Class-D topologies - why most modern amps are half-bridge? Why not multilevel?

[EB1000]

Hi

I've noticed that Hypex, Icepower and Purifi, are all half-bridge based.

Yes, half bridge are the simplest topologies and most easy to control. But still, since a class-D amp is essentially a DC to AC converter (Inventer), an a full-bridge inverter outperforms an half-bridge, and similarly, a multilevel inverter, such as a cascaded H-bridge (series connection of many full bridge topologies), will ensure the lowest THD per switching frequency, as well as lower device stress, lower EMI (dv/dt), higher efficiency and reduced or even eliminated output filter.

Also, full bridge topology is capable of absorbing power from the load back to the DC rail, thus reducing pumping effect of the DC supply, when proper control is utilized. So why most modern class-D amps are based on half bridge? The only full bridge topologies are not really full-bridge but a common mode based BTL connection of independently controlled two half bridge topologies. A true full bridge BTL should be differentially connected, because that why the effective output switching frequency can be doubled.

Attached are different possibilities of output filter configurations for a full bridge topology. Type 1 seems to ensure lowest THD for the same PWM frequency and filter cutoff frequency (BD modulation).

Does it have to do with reducing costs? Are there modern full bridge or multilevel class-D amps out there?

 

[mocenigo]

I do not think this is to reduce costs. Simply speaking, Class D amplifiers can already get good performance as half bridge designs, so there is no need to make them more complicated. Full bridge also makes feedback loops more complicated, so it is better to BTL two modules which have completely separated feedback loops and comparators etc… one could save some in components by designing them full bridge to start with, but other design aspects would be more complicated.

 

[DonH56]

Multilevel requires precision matching of levels to maintain low distortion. For a power amp, that can be a very stringent requirement that is extremely challenging to meet. Multilevel delta-sigma converters have extensive on-chip calibration and trimming; it would be much harder to implement in a high-power amplifier.

Simplistic example: a single-bit delta-sigma ADC is intrinsically "perfectly" matched as there are only two levels. That is, they only need "match" to 50%. Add a bit, four levels, and now those extra levels must be matched to the precision of the output because the "steps" must be as precise as you want the output signal. To achieve 16-bit precision, the levels would have to be matched to within 1/65,535 or 0.0015%. Processes match to around 1% with a few tricks, and more tricks and trims can get to perhaps 0.1% or better, but matching to 0.0015% is often impractical. More complex architectures and extensive calibration processing is used. For a class D amplifier, matching the output levels to high precision whilst generating high power into a complex load would be... challenging.