Class-D outputfilter dimensioning

[soundcheck]

Hi folks.

I am just trying to figure out how to dimension the outputfilter of my class-D amps to match them properly to my load (which is actually 16R). I couldn't really find any recommendations
for 16R.
I looked around a bit and couldn't find a consistant approach how to get the filter dimensioned.

Basic assumption: I assume the filter should deliver a 12db slope (meaning each pole gets the same fc ) and a cutoff frequency in the range of 30-50khz .

I used following formulas:

R=Rload
f=cutoff frequency

C=1/(SQRT(2) 2*PI*f*R)
L=SQRT(2)*R/(2*PI*f)

I ended up with 0.14uf and 36uh (2*36uh=72uh total) for best match of fc=50khz at R=16R.

If I'd use values:

f=50kHz
R=4R

I end up using above formulas with 0,56uf and 9uh per inductor (2). Which is quite close to the recommended Tripath values for 4R loads.

8R (22uf/10uH), as recommended by Tripath for 8R loads, would deliver f1=64khz f2=90kHz . Doesn't look that great to me. IMO the L should be also increased to 14-15uh.


Using e.g. 4R basic Tripath filters on a 16R load, I'd end up with one pole of the filter at 15khz and the other at 180kHz. Which would be more or less useless?!?!?


I hope somebody can tell me if above is correct the way I see it. Did I forget anything?
I am aware that I did not consider the inductance of the voice coil. Should I?
(From what I've seen. 99% of the examples are just talking about load as R=pure resistive)


Perhaps somebody could get me a hint of the ranges I need to stay in and what impact a poor
dimensioned filter would really have. How well have the filters per channel to be matched?


THX for your support.

[soundcheck]

Hi folks.

Hmmmh. Not very much of response yet! No ideas?

Anyhow:

I have a 2nd issue -- the next stage:

The amp output filter will obviously also heavily interact with speaker impedance
changes.

I think it might be very useful to think about a speaker linearization with a Zobel filter.

Can anybody comment on this idea?

Cheers

[smms73]

look at this application note from Texas instruments:

http://focus.ti.com/general/docs/tec...tName=sloa119a


and this one:

http://www.audiodesignline.com/howto...UNN2JVN?pgno=1

hope it help

[soundcheck]

THX.
Interesting to see at audiodesignline.com that leaving the filter out wouldn't potentially really hurt very much. Just a ferrite bead core on the lines and a 1nf to ground on each outputleg would do quite a good job.

Perhaps it is worth a try.


Cheers

[smms73]

Interesting to see at audiodesignline.com that leaving the filter out wouldn't potentially really hurt very much. Just a ferrite bead core on the lines and a 1nf to ground on each outputleg would do quite a good job. Perhaps it is worth a try.

If you leave the filter out, you have to put the amplifier really close to loudspeaker or the cables from the loudspeaker will radiate EMI, not a good idea.

For me the best cores for output filter is the micrometals txxx-02, they have very low core lost and can be use to 40Mhz.
www.Micrometals.com have software to calculate the coil, that you can download.
Search for micrometals on ebay and you will find cores like t130-02, t157-02, t200-02, t300-02. The bigger the better. I use t130-02.

[smms73]

I think it might be very useful to think about a speaker linearization with a Zobel filter.

Yes, its the ideal to have the loudspeaker impedance with the lowest variation in impedance, but if you use post filter feedback this is not so important .