Hi Bob,
Happy new year!
It all depends on what you want, of course. If you want to make class-AB amplifiers with a distortion performance that cannot be obtained with class D, it might help to use a high-order feedback loop like the class-D guys and girls are doing. If you only want a distortion level that is a couple of decades below the threshold of audibility, then low-order loops will do fine, as shown in this thread.
Conditionally stable high-order loops with analogue loop filter are made by the millions: any high-order sigma-delta ADC has such a loop. Compared to a fully analogue amplifier, the nonlinearity of the quantiser just makes analysis more complicated.
Best regards,
Marcel
The efficiency gap between D and AB is not as great when tested at full power. Like I said, the greatest difference and need for higher effec. is at idle and lower levels... required for portable/mobile devices but it is Not an issue for home audio for me and a few others.
As far as high power home audio of very high performance is concerned.... class D will certainly evolve and get better but it aint there Now. And, it needs to be as cheap or cheaper to gain market share while being as good performance.
We cant compare just full power output levels for distortion. Low level distortion is very important and relavent whether Class D or A or AB or B etc.
Thx-RNMarsh
HAPPY NEW YEAR !!!
Last edited:
Timely suggestion Marcel! In fact I have such a beast (4th order compensated class AB) in the pipeline for Vol 7, but may have to move it to Vol 8. Measurements on it pose their own challenges!
Happy New Year all!
Jan
This is so but certain types of 'noise modulation' may be as good as proper TPDF dither in practice.
We first found this conducting Double Blind Listening Tests for Greenfield, Efficient Filter Design for Loudspeaker Equalization. Their box could switch out the Excess Phase Delay EQ leaving just the Minimum Phase Amplitude EQ.
To our surprise, all those who could tell the difference, reliably preferred NO Excess Phase compensation. We found that the Minimum Phase EQ, done by a large IIR, was 'self dithering', while the Excess Phase EQ (a large All-Pass) was an incorrectly dithered FIR.
Greenfield & Hawkesford reported this in On the Dither Performance of High-Order Digital Equalization for Loudspeaker Systems.
We were miffed cos we first brought it to their attention but they didn't acknowledge us. They certainly weren't aware of it before we told them. 😡
I think the first formal use of this was probably the noise shaped 14b oversampled DACs used in the early Philips CD players. One possible reason for their better performance compared to the early 16b Sony CD DACs.
It's not 'true' dither cos the 'noise' dies away eventually ie a form of 'noise modulation'. But the audible effects are excellent and give the expected sonic benefits of 'proper' dither.
I think Lipsh*tz & Vanderkooy did their usual full investigation on the subject. I'm not sure if this is part of their diatribe against SACD. I was in the bush by then.
________________
abra, what type of test signals were you using to test for 'noise modulation'?
Last edited:
Not any I know of, but the AES conference presentation does give some info: http://www.hypex.nl/docs/papers/globally.pdf
There was also something in Bruno's Burning Amp presentation I believe in 2011; I will see if I can find that.
@Bob C: Fig 23 is your fav here: the HF IMD test.
Jan
I'm not sure you're comparing apples with apples here - in the S-D modulator case the dither's going to be applied at the quantizer which is typically quantization to 6bits (sometimes fewer). What IIR or FIR structure did you listen to which is only 6bits wide?
Not devised any signals myself yet but I've inspected others' tests and spotted the tell-tale signs. A fade to noise of a single sinewave shows the effect nicely when the THD+N residual is plotted against stimulus level. See the 2nd plot down at this link : Weiss Medea+ Measurements
I'm just making a guess that the stuff I observed, all on 16b integer DSP chips of the early 90's would be applicable to dis new fangled stuff. It was self dithering behaviour
This is precisely the test we used to confirm our suspicions that the large IIRs were self dithering.
My definition of a properly dithered digital path is one where you can take a sine wave or piano recording 20dB below the noise floor and still hear a clean but noisy sine or piano. No 'chirps' in the noise on each piano note 🙂
If you are doing THD, you can see it rise well before you reach the noise level on a bad system. And you hear the crackly quantisation distortion/noise on each piano note too.
If I believed that switching amps was the best path, then that would be where I would focus my interest. From my experience there's still a lot of ground t be covered. Like CD is still getting kicked by analoge, like streaming (even High def) still has years left to actually musically compete with CD, then digital amplifiers also need years of refinement to really be a serious contender. I have doubts that they ever will, for low power Yes for sure, for bigger voltage swings NO. or MAYBE
Thanks, Jan.
Interesting, a class D that work in voltage follower.., I also applying feedback on the output, but it need some reference path before the filter because the ripple is too small if reduced, voltage follower mode is way to OK, because the ripple isn't reduced.
Anyway, Bob C says that the measurement of class D in HF is unfair because of the output filter
.Attachments
jan.didden
Not any I know of, but the AES conference presentation does give some info: http://www.hypex.nl/docs/papers/globally.pdf
There was also something in Bruno's Burning Amp presentation I believe in 2011; I will see if I can find that.
Amplifier class UcD
http://zalil.ru/34878709
best regards
Petr
Not any I know of, but the AES conference presentation does give some info: http://www.hypex.nl/docs/papers/globally.pdf
There was also something in Bruno's Burning Amp presentation I believe in 2011; I will see if I can find that.
Amplifier class UcD
http://zalil.ru/34878709
best regards
Petr
Thanks, may be too high loop gain for 3EF at low frequency., but lowering it may affecting another too.
@petr1951
Thanks Petr, that showing the wafeform. I am familiar with that wafeform, at low loop gain it has bigger drop than another modulating method but has linear drop naturally and lowest reactive current.
Last edited:
Happy New Year, Richard.
These are good points. Class D has come a long, long way.
One thing to keep in mind is that class D has made deep inroads into the home in the form of AV receivers. Most of those 5.1 and 7.1 channel receivers you buy have class D output stages in them. And for good reason: how else could you get 7 75-watt power amplifiers in a reasonably-sized and priced AV receiver without the use of class D?
Of course, there is a significant distance in SQ target from AV receiver to audiophile amplifiers driving audiophile speakers.
Cheers,
Bob
From where I'm looking at this thread it would be good to try to move on to 3rd order compensation or higher. At present I'm clueless on how to achieve this. No matter how many times I look at Cherry's NDFL paper I make no progress in understanding. 🙁
Quite a few designs on this thread have demonstrated single digit distortion levels at 20 kHz at high powers, and sub 1 ppm at mid to low powers. I fail to see what higher order comp schemes are going to bring to the party. We've been over this low distortion thing ad neauseum and I feel it's simply led nowhere.
More importantly, we should spend time and effort on construction, wiring, layout, power supply, decoupling and so forth.
Just my two cents worth and hope I've not trodden on any toes.
More importantly, we should spend time and effort on construction, wiring, layout, power supply, decoupling and so forth.
Just my two cents worth and hope I've not trodden on any toes.
Bonsai,
You're right and I agree. But its more from a technical / theoretical perspective. I'm wondering whether we can go for lower ULGF and/or loop gains with higher order compensation. Could this then lead to simplification in other areas?
I'm fully prepared to be corrected. This NDFL stuff is really bugging me and would like to gain some sort of appreciation for what it brings to the table (good or bad).
PS Not trodden on my toes at all 🙂
You're right and I agree. But its more from a technical / theoretical perspective. I'm wondering whether we can go for lower ULGF and/or loop gains with higher order compensation. Could this then lead to simplification in other areas?
I'm fully prepared to be corrected. This NDFL stuff is really bugging me and would like to gain some sort of appreciation for what it brings to the table (good or bad).
PS Not trodden on my toes at all 🙂
Can-you precise ?
I believe the comparator to be fully analog and no step in the width of pulses (pure analog width modulation) ?
As well as the feedback loop.
That's most of the people believe...🙂
I would love to make-you a nice blind demonstration between my huge 2KVA trafo, coper bars, big caps, followed (or not) by a zener limited cap multiplier +big caps and a little SMPS... to power my 2X150W CFA at normal levels on my 96db/w enclosures, you would be surprised by the results and your preference...
Ask Lazy Cat about ;-)
Member
Joined 2009
Paid Member
Yes, it's been rather chilly lately. Many years ago I experienced my first freezing rain in Albany (I used to live in NY State). I woke up to find a half inch of clear glass-like substance on my car windscreen. I attacked it with vigour and thought I'd cracked the glass - fortunately just the ice 🙂
About high order harmonics generated by class AB crossing, well, the idea to filter the output signal from them with a passive filter, (class D like) and use the same kind of feedback than Bruno used after this filter can be a good idea. Did somebody tried-it ?