04:11 <bmbeach> hello If you take the current accepted eppl programming

04:11 <bmbeach> values(13 of them) and divide by 1024 you will get the

04:11 <bmbeach> sample rate. Now if you divide the entries recursively by 2

04:12 <bmbeach> and form all the values into a matrix you will find that you

04:12 <bmbeach> have all the sample rates you will ever need including the

04:12 <bmbeach> stock ones 8000, 16000, 24000, the all important cd rate

04:12 <bmbeach> 44,100, 48000 all being exact integer values. So setting up

04:12 <bmbeach> a clock rate scaler that both programs the epll and the post

04:12 <bmbeach> divider each time a new sample rate is need, is all you

04:12 <bmbeach> need. As I say using values that are not recommended is

04:12 <bmbeach> asking for trouble, the pll is not a simple

04:13 <bmbeach> divider/multipler but has physical analog constraints.

04:13 <bmbeach> Again a sound codec on an exynos soc will -> never <- act as

04:13 <bmbeach> master unliess they change things see p 4-304 of the 5250

04:13 <bmbeach> users manual where setting the bits only allow you (for i2s)

04:13 <bmbeach> to set the i2s clocks as outputs. i.e. the soc is master

04:13 <bmbeach> I had a thought about when I got alsa running, I had to keep

04:13 <bmbeach> bumping the frequency up, because the sound was to way slow

04:13 <bmbeach> , javier's sound wasn't too bad, and wizzup complained that

04:14 <bmbeach> it is way too fast. What if the epll isn't being set at all

04:14 <bmbeach> and the value is just a relic left over from the u-boot

04:14 <bmbeach> settup. javier is using u-boot-2015.10, I'm using a u-boot

04:14 <bmbeach> that came from git a long time ago. What u-boot is Wizzup

04:14 <bmbeach> using?

04:14 <bmbeach> I believe I have tracked down the dreaded smoking speaker

04:14 <bmbeach> problem on snow. I need to start alsa, take a snapshot of

04:14 <bmbeach> the codec state and then hit the kill switch fast. The

04:15 <bmbeach> problem is not one of playing the speakers too loud, but

04:15 <bmbeach> something else and very ugly. It comes from the d-class

04:15 <bmbeach> amplifiers capability to deliver enormous power at

04:15 <bmbeach> ridiculously high frequencies. If I'm right I will be able

04:15 <bmbeach> to tell you the registers to disable tomorrow. Muting the

04:15 <bmbeach> speaker does work.

04:16 <bmbeach> If you ever hear a very high pitched whine comeing from you speakers

04:16 <bmbeach> it means that they are melting. The problem is that it can happen with no

04:16 <bmbeach> sound at all.

21:19 <prahal____> audio playback stalls on odroidu2 exynos4412 (if I get back to 3.19 it "plays" in that the hw_ptr go forward), any track to follow at first ?

21:21 <prahal____> I tried reverting pl330 changes, samsung i2s, max98090 and going back from simple-card to odroid_max98090 driver but no go (I ahve been unable to plain revert one component in isolation so it turns out to be an erratic mode of debug

21:23 <prahal____> mind I tried u-boot from hardkernel , from tobias jakobi stable and devel (former with mpll at 800, the latter at 880

21:25 <Wizzup> and mainline u-boot?

21:25 <prahal____> and mileage vary (also my older hardkernel based tree fails to detect the emmc partitions , and the oldest trigger glibc fatal error "kernel too old" ... so I cannot test all with the same u-boot

21:25 <Wizzup> I have an odroid u2 at home, can try when I get home (in a week or so)

21:25 <prahal____> by the way long ago I had a round for samsung i2s psr that fixed playback speed issues

21:26 <prahal____> I saw talks about this issue , it might come in handy

21:26 <prahal____> http://paste.debian.net/343089/ <- here it is

21:29 <Wizzup> worth a shot - can try it tomorrow likely :)

21:31 <prahal____> could it be I "killed" the playback path (by a wrong setting I would have turned of on step) ? I read that alsa could stall if so

21:33 <prahal____> I would welcome a test on u2 , ... though a hint would do well too :)

21:34 <prahal____> there are too many parts to test , I could narrow the scope a bit that would do