How would you stop PSU pumping in a Class D Amp

[Colin James Wonfor]

How would you stop PSU pumping in a Class D Amp.

Most of the techy folk here will have come across this maybe with Tripath chips.

So how did you solve it?

I will put up and idea that works well and reduce component cost.
And no cheating by doing it Bridge Mode OK or limiting the bandwidth to above 120Hz .

 

[fas42]

Not something I've worried about - but power supply behaviour does interest me - looked around, and saw this:

Several things can be done to relieve power-supply pumping. The lowest impact is to operate the two inputs out of phase 180° and reverse the speaker connections. Because most audio is highly correlated, this causes the supply pumping to be out of phase and not as severe. If this is not enough, the amount of bulk capacitance on the supply must be increased. Also, improvement is realized by hooking other supplies to this node, thereby, sinking some of the excess current. Power-supply pumping should be tested by operating the amplifier at low frequencies and high output levels.

 

[DonH56]

I don't have much experience with audio class D amps. Last time I regulated the rails and paid a lot of attention to broadband decoupling (maintaining wideband low-impedance power planes). An interesting scheme I tried with limited success was to add a feedback circuit to track out the supply ripples. It worked, sort of, but was tricky to get stable over PVT and load variations. These days I'd probably just use a class-D PWM/PFM power supply since it incorporates regulation and the switching frequencies have risen well above the audio band.

Oops, just read your post again, no point in going further when I can just await "the" answer.

 

[amirm]

I never finished a class D amp I started (for car audio) but remember reading about power supply pumping in some app note where the current flows at the peak of the sine wave back into the power supply. This in turn charges the capacitors and hence increasing their voltage because the rectifier diodes are reverse biased. I suspect making the bulk capacitors larger would help as would running the two channels out of phase (assuming shared supply). But again, never tried and am interested to know the right solution.

 

[tomelex]

Provide opposite phased (from power supply droop) to the drive signal of the switching devices

 

[Colin James Wonfor]

I tried many way the standard big is best and so on and even a push-pull linear PSU so it could sink the over volts.

Then taken a big step back and the dawn smacked me in the face (forgot sunglasses).

Since at any one time only 1 device should be on and thus switching a wide or small width pulse we get a boost converter due to the output inductance and the switching device which in the case of the none switching FET acts due to the body diode and pumps the rail up.
Now if we could capture and invert this pulse and the dump on the HT rail that is not being pumped the problem goes away.
But doing it for both positive and negative rails could be fun.
I like the principle of "KISS" (Keep It Simple Stupid) and not "Harrison" (https://en.wikipedia.org/wiki/John_Harrison) he found design error but instead of restarting he add another complex part and soon on, until the error was so small to present no problem.

OK now I will find the cct that at the moment as taken a walk in a HDD (I hate confusers)

Back soon sorry mmmgrrrr