JohnSwenson wrote: > So here is where I am standing today: > TT stuff on 3359 seems to work (UAC1.0 async works without hub) > UAC2.0 async has clicks/pops > Kernel that comes with board is not RT and doesn't support SCHED_FIFO > Other kernels will probably support SCHED_FIFO but probably not RT - > need to try them see if fixes clicks/pops > I want to use 3874 processor which does have hardware scheduling, but > not ehci. Not clear if driver makes use of hardware scheduling or not > Would have to test on an eval board. Only eval board costs $2200, a > little pricy for out of my pocket! > May be possible to build my own eval board for $200-$250 or so, takes a > month to do so. > If we can get rid of clicks/pops with software (thread priorities RT > kernel etc) is that sufficient to say the issue is closed and go with > the 3874 without explicitely testing that processor? > Freescale has similar processor which DOES say it's USB host port is > ehci compatible. It has some drawbacks: > single memory controller, harder to do 512MB > single I2S and S/PDIF, not nearly as flexible > no builtin HDMI, needs external chip (but does have VGA on chip) > REAL BIGGY does not have pad and symbol for my board layout software, > have to do by hand, very time consuming and error prone > > Any thoughts of where we should go? > > John S.
Are the clicks and pops with squeezelite or anything for uac2? I think bigger alsa buffers will minimise the impact of no sched_fifo, but want to understand if you get clicks for high rate iso. We do need a different kernel though if that is all it does. I would really like something which is known to support rt_prempt as this means we can get closest to touch performance for sync as well as avoiding xruns. This would be for i2s/spdif as well as USB. Are you looking at imx5 or imx6? Both have ehci and would need a hub for uac1 from my reading of their data sheets but would do usb ok. I tend to think we would be safest to use something we can also get an evaluation board for, rather than something we can't test cheaply. Do this make the 3874 higher risk as we can't do this. ------------------------------------------------------------------------ Triode's Profile: http://forums.slimdevices.com/member.php?userid=17 View this thread: http://forums.slimdevices.com/showthread.php?t=97881 _______________________________________________ discuss mailing list discuss@lists.slimdevices.com http://lists.slimdevices.com/mailman/listinfo/discuss
