I use dmix+snd-aloop for all local applications; then forwarded to the dmix+Layla3G using a custom C program. Despite asking for clock sync, they slowly drift. Am I doing something wrong? $ echo hw:Layla3G,0,0 > /proc/asound/Loopback/timer_source The forwarder drains its buffer and eventually underruns. Loopback is not providing enough audio to feed Layla3G -- drifts by about 9 samples/second at 48kHz. This _feels_ like a regression; I'm pretty sure this was perfectly in time before. I'd need to bisect kernel and alsa-lib changes. Below shows the status during playback/capture. The timer is definitely connected in some way. It's listed below, and aloop ticking corresponds to whether Layla3G is open by any application (not just the forwarder) -- Mark $ cat /proc/asound/Loopback/cable#0 Cable 0 substream 0: inactive Cable 0 substream 1: inactive Cable 0 substream 2: inactive Cable 0 substream 3: valid: 3 running: 3 pause: 0 Playback buffer_size: 65536 buffer_pos: 12800 silent_size: 0 period_size: 512 bytes_per_sec: 384000 sample_align: 8 rate_shift: 0 sound timer: hw:0,0,0 timer open: playback Capture buffer_size: 65536 buffer_pos: 53760 silent_size: 0 period_size: 512 bytes_per_sec: 384000 sample_align: 8 rate_shift: 0 sound timer: hw:0,0,0 timer open: playback Cable 0 substream 4: inactive Cable 0 substream 5: inactive Cable 0 substream 6: inactive Cable 0 substream 7: inactive