[madwolfa]

You got me interested.. The SNR of 101 dB unweighted 2.8 vrms makes me think it could be usable for hard to drive headphones as well as speakers?

 
Found some excellent read on Class D ampifiers.
 
http://www.analog.com/library/analogdialogue/archives/40-06/class_d.pdf

 

[Jason Stoddard]

To answer the questions about our prejudice against Class D, we'd have to go into both practical and philosophical domains. 
 
In the practical domain, we have two problems:
 
1. Radiated noise and FCC compliance. Although we're past the "exploding parakeet" stage, high-frequency pulse modulation at high amplitudes is going to generate a ton of noise. To the best of my knowledge (and based on some experience), many, if not all, high-power Class D modules are not FCC compliant. And, even if they are, they still have to be tested after they are put inside of an amplifier. This is why you'll see a lot of modules sold, but not a lot of finished amps. And, as I've mentioned before, FCC compliance is really the one thing you have to do for sales in the USA. Otherwise, very bad things (nasty visits, big fines, business shutdown) will happen. We'd prefer to avoid this.
 
2. Nonlinear control loops and high feedback. Yep, there are some modules that have good measured performance, but this is thanks to tons of feedback--same as with most other amps that have good static measured performance. However, in this case, you're talking nonlinear control systems and z-domain analysis, which has different stability criteria than analog control systems (that is, typical multiloop amp feedback. Again, something we'd prefer to avoid. Add to this the fact that these nonlinear control algorithms are usually the "secret sauce" of the Class D amp designer, and it makes us even more nervous about "black box" applications
 
In the philosophical domain, Class D amps are mathematical transformations of analog to pulse modulation, which is not really congruent with our approach of "preserve the original bits/signal." It's the same as our objection to DSD. 
 
What would be interesting is a comparison of Class D designs with a multitone test signal (not just IMD, but something with, say, 8 sines) to see how the nonlinear control loop deals with it. But even if it did well, we'd probably shy away due to the radiated noise problem.

 

[Tuco1965]

Thanks for the explanation.

 

[Makiah S]

 
Now, for all of you who think Mike is a Yoda-like sage who comes in with words of wisdom and a perfect plan, consider this conversation…
 
Final word: Tony. To this day, Tony is our lead tech. He’s tested and/or programmed tens of thousands of products. Pretty good after Mike’s first intro, right?

... I still think mike is yoda, actually this is who I picture when you say mike 
 

... and for me the image is be fitting 

 

[Makiah S]

 
But, you know what? The main reason we went with a balanced design is that, in our experience, balanced designs offer better sonic performance than single-ended designs…as long as it is a purely balanced design.
 
 
Yes. True. And that comes down to the First and Second Laws of Audio, as esposed by John Chen at Grado:
 

• You can never have too much power.
• See the first law.

And it is for that reason I went to all balanced solid states amps and never looked back, they really do sound wonderful! 

 

[Maxvla]

What would be interesting is a comparison of Class D designs with a multitone test signal (not just IMD, but something with, say, 8 sines) to see how the nonlinear control loop deals with it. But even if it did well, we'd probably shy away due to the radiated noise problem.

I hope you pursue this when you get time. I've always been interested in this class of amp and wondered what could be done with it if serious effort was made to produce a high end Class D amp.

Going back to your previous post from yesterday, talking about advantages/disadvantages of design, aside from the radiated noise, are there any other disadvantages, and aside from efficiency (and reduction in size/weight/etc therefor) are there any advantages? Things such as power supply or part count/quality or layout complexity, protection circuit incorporation, inherent balancing, and whatever else you can think of.

When I got my most recent smartphone one of the things I tested was the audio quality. The phone I ended up liking the best as far as audio (and all the other features as well, as it happens) was marketed as having a high power PWM amp on board than is usual for such applications. I don't know if this actually affected the performance or not, but I've tried many friends' phones and several demos at stores and nothing sounds nearly as good. I can even run HD800s from it and it sounds 'alright'. It's an HTC DNA, btw.

 

[Armaegis]

Digital music formats used to be bad, as were switching power supplies, but both have come a long way. I'm sure Class D is getting there (if not already).

 

[ab initio]

The reason to avoid class D amps is because the test poorly. In fact, they're barely a passing grade. Everybody should strive to get an A.
 
Cheers

 

[walfredo]

  Digital music formats used to be bad, as were switching power supplies, but both have come a long way. I'm sure Class D is getting there (if not already).

 
They are there already.  peachtree and bel canto are two good examples.

 

[davidsh]

Jason, by noise are we talking about noise leading to high noise floor and bad SNR?

 

[BeyerMonster]

 

  Digital music formats used to be bad, as were switching power supplies, but both have come a long way. I'm sure Class D is getting there (if not already).

 
They are there already.  peachtree and bel canto are two good examples.

 
IIRC, Bel Canto just repackages the B&O Ice modules at a pretty good markup. Nothing wrong with that, but they aren't really doing the innovation themselves.

 

[Maxvla]

Jason, by noise are we talking about noise leading to high noise floor and bad SNR?

He means ultrasonic noise (above human hearing) around the unit, not something that would affect playback, unless you were using a tube amp and the tubes picked up the oscillation in the air.

 

[RudeWolf]

Both Peachtree and Bel Canto use ICE modules. Not bad per se, but you can do better with Class D.
 
As for better ones there isn't only Hypex. "New Class D" also look pretty respectable. If we look at the traditional suite of tests, then they tend to do very respectably. However I'd be very interested to do the tests Jason proposed.

 

[UmustBKidn]

For some reason, this chapter reminds me of this joke:
 
If Architects Had To Work Like Software Engineers
 
Also, every time I see the word "Circlotron", my feeble mind confuses it with the Cyclotron, a tube device. Go figure.
 
Finally, I am somewhat gratified to see the number of engineers who replied with some sort of engineering issue. Your audience is composed of many more engineers than you imagine.

 

[Ableza]

Circlotron topology always reminds me of something that was drawn up after a couple dozen beers by a group of engineering grad students.  You know, the sort of thing that usually fails the smoke test after uttering the phrase, "Now watch this schiit!"