Development of a "reference" class D starting point

[Jan-Peter]

Bricolo,

A full-bridge is more complex and by testing/experimenting you destroyes 2 times more Powerfets

To keep it simple I advise to go for 2 powerfets output. The only possible problem can be powersupply pumping. But with a good powersupply and some 10.000uF capacitors this is easy to avoid.


Regards,

Jan-Peter

www.hypex.nl

[classd4sure]

Originally posted by Bricolo


Here's the patent
http://l2.espacenet.com/espacenet/vi...h&LG=en&DB=EPD


What I want, is to understand how what I build works. That's all... (maybe that's not as easy as it seems)

As for the power, can't we simply agree on a design, and adapt the PS rails (and the output mosfets) for more or less power?

It certainly can be a source of confusion, given the limited info available, the poor quality of it, I find it's either too beginner or too undergraduate. Multiply that confusion factor by 100 when you start looking at different types of modulators.

I think in this case the patent explains the operation very well, but you just can't visualize it, can you? That's what simulators are for, and I think "simulationable" should remain a design goal for that very reason.

You're right you could just scale up the rails to an extent and get more out of it with a few small component changes but the issue is at 20 watts, it's happy running on a cheap plastic protoboard, or p2p or something, happy enough to impress you with it I'm sure. Start turning up the power though, things get allllot more critical.....dual layer pcb required....critical layout..SMD..etc.

Although Bruno's idea of seperate PCB's could help that, each pcb still needs to be done right, just a huge difference in the design requirements for a 20 watt max circuit to a 20 to 100+ one. It's still an excellent way to go for any DIY'r though, nothing like a good test mule.

Regards,
Chris.

[classd4sure]

Originally posted by Jan-Peter
Bricolo,

A full-bridge is more complex and by testing/experimenting you destroyes 2 times more Powerfets

To keep it simple I advise to go for 2 powerfets output. The only possible problem can be powersupply pumping. But with a good powersupply and some 10.000uF capacitors this is easy to avoid.


Regards,

Jan-Peter

www.hypex.nl

Hi Jan,

I can't help but agree with you there. It's enough of a job to drive just two mosfets properly. I don't think a little 20 watt amp could pump the psu all that much anyway could it? Another reason for keeping it small and simple.

Bricolo if you just want 100W.... see the link above, www.Hypex.nl but if you're more interested in learning about it..I'm thinking 20W.

It has to stay low or it will get complicated fast. That's a later project

Regards,
Chris
Seriously crashed now.

[johanps]

Originally posted by Bricolo
In fact I asked for a basic schematic. No power/THD defined in it.

For a block schematic, I think the one I'm attaching here is useful.
A first order noise shaping filter (one integrating op-amp) is highly desirable to get good performance over the entire audio frequency range. Higher orders should be even better, but introduce stability problems. (Probably more problems than benefit for us at this point?!?)
The "output driver stage" contains the output filter as well.

By the way, I also like Bruno's idea of separating control/feedback from the output test mule . The first PCB could be constructed with a line somewhere separating the parts, so you can chop off and replace one part later...

If no-one else beats me to it (or protest), I will try to post an Eagle schematic tonight with a few proposals in place - to get something more concrete to discuss around.
Meanwhile, I would very much like to get suggestions on how the output driver stage should be constructed. I am familiar with circuits like IR2110, MAX626, HIP4080, and a few others in the same category. I like the IR2110 for the kind of output stage I have in mind, but I wouldn't mind switching it for some chip with a lower number of pins, a bit higher switching performance, possibly at the expense of lower voltage tolerance. (500V isn't needed here )
Or is the discrete driver from the UCD patent useful? (Haven't simulated/worked on it myself, so I don't know how it performs...)

[phase_accurate]

I think in this case the patent explains the operation very well, but you just can't visualize it, can you? That's what simulators are for, and I think "simulationable" should remain a design goal for that very reason.

Your wish is my command !!!

Regards

Charles

[Bricolo]

Originally posted by johanps

loop.gif

I don't see what causes de self oscillation in this diagram

[phase_accurate]

That's the circuit BTW.
Transistors and FETs are the ones I just had on hand (i.e. my SPICE library). In real life there are of course better suited models available. Gate drivers were modelled as voltage controlled voltage sources.

The dimensioning (and ciruit) does of course leave much room for improvement.

Regards

Charles

[johanps]

Originally posted by Bricolo
I don't see what causes de self oscillation in this diagram

Uups, noticed that the "+" and "-" inputs are not marked on the comparator in that picture.
Anyway, first ignore the noise shaping aspect (just think "amplification stage"). Then, if the momentaneous voltage on the output is lower than the input (times the feedback factor beta), that difference will be detected by the comparator, causing the output to go "high". This will quickly drive the output to a voltage slightly higher than the input (times beta), and so on... The mean value of the output will track the input. The precision with which it does so depends on the loop delay (i.e. the self-oscillating switching frequency, and the order of the noise shaping filter.

[Bricolo]

If I compare your schematic to Johanps's loop.gif, I can't see the noise shaping filter (integrator).
Is it formed by C2 in the feedbacl loop?

[Bricolo]

Originally posted by Bruno Putzeys

In the end I managed to order diamond coated burrs in a local shop (20 euros apiece or so. Manufacturer is called Pferd).

Like we say in french "Le monde est petit!"
Pferd's french "section" is 2 minutes far from where I live, and a member of my family used to work there!

They make great stuff