Hi,

 

I finally put together a simple program for testing what happens to the pipe length out from software, through a TDM400 card, and back to software again. I haven't done much with it yet, or really studied in depth if there are pitfalls in what I have done. However, you can grab the current source from ftp://ftp.soft-switch.org/pub/sliptest.c, and compile it with

gcc sliptest.c -lspandsp

(which, obviously, means you need spandsp installed). When you run it, you need to give it a parameter which is the channel number of a circuit on a TDM400 card. Don't have anything plugged into the card. The software listens to the echo of what it sends. When it runs, it lists 50 loop lengths per second. The loop lengths are measured in audio samples.

The initial results I get are odd. I suspected I might get jumps of 1, 8 or 160 samples. These seem to be the significant sizes of things in the zaptel world. However, I get a variety of oddly sized changes, that I haven't yet explained.

The loop length is evaluated by looking for the peak in the cross-correlation of the Tx and Rx signals. This isn't foolproof, so you may get a few samples which hiccup a little. However consistent changes in loop length are real.

Treat this code as an interesting (or possibly very dull) experiment, rather than a proof of anything.

Regards,
Steve