[Asterisk-Users] Echo problem update - POSSIBLE SOLUTION

[Steve Underwood]

John Galt wrote:

>could  one at least in the case of the fxo/fxs cards just call out one
>port and be looped back into the other, record the outgoing and
>incomming call (one recording / port) then compare the phase
>difference of the 2 recordings?
>
>-Galt 
>  
>
That is probably the simplest way to achieve the required analysis. Any 
dropped, or inserted, audio sample represents a large phase jump. Send a 
single tone at a frequency which doesn't correlate well with the 8000 
samples/second rate (i.e. things like 1kHz would be a poor choice). Then 
just run a sliding medium term correlation (simply the sliding dot 
product of maybe 20ms of audio is adequate) between the transmitted and 
received audio, and look for significant jumps, after an initial 
settling period. That should be a solid, noise tolerant, way of looking 
for these phase jumps.

With digital interfaces (BRI, E1, T1) you should get back precisely what 
you send. It is better in those cases to loop back, send random numbers 
and check you get back the same random values. The path length is not 
very long, so its not a big problem to sync to the delay between the 
outgoing and incoming stream. Once things initially settle, any wrong 
value, and especially any change in the delay between transmit and 
receive is a bad thing.

A simple tool like this for people to check out new installations has 
real long term value, beyond the short term goal of fixing echo 
problems. The number of people having data slip problems affecting their 
use of my SoftFax has made me consider writing such a tool recently. If 
anyone would like to save me the trouble, please do. :-)

Regards,
Steve