On 2013-05-14 08:50, David Henningsson wrote: > On 05/13/2013 11:19 PM, Christian Iversen wrote: >> On 2013-05-13 05:41, Arun Raghavan wrote: >>> On Sun, 2013-05-12 at 17:18 +0200, Christian Iversen wrote: >>> [...] >>>> Is there any way to have pulseaudio keep the device open at all times? >>>> Or some other way to debug this? >>> >>> Disable the loading of module-suspend-on-idle in /etc/pulse/default.pa. >>> I'm assuming you don't care about the resulting power consumption. :) >> >> Ah, there is was! :) >> >> Well, that's great. Thanks! >> >> I don't think the Scarlett 2i2 really supports any meaningful power save >> mode anyway. It's quite a low-power soundcard to begin with (powered >> entirely by the USB bus). > > Not so low-power IMO, it takes about 2-3W of my battery when connected > (only tested with phantom power on). I think it does that regardless of > stream running or not, but have not really checked. (The direct monitor > requires headphone amp and mic amps on anyway.) I should have been clearer. I think there is, at best, only a very slight amount of power saved by idling this card. But the effects to the user are quite severe. >> Anyway, even with the official driver for windows, this card has >> terrible clicks and pops when starting up from suspend (or power off). > > Ok, good to know. > >> Does module-suspend-on-idle have a blacklist? It should! And this card >> should be on it. > > There are other reasons for wanting to release the sound card when not > actively using it - it makes compatibility with other apps easier, i e, > everything that wants to use ALSA directly (except jack2 which has > device reservation). Point taken. But it's still quite annoying ;-) >> What does pa use for identifying backends? I mean, if I wanted to >> blacklist certain modules for a specific PCI or USB id, is that >> something that is reasonably easy to do? > > Not currently, but it sounds very useful if we had a smart way of > accomplishing this. Agreed. Just like the kernel drivers have ample support for identifying device quirks. I'm almost sure we could benefit from a similar scheme? -- Med venlig hilsen Christian Iversen
