Help with Class D Amplfier Design (feedback)

[classd4sure]

You can also have intermodulation products between different channels that tend to produce annoying "whistles". That's a defect of every self-oscillating Class-D amplifier.

I wouldn't go so far as to say it's a defect, but sure it's a possibility. I understand this same problem exists in poorly implemented clock based designs as well, there's always a work around isn't there? You have to admit, the definite pro's far outweigh the possible cons.

Regards

 

[hypnopete]

Does anyone know a good and reasonably simple way to set up some clipping led to a digital amplifier, ie the design i have come up with.....?? is the easiest way to use a comparator for example, when the output voltage reaches 99% of the rail voltage... any other ideas...?


Regards

Peter.

 

[classd4sure]

Originally I'd clamped mine at the input, two pairs of series diodes for about +-1.2 volts limiting. Later removed them I found they weren't needed in my case.

Regards

 

[hypnopete]

Perhaps you misunderstand.

All i want to do is give some visual indictation that clipping is occuring..

Regards

Peter

 

[classd4sure]

Yes I did, and that's a good idea.

How about a PLL set to lock on 150Khz

 

[subwo1]

For clipping indicators, consider two zeners, one at the upper and one at the lower rail. A resistor connects them together. The LEDS attach to the inner connections of the zeners, observing proper polarity with series 1amp 200PIV rectifiers in series with as well as the necessary current limiting resistor for each LED. The higher the voltage rating of the zeners, the farther from the rail the indicators will sense the audio signal. The LED assemblies are connected to the audio output.

 

[hypnopete]

Okay I'll Give that a go!!.

I think i could probably get away with just using one led, and monitoring on only one of the rails, probably the top.

Thanks.

 

[subwo1]

Sounds good. The rectifier should be an ultrafast or two 1N4148 in series.

 

[phase_accurate]

Clipping detection within class-d:

If one drives a class-d amp into clipping it stops switching. I.e. one output device is kept on and the other one off. This situation can be detected at the input of the mosfet driver by means of a retriggerable monoflop.

Regards

Charles

 

[hypnopete]

I have just finished winding my new Torroidal Inductor, which is for use at around 300kHz.

As i was unsure of the core material, mix. ( powdered Iron)

I set up a parallel resonant circuit with the inductor and a 1.0uF capacitor.

The resonant peak was measured at 11.5kHz.

According to my calculations, ie L=1/(((2*pi*f)^2)/C)

The inductance is 191uH!... This is much too high i beleive . There are 42 turns on the inductor.

Should i just use say 10 turns instead of 42? to get something in the ball park of 47uH?

Any help would be much appreciated. I am not sure of the exact value i need either.


Thanks

Peter

 

[phase_accurate]

Inductance is a function of the square of the winding numbers (amongst other things of course). If you remove HALF of the windings you'll get around 48 microhenries.

The filter inductance depends on the desired cutoff-frequency and the expected load impedance.

Regards

Charles

 

[hypnopete]

Thanks Charles,

Okay, I realise it is a function of the square of the the number of turns.

For a output filter designed to run around 300kHz, with a -3dB point just above 20kHz say. The expected loads range from 8ohms down to 2ohms. It doesn't matter too much in this application if the -3db point is a little lower.

Would 48 uH be a good value. I am yet to work out the value of bypass capacitor needed, as this will depend on the inducance of course.

Thanks

Peter

 

[Pierre]

Hello, Peter.
I have a small DOS program that calculates any order LC filter (highpass, lowpass, whatever) that is perfect for this application. You give it a -3dB freq. and a load impedance and it calculates L and C values for Butterworth, Bessel, etc. It also shows the freq. response obtained, etc.

If you are interested, I can try to post it here. I am sure it will be very useful for Class-d designers.

 

[hypnopete]

Hello Pierre,

That would be fantastic if you could do that....

Would make the design a bit easier for me

Regards

Peter

 

[Pierre]

Wow! We are really online!
Here you are. I have split it in 3 files. You only have to unzip them in the same folder.

Here is the first ZIP:

 

[Pierre]

...the second one...

 

[Pierre]

... and the last file.

The main program is RFDESIGN.EXE
A small note: in order to design the filter, type "0" as the source impedance and whatever you want (4, 8 ohms, etc) as the load impedance. You want the standard (not the dual) network.

Enjoy!

 

[hypnopete]

Thanks for that pierre,

I just installed it and am playing with it. seems to actually be very helpful.


I will have anothing play tomorrow, and let you know what i come up with.

Thanks.

Peter

 

[hypnopete]

Help with output inductor...

After using the software Pierre Posted,

I have decided that 54uH is about right for what i need.

I have a powdered iron toroid with 22 turns giving 54uH ( measured). The core is colored yellow / white.

I have tried two variations, one with 2mm wire, the other with 4 strands of 0.9mm wire.

On each the windings get really hot, quite quickly. Its not the core thats getting hot, just the wires.

The pwm amplitude going into the inductor is -80v to 80v, and carrier frequency is 300kHz. The output filter capacitor is currently 0.47uF. The output voltage ripple at 50% PWM is about 2v peak-peak.

Any suggestions would be great.

Thanks

Peter

 

[Pierre]

What load are you using? I mean, what's the RMS current to it?
If you are delivering one thousand watts, then you can expect the wire to heat up a bit, don't you think?

What is the lenght of the inductor wire? I can calculate the wire resistance and hence approximate the losses in it basing on its length and diameter, if you want.

Best regards,
Pierre