I think Lipshitz and Vanderkooy equalized it within +/- 0.1 dB for double-blind tests meant to test for something else, as some subjects could hear 0.2 dB errors. I'll see if I can dig up their article 'The great debate: subjective evaluation'.
I think it is better to make it the best as possible and the threshold of average people about THD, S/N ratio, etc. should be treat as reference only.
If it is for DIY, it should be not too complex. Of course, it is subjective but you should decide it. For me, input cascode, with or without current mirror, VAS buffered/cascode, and output 3EF is enough. But passive components are unlimited, you can use complex compensation.
If it is for DIY, it should be not too complex. Of course, it is subjective but you should decide it. For me, input cascode, with or without current mirror, VAS buffered/cascode, and output 3EF is enough. But passive components are unlimited, you can use complex compensation.
In the top graph, the power increments (driven from the source signal) are too large so the clipping profile is rounded at the bottom of the trough just before the amp goes into clipping. To fix this, reduce the source signal amplitude steps.
I agree with you - distortion vs power must be added. I’ll do that when I update the first post and also add the frequency response updates.
Good point re SOAR.
Re the speaker relay - that should read ‘electromechanical relay’. Solid state relays are what well designed amps use - for example, Accuphase.
HF IMD and FR separate the good from the ugly for class D, and these are criteria already taken into account. So for me, nothing to be changed as long as FR is measured with different loads.
Of course its does. A test should be defined in a DUTs external interface. A good test never anticipates or takes any inside aspect into consideration - it just explores a DUTs characterstics by observing it form the outside - i.e. the same as a user would have access to.
//
#9 is a usually a separate circuit , not really related to the core amplifier design/circuit.
#10 should be expanded to how the amp responds to failure , in general. What blows (even through error) , any design considerations and de-rating that would keep a design from becoming a fire hazard.
Rail sticking ?? .. I have designs that never stick ! Clip all day , not even device saturation ....
Also , the general "abuse tolerance" of the design.
Ease of repair ... if failure does occur.
OS
Maybe is sign of output power transistors thermal distortion drift efect if OPS is biased in AB class ? , or maybe is something wrong with PSU ?
I guess as a good designer you are aware of these things Ostripper. A lot of designers- both DIY and commercial- are a bit behind the curve.
Protection is included as the proposed standard will be applicable to completed amplifiers ie we are testing a ‘black box’
🙂
I way to catch this is to specify the amp runs at 1/3 power for 1hour before distortion tests are done. 1/3 power in a class AB amp is worst case power dissipation. This won’t really affect class A or class D, but they should also be preconditioned in this way.
Quoting Stanley P. Lipshitz and John Vanderkooy, "The great debate: subjective evaluation", Journal of the Audio Engineering Society, vol. 29, nr. 7/8, July/August 1981, pages 482 ... 491:
Page 484: "Our tests, and those of others, have shown that, under controlled conditions, differences in level or frequency response of 0.2 dB over an octave or so can reliably be detected by some listeners. Furthermore, a polarity reversal of the signal can also be discriminated on suitable, non-phase-garbled source material"
Page 489: "Our findings are that level differences of 0.2 dB are audible if present over a fairly wide band, and we therefore suggest that A/B tests must have linear differences matched to 0.1 dB. This represents an amplitude difference of about 1% as the minimum amount detectable by the trained ear."
So, 5 points when the differences including the loudspeaker model stay within +/- 0.1 dB, 0 points when they exceed +/- 0.2 dB?
By the way, for low-order harmonic distortion, Lipshitz and Vanderkooy claim that a few tenths of 1 % is audible on sine waves, about 1 % on music.
Last edited:
<1% H2 is audible on 1kHz sine, at proper level that must be below ear intrinsic nonlinear distortion. The limit is between 0.1%-1%, based on statistically relevant DBT test. I doubt that 1% H2 is audible with any kind of music, at any level. If someone states yes, I would like to see a link to relevant literature, with the experiment exactly described.
A shame ... with Class D and SMPS advancing so rapidly , class AB should be "updated" / refined to at least match the "curve".
As far as protection , what ? Standard uPC1237 triggering SS or mechanical relays ? With or without VI limiters ? Or both.
My point is that there are too many protection schemes , they need to be rated separately. Or , just ... does it have one ?
VI or any other ill conceived scheme could detract from the other points (THD).
As far as 1% THD , a subwoofer will approach that being a crude electro-mechanical device. Would we hear that ?
Amp at 40hz PPM would not matter.
For the bass , damping and other factors matter.