best compromise for LC filter design, driving 4, 6 and 8 ohm loads

[Oil]

Hi All

I'm currently in the process of designing a small amp based on a TI TPA3106.

I'd like to have the output filter designed so that it will run 4, 6 and 8 ohm loads satisfactorily, without excessive peaking / THD implications (as little as possible anyway)

I have read through this: http://focus.ti.com/lit/an/sloa119a/sloa119a.pdf and it seems as though the tables on pages 11 and 12 show 15-20 uH with 1.5-2 uF would be a reasonable place to start.

I'm hoping that somebody can offer some suggestions on the values to use?

Given that I don't have the equipment to measure the actual results, could anybody recommend a suitable software package to simulate the filter design and implications?

Lastly, I see sure electronics put a 470uF coupling electrolytic after the filter on their TPA3123 board: here (page 15) (they use a 33uH / 1.47uF combination). What is the benefit to doing this and how would you guess they justified the filter selection?

Many thanks

Oli

[theAnonymous1]

The Sure TPA3123 board uses 470uF caps on the output to block the DC bias since it is a single-ended design running off a single supply. The TPA3106 runs BTL so these caps are not needed.

Here is an online filter calculator page you can mess around with....

Filter Design and Analysis

On my TPA3106D1 boards I use 15uH + 1uF. I haven't measured the response, but it sounds fine to my ears running 4-8R loads.

[nigelwright7557]

Just make sure the inductor will take the current for the 4 ohm setup OK.

My class d amp uses 22uh and 470nf and runs into 4 ohms very well.

[killerbobjr]

You can design for a nominal impedance of 4 ohms and use a Zobel network to minimize impedance mismatches. If you design the output filter for an Fco of around 1/5 the switching frequency (typically around 250KHz for the TI TPAxxx chips), you can pretty much get away with using LC values for a 4 ohm load and not have an audible HF peak when attaching 8 ohm loads.