Development of a "reference" class D starting point

[classd4sure]

http://www.alltronics.com/download/1373.pdf

Have a look on page 1.

The diagram shows a charge pump and a bootstrap, they only show a "boot" cap the rest is internal.

Typically on driver IC's that have both, the bootstrap does the grunt work and the charge pump is only there to maintain the bootstrap's charge so that it can be used for very large duty ratio's/low frequency or whatever.

[phase_accurate]

Me again with the reverse recovery issue.
There are circuit topologies that don't know reverse recovery at all since they avoid the diodes going into reverse conduction.

One solution has been proposed by Brian Attwood and it was used in the Peavey DECA series amps.

Regards

Charles

[classd4sure]

Hi,

You mean DirectFet?

There have been others which claimed to cure this too and they didn't.

It would be a shame to have to restrict topology just to avoid that. BCA might be one but it has other issues.

A strong driver is required.

Tripath's patents are usually vague, do they go into any detail about that?

US5408150 is worth a read.

[subwo1]

Some designers place a Schottky diode in series with each MOSFET to prevent body diode forward biasing. I have not done much simulation and even less physical experimentation concerning that technique.

Charles,

Yesterday it finally occurred to me how to delay a pulse wave, phase shift it, without shortening the on-time. It is for my SMPS design, so I think it would not work as well for class D since the resistor, for variable delay, would be replaced with something like an H11Fx bidirectional FET output optoisolator due to the fact that the resistor sees dual polarity. It is possible to insert a transistor inside a bridge rectifier where the resistor goes, but the results suffer.

[subwo1]

Oops, the phase shifter above works on the simulator because the input gates do not include the over-voltage protection shunt diodes. In real life, those diodes dissipate some of the charge on the capacitor, decreasing the on-time and ruining the duty cycle preservation. The result is that the greater the resistor and attempted phase shift, the worse the problem.

[Kenshin]

Originally posted by subwo1


Indeed, reverse breakdown can cause internal hot spots and then failure.

I didn't mean by diode failure--I mean GATE failure.

Note: I didn't connect the bootstrap cap -- some diagram has it unconnected,and I didn't know what is it for.

[subwo1]

aahhh

[classd4sure]

Nice tweak for the upper driver is put a cap on it with a diode over it, it'll pump the gate below zero and provide about a volt of negative bias.

Sadly, it only works on the upper driver

[Kenshin]

N+P bridge is more simpler,but the P-channel FET has an Ron more larger...unsymmetric and low current capability.

use a full bridge can get back the symmetry and have higher power output.

is it worthy to use a N+P design,with a transistor gate driver instead of chip,and use the saved cost to build a full bridge?

full bridge could also simplify the PSU,and use a smaller filter cap.

which design have less overall cost?

[classd4sure]

Originally posted by Kenshin
N+P bridge is more simpler,but the P-channel FET has an Ron more larger...unsymmetric and low current capability.

use a full bridge can get back the symmetry and have higher power output.

is it worthy to use a N+P design,with a transistor gate driver instead of chip,and use the saved cost to build a full bridge?

full bridge could also simplify the PSU,and use a smaller filter cap.

which design have less overall cost?

I think it would be very difficult to properly do a reliable full bridge with an ordinary discrete bjt driver. It is difficult enough to get one working with a half bridge and symetric N channel switches, and as you say, P channels will be slower to respond.

High power P channels are also still alot more rare than N channels and are more expensive than N channels, so I don't really see where you're saving in cost. I think the usual practice is to oversize the P channel with respect to the N channel as well, as an attempt to try and match Ron. Therefore you'd need an even more expensive mosfet.

I'll confirm this in a few minutes, but I think it takes transistors to drive a high side N channel as well!

I think all you save with a P channel is the bootstrap circuit, which isn't substantial by any means.

It seems it is best to use dual LC filters in full bridge as well (better rejection of common mode signals), so you dont' get away with a smaller cap, but require another identical one.

Less overall cost? With the circuit I worked on before (full bridge) it only doubles the output stage (drivers, boostrap/supply, mosfets and filter).

So that's not alot of added cost! Also, you can design a full bridge to handle 4X the power of a a similar half bridge.

In that respect the PSU isn't really simplified, but you can use smaller and far cheaper parts for it than a similar half bridge design would require (to achieve the same power level that is). Since things like PSU toroidal transformers and large can electrolytics of high quality are big ticket expensive items, this would be a substantial savings!

So with that in mind, my opinion, for any given power level, a full bridge would pay for itself, and is well worth doing.

However a robust/precise driver is likely even more important.