Class D

Besides the other things, I do the Class D amplifier designs.

1. Leaving aside the audiofoolery and the pro audio market (which are the special niches), the audio is really about low cost. The main purpose of Class-D is the cost saving on the heatsink material. Specmanship doesn't really matter as long as it is reasonably good. As such, everything should be simple, cheap and reliable; the elaborated designs are missing the purpose.

1. The typical nonlinear distortion of a Class-D process is in the several percent range. That is somewhat higher then the self distortion of a classic linear amp (~1%). However, with the linear approach, there is 40...50 dB of negative feedback to bring the distortion down; whereas the Class D can only allow for some 20...30 dB of feedback. As such, the specmanship for a class D amp is typically by order of magnitude worse then that for a linear amp. However it doesn't matter as long as it is reasonable.

2. The high switching rate (MHz range) is a cure for many problems of the class D; it is easy to achieve any specmanship whatsoever. However it brings down the efficiency almost to the same numbers as for the classic AB-amps, and the EMI becomes a hell of a problem. High switching rate is what they like to demonstrate at the shows and write in the papers; however in the reality one would like to get by the lowest switching frequency. The ballpark is: optimum switching rate ~ 20 x maximum audio frequency of interest. If the sw. frequency is lower then that, it is difficult to achieve the performance; with higher rate, there is a hit to efficiency.

Vladimir Vassilevsky DSP and Mixed Signal Design Consultant

Just a couple questions, as I hope to do some Class D amp design in the future. Do your designs take the feedback only by lowpass filtering the junction of the output filters, or do they also incorporate feedback from the output post-filter? Do you use the self oscillating topology, or do you have a separate oscillator? Thanks.

It depends. What, when, where, what is available, how and how much. Each approach has its merits and limitations.

Vladimir Vassilevsky DSP and Mixed Signal Design Consultant

Phil gave you a *good* summary. He could have phrased the last bit better but it is Usenet. I don't see anything to object to.