Low Pass filters for class-D audio testing??

[Zero Cool]

I need to build a better set of low pass filters for my test setup. I need to build a set of high power filters for my scope and distortion analyzer inputs that will allow me to test class-D amps up to 500 watts from 20-20khz.

What I have now is 12db oct at 20khz and that allows me to test up to 5khz at full power, if I go over 5khz at more than a few watts i burn them up which has happened more than a few times.

SO. Suggestions? maybe 24db at 100khz? with an 8/4 ohm resistive load?

Can I use standard speaker crossover calc's to do the math??

where Eva?

Zc

[KSTR]

Check specs of AP's AUX-0025 Filter Unit (attached).

Note that you DON'T attach the Load to the filter output, rather the other way round.

A speaker Software would certainly do it (LSPcad demo version), or LTspice.

[Zero Cool]

I will check it out. I am not using a software based test set up. and right my filters connect to the input of my scope and my dist an. but my filters are made up of a series resistor / inductor and a capacitor to ground and as you start getting above 5khz there is sufficient energy to burn up the low wattage resistor and inductors as you are getting into the tail of the 12db down points of the 20khz filter

[Zero Cool]

Ok well it's great that AP makes that product but i am quite sure I can't afford ANYTHING with the AP name on it. I need to BUILD one!

[KSTR]

Well, the AP filter is on the order of 800EUR and quite unlikely to be found used. I have one at work but it's still having the screws sealed so no way to peek inside...

[ChocoHolic]

I am typically using passive RC filters in the range of 3rd...5th order. Similar to the 4th order filter in the attached picture.

Chose a 5W type for the 100 Ohms and 1 W for 330 Ohms, all the rest 1/4W resistors. Caps should be MKP types with 100V rating (lower quality will still work....).

[Zero Cool]

I need to get out to 20khz flat before roll off. part of my testing involves full power frequency response so i need to make sure the filter is not causing problems at 20khz

[KSTR]

I would think with a well-known roll-off you could account for that easily in the measurements. Flat response (how flat, <0.1dB?) and sharp roll-off complicates filter design and construction quite a bit.

[ChocoHolic]

Of course, you can adjust the transferfunction according your taste.
The shown values were just anticipated from your first posting to fit -3db at 20kHz. Feel free to make the filter faster unless your analyzer gets irritated because the switching residuals are not suppressed sufficiently anymore.
Or simply measure the full power frequency response with a function generator and the scope directly, you do not need any filter for that at all.

[Zero Cool]

Originally Posted by KSTR

I would think with a well-known roll-off you could account for that easily in the measurements. Flat response (how flat, <0.1dB?) and sharp roll-off complicates filter design and construction quite a bit.

Yeah within .1 to .2db accurate to 20khz. most of the switching frequencies seem to be 300-400khz or so so I figure if i have a 100khz low pass filter that should be good enough to keep the junk out of the analyzer yet give accurate results.