Digital X-over

[borges]

I'm thinking about taking my DAC project one step further. This time around I want to try out digital X-over and separate DAC and amp for each speaker element.

My experience is mainly in DSP and small-signal Laplace stuff with opamps and RC filters. I've never made an X-over before, so I hope you guys can get me up to speed on a few topics.

In an X-over, how do you consider the speaker element? Resisistive / inductive load?

With 1st-order filters it is theoretically possible to achieve linear phase because the sum of the lopass 1/(Ts+1) and the hipass Ts/(Ts+1) is exactly 1. But what I also heard somewhere is that this feature can be found again at 4th order, but not at 2nd and 3rd. I would love some theory on this!

The basic design idea I have right now is to do all the digital filtering in an FPGA; FIR oversampler and IIR X-over closely matched to an "ideal" analog X-over which sums to linear phase. The DACs will use some interpolation theory I've been working on. They will be PCM 1704 or 1794. Not sure exactly how I will do I/V conversion and ampilification. I could also need some input on digitally controllable volume in a high-quality amp. But that'll probably be in a different thread.

Cheers,
Børge

[Sheldon]

Look down the list a bit:

A how to for a PC XO.

[jackinnj]

The driver's mechanical and electrical characteristics can be transformed into an RCL Network.

For simplicity's sake, however, consider 1 mH in series with 20uF in series with 8.2R.

Joseph D'Appolito wrote a book with all the math "Testing Loudspeakers" it is available from Old Colony Sound. It's worth the money.