Been playing with a couple of these Class D amplifier modules. DC rails are +/- 67VDC. On a tone burst they will put out 225W RMS into 8 ohms. I just thought I’d measure the power consumption at standby. With two modules connected, the AC power from the mains is 31 watts! I think a lot of it is going into heating up the output HF filter inductor. After idling for 1/2 hour their temps get to 85 deg C in a 15 deg ambient. Not exactly the last word in efficiency! I think the switching frequency is about 375kHz so there might be a lot of skin effect losses in the inductor. Might be interesting to put on a bigger external inductor and see what happens.
To an extent, the quiescent power consumption of a class D amp scales linearly with the power supply voltage. At the same time, the maximum power output of the amp is directly limited by the power supply voltage. Normally, this means choosing the balance of maximum output vs. efficiency at low output levels. Indeed, I’m working on an audio project at work involving precisely this tradeoff, and I thought to myself, “you’d actually want to vary the supply voltage depending on the power requirements”, and I thought about using the MCU, which will already control volume digitally, to control the voltage of the PSU. Right after this, I randomly came across some Texas Instruments reference designs for something even better: they call them “envelope tracking power supply” designs. They simply use some op-amps to prescale the audio signal into a feedback signal for a DC-DC converter, causing the supply voltage to closely track the audio signal. Apparently this is used extensively in some RF applications, but is practically unknown in audio. I found some research papers discussing it, with quite poor experimental results, but the TI designs from just a few years later have excellent performance according to their test results.
Maybe one of TI’s designs could be adapted to your application.