB. I suppose for any given power level it could be argued: why not just do the same thing but without the soft-clipping? Because with the same resources, it's an open question which will bring better results, say:
A) 400W standard amplifier
or
B) 50W with a 'clean' 10W + 40W soft clipping.
With option A, it may be a lot of effort for a tight class-B design that manages a low idle current and low distortion at the same time, and lots of padding in case someone decides to run continuous 400W signals through it for testing purposes.
With option B, the whole thing becomes smaller and lighter, and you can have class-A with a high bias. And if anyone complains about subtle crossover distortion past 10W, well it's already clipping, isn't it?
Well, we should not be too prescriptive. Maybe a clipping indicator might be a better way of dealing with this? I have a fantastic DG Claudio Abbado LSO recording (Stravinsk’s Firebird Suite) but the recording clips on 1 or 2 of the crescendos. Nevertheles, it doesn’t stop me from putting it on and really enjoying it. It’s the same for an amp - if it clips you will soon tell and then accept it, or reduce the volume. I understand some people like their music loud, but ultimately there must be a limit somewhere- it’s either the recording dynamic range, the amplifier power or your ears.
🙂
🙂
This is an implementation aspect and should therefore not be part of the AFOM I think.
//
This should not be an amplifier assessment criteria. The criteria would simply look at this amplifier as a black box and then report the results accordingly. How the amplifier is implemented, other than protection features, what topology, solid state or tube should not come into it.
What we are aiming for IMV is a simple set of measurements that can be applied to commercial and DIY amplifiers, that if largely scoring well, will assess the amplifier as all round capable, rather than the single dimension distortion figure objectivists have gravitated towards. People will certainly game some aspects, but that will hardly change their score if they have neglected other parameters in doing so.
As already agreed, the results will be presented in graph format with a top level figure of merit normalized to 0 to 10.
Any multicriteria optimization, according to Pareto’s law (which has never been refuted by anyone) is possible only according to one criterion.
Any “convolution” of a multi-criteria parameter space into one integral assessment based on “weighting” of criteria estimates is biased because it is based on a biased choice of the weight of each criterion.
All this only prevents people from engaging in free creativity.
You may gain in the power density of a Class D amplifier and in the temperature of the amplifier, but you will lose in something else (for example, in a non-flat distortion versus frequency curve, with a characteristic increase in distortion in the range of greatest sensitivity of human ears).
You can get incredibly low distortion and great dynamics from a Class A amplifier, but you will lose in amp weight, steady-state current, and temperature.
You can make an amplifier from high-end parallel op-amps, but you will lose in output voltage swing (output power). You can add floating power to this amplifier from a Class B amplifier, but you will lose the simplicity of the circuit design.
It is better to design simple good amplifiers (such as Elvee's Circlophone, such as many interesting solid-state or hybrid circlotrones) and not waste time on building dubious rating or estimation systems, especially since these rating systems are completely abstract, they are based on no real goals, they are an end in themselves.
I'm sorry, but I think so.
Any “convolution” of a multi-criteria parameter space into one integral assessment based on “weighting” of criteria estimates is biased because it is based on a biased choice of the weight of each criterion.
All this only prevents people from engaging in free creativity.
You may gain in the power density of a Class D amplifier and in the temperature of the amplifier, but you will lose in something else (for example, in a non-flat distortion versus frequency curve, with a characteristic increase in distortion in the range of greatest sensitivity of human ears).
You can get incredibly low distortion and great dynamics from a Class A amplifier, but you will lose in amp weight, steady-state current, and temperature.
You can make an amplifier from high-end parallel op-amps, but you will lose in output voltage swing (output power). You can add floating power to this amplifier from a Class B amplifier, but you will lose the simplicity of the circuit design.
It is better to design simple good amplifiers (such as Elvee's Circlophone, such as many interesting solid-state or hybrid circlotrones) and not waste time on building dubious rating or estimation systems, especially since these rating systems are completely abstract, they are based on no real goals, they are an end in themselves.
I'm sorry, but I think so.
Having been a part of a number of standards committees, I have one suggestion to offer to perhaps quell the dissent re: "ratings".
Consider not worrying about a "total" rating. Certain people do value certain characteristics above others, and their decisions are theirs to make based upon the criteria they value most.
Although a "spider / radar chart" can be a bit busy with 10 or so factors, to me, it seems a nice choice for this application if people would like to compare properties across amplifiers for multiple criteria. The scaling looks better if the minimum is 1 vs. 0 (just makes things easier to view), but that's not particularly relevant at this point.
Then... the discussion of which amplifier is "better" "in total" could perhaps be put to rest...
Example below....
Consider not worrying about a "total" rating. Certain people do value certain characteristics above others, and their decisions are theirs to make based upon the criteria they value most.
Although a "spider / radar chart" can be a bit busy with 10 or so factors, to me, it seems a nice choice for this application if people would like to compare properties across amplifiers for multiple criteria. The scaling looks better if the minimum is 1 vs. 0 (just makes things easier to view), but that's not particularly relevant at this point.
Then... the discussion of which amplifier is "better" "in total" could perhaps be put to rest...
Example below....
Looking at the frequency response of most speakers makes this a no no and can result in buzzing, uneven response and speaker burnout
Imagine (with solid state) where this no longer matters. You won't blow tweeters and it won't drive you from the room. Just like a
KT66 pair(valve) being over driven.
We need AI. I feel (subjectively) , that this will be hard to do.
Thanks!
I think with burst recovery it is about AC coupling of any kind. DC coupled NC252MP has stellar behaviour in this regard, but has other issues (capacitive load - rise of distortion).