I have an AVStream virtual audio driver, based originally on the "avsamp" sample but almost completely rewritten. I serve exactly one format at a time, usually 22050-M-16. The only properties I support are the two jack descriptors. If I connect directly to the device in graphedt under the "WDM Streaming Capture Devices" category, audio plays fine. But if I build a graph using the instance under "Audio Capture Sources", it appears that the system audio engine gets involved. One minor side effect is that the pin always says 44100-S-16, even though I don't support that format. The audio does play, however (so someone is converting). The larger problem, however, is that my audio is being boosted to the point of distortion. I don't understand where this is coming from. I'm currently attenuating my own signal by 50%, and the resulting stream is still painfully overdriven. Where is this coming from? Is there some AGC in the audio system? Does it sample a short period to decide what my average volume is and boost to compensate? How can I stop this? -- Tim Roberts, timr@xxxxxxxxx Providenza & Boekelheide, Inc.