Choice of caps for Class-D output filter ?

[soundcheck]

Hi guys.

I'm trying to tune my Class-D output stage a little.
I just ordered some Würth coils as recommended in another thread over here.


I am wondering if I should also have a look at the output filter caps!?!?

What would be your prefered choice?


THX

Cheers

[nigelwright7557]

I just use standard polyester caps for my output filter.
Along with a t106-2 core with 2 metres of enamelled copper wire.

[Aucosticraft]

Yes, Plain poly caps work well. Just use its voltage rating 2X of the peak unfiltered voltage at output.

[lorylaci]

Originally Posted by soundcheck

Hi guys.

I'm trying to tune my Class-D output stage a little.
I just ordered some Würth coils as recommended in another thread over here.


I am wondering if I should also have a look at the output filter caps!?!?

What would be your prefered choice?


THX

Cheers

One of best filter cap for class-D is polypropylene. See this page: WIMA
As you can see polyester can have high dissipation factor at high freqs, they serve well usually, but polypropylene is better, specially for high power.

[soundcheck]

Hi folks.

I now plan to go for this setting:

Filter:
Roederstein KP 1832 0,52uf ( I had them around from earlier projects), I could also use Vishay MKP1844 at 0.44uf.
Decoupling:
Kemet COG 1206 0,068uf
Inductors:
Mundorf aircoils 2*10uh (replacing (for now) my earlier choice of Würth coils)


Now I'd need some support to verfiy the dimensioning:


Goal:

Q:
A 0.707 > Q < 1.0 would be acceptable I guess. A little overshot would be acceptable if going towards the upper value.

With above BOM I'd end up at a

Q=0.8
and
Fo=49khz

assuming an Rl=5R.

Things get slightly better if I'd consider the inductvity of the drivers.


One thing I'm not sure about. Calculating/figuring out a rather exact Rload value I'd have to apply.

I'm running a 2way system (Bastanis). One driver at 11Rdc (no crossover parts) one at 9Rdc (hp with 1uf cap). Can I just use these DC values for a rather reasonable dimensioning? Above 5R is the result of that calculation.


Thx.

[soundcheck]

The output coil thing is still bugging me.

I know -- over the years we've seen endless discussions over here, what inductors to use on the amp output, how to wind them, core materials, wire asf. asf.


The majority of those typcially used coils on class-d amps which I looked
at are ususally not made for that particular audio HIGH-Power NF purpose.
Many of those are a kind of spin-off from normal coil production.
These coils do reside at the same position as the traditional crossover inductors.
From that perspective I just anticipate that the ususal CLASS-D output coil performs worse than well done crossover coils on sound related attributes
-- since these crossover coils were made to perform best in the audio frequency range . ( Has anybody ever seen Würth or Coilcraft inductors used in crossover designs?)

I tried those Mundorf air-coils, since these are the result of years of
evolution in the speaker crossover design.
I'm well aware that these are a comprise in terms of filtering the HF stuff out. Though I should mention that I'm in the comfy posititon running speakers with a 98db/SPL.
That leads to a rather low power draw and avoids excessive radiatiton.

My latest listening tests have shown and confirmed again ( I used to use them earlier) , that the sound signature becomes really pleasant without loosing anything.
They sound IMO that good that they gonna stay...

...for now - I'll keep trying.

As usual in audio land - It's always about finding the best compromise.

Enjoy.

[lorylaci]

Soundcheck: Is this a fixed-freq PWM amp, is there feedback, or is it self-osc, hysteresis, UcD...?
If UcD or other post inductor feedback, the impedance of the load does not matter so much. Even the type of output inductor doesn't count so much there.
If its pre inductor feedback, or without feedback, then you cannot really optimise the output filter, unless you set corenr freq high, so any peak would happen off audio range.
You should forget air-core inductors in class-D, unless you want a noisy radioantenne.
A speaker crossover should separate different audiobands (without any >20kHz material), so they are made to perform this the best. Here you must separate whole audio band from the HF band. Normal crossover inductors are not made for this, so different inductor is needed.
The inductor you shoved are multilayer ones, with so much parasitic capacitances. They will have a hard time filtering, and they will radiate much.

How do you check sound? Do you make a bode plot measurement? Do you do blind-tests? Otherwise you just hear back your preconception.

[soundcheck]

IMO as long as the signal travels through that inductor, it's about physics
impacting the signal. Doesn't matter what feedback scheme you're running.

The filter performance is of cousre a different issue. And that's what I mentioned in my post.
My choice might not be the best filter (I'd need to measure that) for now it sounds better then the alternative
solutions I have on the table.


BTW: DIY-Paradize has produced and sold hundreds of really nice sounding Tripath amps with air-cores on the output over the years.
On low power amps air-cores are not that bad.

[lorylaci]

Feedback can have much greater impact on signal than the inductor.
A post-inductor feedback will (try to) decrease the impact of the inductor on signal. (for example see a bode diagramm of an UcD amp with different loads, and compare it to an amp with pre-inductor feedback) Usually the differences between inductors in a post-inductor feedback amp will not be audible, only measureable.

Sound better means only listening test? (I asked whether its blind-test, since if its not then you cannot even know it is really better, or your bias is talking)

What is your other inductor choices?

[soundcheck]

Quite some people and even companies (e.g. Dexa-Newclassd) consider pre-filter-feedback the prefered choice, because it just sounds better.

In this case I 'dbetter take care on matching the output stage
with the drivers, to get the impact as low as possible.


If you look some posts back, I had asked a question how to calculate "properly" the Rload for that pre-feedback design.


Thx.