Audio Power Amplifier Design book- Douglas Self wants your opinions

Status
This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.
Yes there are amplifiers that accept digital signals or internally do a A/D and at that point revert to digital processing.
Don't see why it is a sticky wicket?

The problem is that when you use the wrong term 'digital' for class D, you forego the fact that it is useless to try to design a class D amp unless you are very, very good in analog design.

One of the nice things with A/D and D/A designs is that it taxes the designers abilities in both analog and digital design, which makes it pretty hard to design a first rate DAC, for instance.

One issue that still makes it difficult to design first rate class D amps is that there are very few designers in audio that are experienced in high speed analog design. Feedback and control in class D is also something that goes a bit beyond the ubiqutous (sp?) miller comp cap ;)
And if you look at the issues with grounding and power supply routing in class D, that's an analog issue if there ever was one, and very critical.

Edit: I think what I want to say is that when you are very good in analog design you can probably design a passable class D amp on first try. If you are a digital systems designer, no way you will be able to design a good class D amp!

jan

It is always interesting to debate the analog vs digital nature(s) of class D, and also the nature of its sampled-data quantization. PWM and sigma-delta class D are both quantized, but in somewhat different ways.

I have often said that when something is wrong with a digital circuit, it is usually a problem that is analog in nature. Signal integrity of digital circuits, especially at very high speeds, is an analog issue. In my day job I work with LVDS digital signals on copper traces at 10-28 Gb/s, and it becomes basically an analog microstrip thing. A via stub 30 mils long can kill us!

It is also true that in class D there are distortions not unlike class AB crossover distortion as a result of switching times of the MOSFETs and critical timing between positive and negative turn-on times.

I think that Bruno is a good example of someone who has the necessary expertise and experience in BOTH the analog and digital domains to be successful in class D design, and his products certainly show it.

I guess if I really wanted to go digital in my book in the classical sense I would have material on the application and programming of DSPs for use in audio. Of course, DSP is already of fundamental importance to many of the class D schemes out there.

Cheers,
Bob
 
Well you ARE creative, I give you that. You have just described a sampled system. It becomes only digital if you code those step values in a digital byte or word and operate on that.
Like the output of an ADC - the input is analog, and the output is a digitally encoded value.
There are digital amps that take such a digital encoded value as input and those could be called digital amps.
Class D accepts analog in, gives analog out, and at no time is the internal processing anything but analog.

jan

There s a confusion between modelisation of digital circuitry
functionality and how it actualy works, wich is in fact all analog
circuitry....
 
For me, an analog magnetic tape contain sampled music. Can-you explain further what you mean ?

Easy . The recording only happens when the energy reaches a certain point of the magnetic curve ( top ) . Effectively like digital . Tim de Paravacini owns that idea I think ? The bias frequency is the sampling rate . As Tim said if only they had looked at that first digital would have been better . Better than digital it needs no playback clock . Too good an idea not to say it . Perhaps said previously ?
 
The strangest experience I had of this was a MM preamp I knocked together out of junk . Good junk . 10 nF 3 cents Mylar capacitors measure to 0.5% ( green 100 V ) . TL074 . The design was passive ( 75 uS ) + active . The 318 + 3180 correct to 0.2 dB . Identical between channels . Dead bug . The cartridge Linn K18 ( LP12 EKOS ) . I listened to it for 10 minutes and gave up . Some months later it was still running . My friend discovered it and announced how wonderful it sounded . Produced a Schroeder pick up arm and Tracy Chapman was in the room . Dammed if I know why it was like that . TL074 was what I had rather than a choice . Placebo or the anti Placebo ?

Some of the answers about good audio lie in that simple example, and of course the value of those types of experiences will be ignored, considered irrelevant by the vast majority, as always ... a major reason why I made the headway in the areas that I have interest in, is because I put a lot of effort into trying to understand some of these behaviours.

In general, audio enginers seem to dismiss the time element in the workings of a system. Give something a few minutes to warm up, and from then on the system behaviour will always be identical - subjectively, and objectively, nothing will change, whether in an hours time, a day, a week, etc - that's the common belief ...

If I thought that way I would have given audio away as a bit of a waste of time, not worth it, decades ago ..
 
Last edited:
In general, audio enginers seem to dismiss the time element in the workings of a system. Give something a few minutes to warm up, and from then on the system behaviour will always be identical - subjectively, and objectively, nothing will change, whether in an hours time, a day, a week, etc - that's the common belief ... If I thought that way I would have given audio away as a bit of a waste of time, not worth it, decades ago ..
Hum, by habit, we never unplugged our desks and power amps. Unless long holidays.
Lot of reason, including reliability.
Anyway, in the good old analog time, it was anyway long enough to set the bias and response curve of the 24 track analog tape recorder and the mixing one, before to begin a session, for the capacitances to be fully charged and everything stable in temperature.
Please, consider too that the situation is not at all the same than for a hifi system reserved to the reproduction: As we *create* the sounds, the only thing we have to rely on in terms of 'fidelity' is the monitoring system.
Some mikes change their characters with humidity and temperature, some don't like to be powered too long in front of the spittle of a singer and we are obliged to change them every hours (specially some Neumann).
Believe-me, when, in the morning, you feel a difference, listening to what you have recorded the day before, you begin to ask-you questions, about machines and even about yourself...
I remember, in my post production movie mixing theater, a sound engineer telling us that something had changed in the monitoring chain. He was right: during the night, we had added +1db to the normalized acoustic level (with our sound level meter) for the 0db of the desk. Go figure.
 
Last edited:
+1 dB is well within human resolution - typically not recgnized as a a loudness diferenece but rather modified frequeny balance - Bob Katz discusses in his blog, book


and in the context of this being the Doug Self Audio Amplifier Book thread - how about reading at least the last edition before "helpfully" bringing up "what most engineers ignore" that has been known for many decades, is treated in Doug's books, articles
 
Since the results of that will live "forever", it makes sense that the recording equipment should be fully stabilised before each take - it seems of late there is a lot of sloppiness around, especially in the classical field - a recording engineer said the 'time = $$$$' factor is a key issue these days ...

Borrowed a very recent recording of two violins playing numerous short pieces, on a respected classical label. The variation in tonal quality from one track to the next was enormous - I think I would have done a fair job of estimating how long from dead cold the recording chain had been switched on before the take of each track was done ...
 
and in the context of this being the Doug Self Audio Amplifier Book thread - how about reading at least the last edition before "helpfully" bringing up "what most engineers ignore" that has been known for many decades, is treated in Doug's books, articles
I presume you're talking of thermal behaviours, compensation and dynamics. Yes, they're part of the equation, but the time constants are minutes, or seconds - the worst 'offenders' are probably huge heat sinks, which will take considerable time to reach roughly stable temperatures. But I'm talking of other factors beyond the generation and dissipation of heat ...

Note that Self himself states:

It is often stated in hi-fi magazines that semiconductor amplifiers sound better after hours or days of warm-up. If this is true (which it certainly isn’t in most cases) it represents truly spectacular design incompetence.

Since some very highly regarded amplifiers have suggestions even in their manuals about 'burn-in' time, stretching over days or weeks ... and it's an accepted part of high end "knowledge" - there's obviously a lot of incompetence floating around ... :)
 
Dave Zan said:
I could be incorrect, perhaps people will say "yes, I checked my impressions with a blind test to remove placebo effect and I found..."
Perhaps JC will do this with Blowtorch vs a simple JRC4558 preamp .. and MikeK will admit that after my 50 page monograph on VAS/TIS, his own pontificating adds zilch to the store of human knowledge ... perhaps before the end of this Millenium ..
___________________

Why do you use the dual feedback network? You are creating unecessary complexity. I can assure from a practical point of view, that the single network works exceedingly well. The sx and nx-Amps are 100% DC coupled.
As I've said, my interest is in the Low Noise potential, simplicity and what you reported about DC offset in your CFA designs.

Alas, the 3rd requires evil diamond inputs which negates the first two. :mad:

I can't see how to use a 'single network' without even greater problems with DC offset. Besides my 'dual feedback network' is single at audio frequencies but at the cost of evil caps. :mad:

(I'm actually quite impressed by Ovation SX. It's simple and quite LN though not as LN as I'd like. If I ever did a CFA for real, it would be based on this.)
______________________

I don't adopt new topologies or extra complexity for religious reasons but only if I need something that tried & tested stuff can't provide.

This includes 'sound'. Yes, even the true golden pinnae, (and my Blind Listening Test panel has included some of the best ears in the business) are as prejudiced as the most rabid Golden Pinnae reviewer. But unlike the pseudo Golden Pinnae, the true golden pinnae can reliably confirm their prejudices in Double Blind Listening Tests bla bla ..

From experience, I've learnt to take their 1st impressions of something new very seriously.
__________________

Bob's different attitude towards CFA and Class D is IMHO, an example of this pragmatism.

He can see stuff that Class D potentially provides that he wants. Yus CFA fans haven't shown him anything that he wants or thinks worthwhile.

My own take is that Bob can't give either subject the depth and authority evident in his analogue chapters ... cos he hasn't burnt solder for 'real life' Class D or CFA.

For Class D, you need Bruno cos he's burn't loadsa solder. Dunno who is the person to do this for CFA. Any takers? :D
 
RNMarsh,
I think kgrlee is alluding to who is qualified or knowledgeable enough to write about those subjects. For some reason I would think you would be more than qualified but perhaps not interested enough to write an entire book on the subject of cfa feedback type amplifiers? As for class-D I'm not so sure that there are that many diy folks that would want to take that on even if there was design information. There are some threads here but I think those are more to do with commercial products than design of an original circuit.
 
Disabled Account
Joined 2012
When the winter comes and I am stuck inside somewhere is the time I do my best thinking, tinkering and writting. I will think about an article though - or multi-part article to get the most important elements of CFA laid out. But a whole book? I dont know. I did just get some test equipment that would do the subject some justice (ShibaSoku AD725D and Audio Precision 2722 etc)..... get back to ya on that one.... but I agree someone needs to start one on the subject..... long over-due.

Thx-Richard
 
www.hifisonix.com
Joined 2003
Paid Member
kgrlee

The offset is LOW. For both SMPS. The sx goes from ambient to circa 55c in about 20 to 30 minutes. The initial offset is about 50mV, but with 2 or 3 minutes its at 20mV and just gets smaller until its at operating temp where it's around 1mV. This of course is because I adjust the offset to 0 mV after the amp is warmed up.

The nx-Amp is even lower. It starts out at 24 mV and settles to 0 mV after about 10 minutes.

To use a single feedback network, keep your buffer degen resistors low and feed to the junction - just like the IC version, although they don't use the degen R's.
 
When the winter comes and I am stuck inside somewhere is the time I do my best thinking, tinkering and writting. I will think about an article though - or multi-part article to get the most important elements of CFA laid out. But a whole book? I dont know. I did just get some test equipment that would do the subject some justice (ShibaSoku AD725D and Audio Precision 2722 etc)..... get back to ya on that one.... but I agree someone needs to start one on the subject..... long over-due.

Thx-Richard

Interesting thing is that writers of books about audio power amplifiers are ignoring CFB but numerous companies (including high end like Krell) are using it. That's paradox, isn't it? But there must be something good in CFB since companies depend on sales and sales depend on human consumers.
 
Interesting thing is that writers of books about audio power amplifiers are ignoring CFB but numerous companies (including high end like Krell) are using it. That's paradox, isn't it? But there must be something good in CFB since companies depend on sales and sales depend on human consumers.
Commercial designs have a prime goal: Get acceptable results with as few parts as possible. CFA meets the criteria of fewer parts. There you go :)
 
As for class-D I'm not so sure that there are that many diy folks that would want to take that on even if there was design information.

Now that class D is big thing of the future, it's unlikely that the competent designers will be ready to give free gerber files to diy community and lose income. Since much of class D is specialized knowledge about layout, routing, etc., there is not much there for diy-ers, unless simple connecting of modules to PSU is counted as diy.

(Bonsai is one exception with his free gerbers for Ovation nx and sx CFB amps. Much obliged!)
 
Status
This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.