Class-D amp設計

在Elliot Sound Produtct這篇中，提到

any error as delays or rise-time of the MOSFETs will ultimately affect efficiency and audio quality. So all the involved components must be high-speed. Dead-time also affects performance, and it must be minimised. At the same time, the dead-time must be sufficiently long to ensure that under no circumstance both MOSFETs are on at the same time. Typical values are 5 to 100ns.

Filter.......speaker wires can become an antenna and affect other equipment. In fact, although a couple of volts RMS of ripple can seem low enough to run your speakers safely, EMI can be a concern, so the less carrier level you have, the better.

Filter.....the response is dependent on the load, and in fact the load is part of the filter. .....only a handful of good Class-D amplifiers use feedback techniques that include the output filter to compensate for impedance variations and have a nearly load independent frequency response, as well as to reduce distortion produced by non-linearities in the filter.

Inductor core shape........toroids are preferred because they feature both closed magnetic field that helps control radiated EMI, a physically open structure that allows proper cooling, and easy and economical winding, as they don't need bobbins.

Implementing feedback in a 'pure' digital design is at best difficult, and may be impossible without using a DSP (digital signal processor) or resorting to an outboard analogue feedback system.

The 'pure' digital solutions described above have another shortfall, and that's the fact that the number of different pulse widths is finite, and determined by the clock speed. A digital system can only switch on a clock transition. Based on currently available information, only around 8 x oversampling is possible if a digital noise shaping filter is added to the system.