I have incorporated the speex echo cancellation into my softphone project, but wasn't getting great results. I decided to simplify the problem, by recording speaker and mic files and running testecho. I've been pretty happy with the preprocess module and I wanted to focus on the mdf, so I commented out the call to preprocess in testecho. When I run testecho my output file looks almost identicle to my mic file. I don't see any reduction in the power level of the echo. I've looked through some of the previous posts to this group and have tried the obvious, such as increasing my filter length and reversing the order of the files passed into the application. Looking at the files in Audacity, the echo lags by about 180ms. I have experimented with filter lengths up to 2 seconds. I'm open to suggestions. If anyone would like to try my test files you can download them from: http://24.5.236.242/speaker.bin http://24.5.236.242/mic.bin http://24.5.236.242/output.bin These files are all 16-bit PCM, Little-endian. -Jason