Nominally Fixed Frequency Ripple Reduced Class D Amplifier

[SMPSGuy]

This is DIY but may live elsewhere and I need to work on the title to get a nicer Acronym.

I've been dreaming about for this for some time and having gone through various iterations of ideas that ultimately failed I ended up with this one. Your caveat is that it is a Spice Model without the kitchen sink to get it working but I need to break it harder to see if it is worth building one.

It is an incremental thing in as much as I poke it to break it then look to see why it is broken and try to fix it. I have many more things to break and fix but I think I may be getting close. No doubt I will never be able to Iron Out The Rough Spots and it will never see the light of day but this is what I have at the moment.

Obviously numbers are fantasy but this it what it does at just over 800W into 4Ohms @ 1KHz normalized to 0dB. It's about 10V off of going into clip.

I'm sure things will only get worse in real life but this is the HF ripple. Target FS was 500KHz.

Again and on the basis that it is just a Spice Model without the warts here is the "Dashboard". Could be wrong but 6VP-P into 4Ohms @1KHz.

Apologies for the file names. I was trying something out to fix another thing.

 

[SMPSGuy]

An add after further head scratching. One of the big problems has been working out how to stop the modulator losing the plot up at 20KHz plus. A partial fix lets me generate some intermodulation plots. The first is 19/20KHz close to clipping,

Normalised to 0dB

Of course things are much nicer at lower output levels.

As a last one, until I sort out some other quirks, here's a plot of load dependency. 2R, 4R, 8R, 16R, 32R 4096R with 6Vp-p out. In this case, and as mentioned elsewhere, if I hit things too hard the modulator looses the plot.

This is with a second order 40KHz filter on the input.

Ooops. It is still overall inverting and I also might have problems with it providing a 'useful' response out to 100K.

Whilst I hate the term because I do not fully understand it we also get gain peaking which may have resulted from me adding a pole to the feedback loop chasing distortion numbers.

If I knock that one on the head, waits for the analysis to run, I get one of these,

So, if you believe it, that's nice(r). I have cheated and only plotted up to 250KHz

Oh. This is post filter feedback and in this case it's behaving itself without a load because I have used current feedback from the output filter capacitor to tame or kill the filter resonance.

 

[SMPSGuy]

Yesterday I got slightly excited because I managed to lock the modulator to an external signal not using a phase locked loop. Yes I have tried that and it is a road to fail especially if you expect any bandwidth from it. Most likely because I am doing it wrong. Then I poked about and it fell over so that will be another fail.

If I am gentle with it then I can get a plot of loop gain. Apparently LTSpice has moved on since Mike told me how to do it and rubbished the idea but I can get pretty pictures.

Without a pole in the Voltage Error Amplifier...

With a pole in the Voltage Error Amplifier...

I'll resist the temptation to assign any meaning to those. Oh that was with a 4 Ohm load.