Class D Amplifiers 101

[Offler]

You mean this distortion?

Precisely . The argument went like that amplifier which does not use triangle wave will always be better, or because "D" stands for "digital" which I later found out to be wrong.

Even relatively cheap and simple UcD102 has THD+N at -80dB.

 

[boXem]

...
Even relatively cheap and simple UcD102 has THD+N at -80dB.

Cheapness may be relative, simplicity is just amazing. This is the amp that made me become a Bruno P. fanboy

 

[Kokoriantz]

I tried out my idea on post 62 to see the feasibility. I simulate it at standard 400khz and 50ns dt. This is without correction.

The distortion.

A roughly adjusted gives

distortion.

 

[Kokoriantz]

What makes the amp better sounding at high switching frequencies is the very low ripple current. The kinks observed are separated by PP current of the ripple, running it at 2Mhz the the kinks will have shorter separation acting at very low power, very similar to under biased class AB. There the feedback acts as usual once at every zero crossing.

 

[AdamG]

Thread Sticked. Nice work @DonH56!

 

[pma]

I want to add on this reference thread the dead time issue in class D amplifiers.

Similar to class AB crossover distortion, though not exactly same. Odd and annoying in TPA32XX related circuits, when frequency gets higher and dead time is more pronounced.

 

[Kokoriantz]

Similar to class AB crossover distortion, though not exactly same. Odd and annoying in TPA32XX related circuits, when frequency gets higher and dead time is more pronounced.

It looks like the dt correction is voltage sense as it is optimized for 3ohms load.

 

[Kokoriantz]

Here is 10ns dt and 2Mhz Fs. The pair kinks got merged to make a smooth zero transition, only each polarity now is lacking an offset as the distortion is about 0.5%. The feedback can operate now much better in the lack of abrupt kinks, as the frequency is high 60db NFB can be easily applied to bring the number 0.0005%.

 

[Kokoriantz]

A suitable inductor can be ED0006 AL from Coilcraft

 

[JohnYang1997]

It looks like the dt correction is voltage sense as it is optimized for 3ohms load.
View attachment 153395View attachment 153397View attachment 153398

SE configuration has high noise floor giving "better" looking graph. Then the analyzer configuration is different for these graphs. The 3ohm one is 20khz BW. The other two are 80khz BW. If you read the 20khz BW ones only you'd see the distortion at 20W rises to 0.02% or so for all different loads.

 

[boXem]

How to ruin a good thread 101

 

[Kokoriantz]

How to ruin a good thread 101

Post 80?

 

[NTK]

How to ruin a good thread 101

Post 80?

I got a like from the thread starter

 

[boXem]

I got a like from the thread starter

And now you got a new one

 

[Karls_speaker_works]

How to ruin a good thread 101

Was interesting reading until you seem to have an issue - very interested in the technical discussion.

 

[boXem]

Was interesting reading until you seem to have an issue - very interested in the technical discussion.

Indeed. That is now completely out of place since the messages I commented have been deleted.

 

[Karls_speaker_works]

Indeed. That is now completely out of place since the messages I commented have been deleted.

Aaaah, my apologies.

 

[DonH56]

Was interesting reading until you seem to have an issue - very interested in the technical discussion.

@boXem | audio designs these things so I tend to pay attention to his issues. In this thread there were a few diversions out of scope of the (my) original intent, and he corrected a few things I and others said. I intentionally used a (arguably) simpler design that was not self-oscillating, because I thought it was easier to use in explaining the operation, and some folk took issue with that decision. Oh well. Some other comments (mine and others') revealed a lack of understanding of some very basic principles and/or a glossing over of some details (again, intentional in my case, to suit the intended audience -- which just means I didn't know it well enough to explain it simply). Some of the comments were just wrong, at least in the context of modern amplifier design. A lot of the discussion really belongs in a "Class D 201" thread or better, which would be an interesting read for me. Alas, I need to get back to my day job, which is running deep into the night... - Don

 

[Karls_speaker_works]

I design DC-DC converters for my day job, many of the issues discussed ring a bell for me and I've often found the similarity between DC-DC and Class D interesting, I see errors all over the place (surely not with just one poster), still was interesting to me.

 

[DonH56]

I design DC-DC converters for my day job, many of the issues discussed ring a bell for me and I've often found the similarity between DC-DC and Class D interesting, I see errors all over the place (surely not with just one poster), still was interesting to me.

Thanks! Any flat-out errors in the original post, I would love to correct. The rest of it, meh, hopefully there's enough "goodness" to let it ride.

A friend who designs DC-DC converters once commented that class D amplification was somewhat similar to modulating the reference of a SMPS design. Although I worked (briefly) there, I was not doing DC-DC converters, so was (am) not competent to comment.