msimpson has quit [Remote host closed the connection]
gaston1980 has quit [Quit: Konversation terminated!]
fl_0 has quit [Ping timeout: 256 seconds]
fl_0 has joined #linux-sunxi
tnovotny has joined #linux-sunxi
dev1990 has joined #linux-sunxi
Mangy_Dog has joined #linux-sunxi
<jernej>
anyone have an idea how to measure RTC RC oscillator frequency? Possibly programmatically?
kilobyte_ch has quit [*.net *.split]
clementp[m] has quit [*.net *.split]
leio has quit [*.net *.split]
quark_ has quit [*.net *.split]
kilobyte_ch has joined #linux-sunxi
quark_ has joined #linux-sunxi
clementp[m] has joined #linux-sunxi
leio has joined #linux-sunxi
JohnDoe0 has joined #linux-sunxi
JohnDoe_71Rus has quit [Ping timeout: 264 seconds]
<mru>
jernej: which chip/board?
<mru>
it should be somewhere around 32 kHz
<jernej>
I'm talking mostly about 40 nm chips, like H3, H5, A64, etc.
hlauer has quit [Ping timeout: 258 seconds]
<jernej>
yeah, that's the thing, it's not close enough for HDMI-CEC to work
<mru>
no, it's useless
<mru>
most boards have an external crystal
<mru>
use that
<jernej>
if I manually lower divider from 16 to 15, CEC works
<jernej>
most H3 boards don't have external crystal
<mru>
then don't use them
<jernej>
:)
<mru>
the rc oscillator is too unstable to be useful
<jernej>
it's not not me, it's many LE users :D
<mru>
but if you want to measure it, send it to a pin and poke it with a frequency counter
<jernej>
I did that, but my, albeit cheap, frequency counter always showed 32 kHz
<jernej>
even if changed divider in extremes
<jernej>
anyway, it would be nice to somehow compare RTC RC oscillator to timer or hstimer which would allow automatic adjustments
<jernej>
but I'm not sure how to do that
netlynx has joined #linux-sunxi
netlynx has quit [Changing host]
netlynx has joined #linux-sunxi
<mru>
you'd need a counter for that
<jernej>
trusted watchdog could be abused for that (LOSC source, counter, interrupt on match), but that's too hackish
<mru>
probably too much jitter in the interrupt handling to be useful too
<jernej>
well, 32 kHz is not that fast or precise
<jernej>
and in most cases prescaler should be shifted just for one up or down
tnovotny has quit [Quit: Leaving]
cnxsoft has quit [Quit: cnxsoft]
<mru>
hmm, h3 doesn't have a clock output pin like a20 does
mripard has quit [Ping timeout: 260 seconds]
<jernej>
it has - X32KFOUT
<mru>
ah, it's a fixed function pin
mripard has joined #linux-sunxi
<mru>
and not connected on any of the h3 boards I have
_whitelogger has joined #linux-sunxi
warpme_ has quit [Quit: Connection closed for inactivity]
cnxsoft has joined #linux-sunxi
elros1 has joined #linux-sunxi
[ElectricShit] is now known as [GrinToLife]
ndufresne has joined #linux-sunxi
azend has quit [Ping timeout: 256 seconds]
azend has joined #linux-sunxi
<smaeul>
jernej: on H6 there is a hardware calibration circuit. on older chips, you have to set the divider manually
<smaeul>
the way I do it in crust is to set CPUS to IOSC, and then count HOSC cycles vs IOSC cycles
<smaeul>
the AR100 has a built-in timer that counts at whatever speed the CPU clock source runs at
<jernej>
smaeul: I know that H6 has calibration mode, but I'm interested in older, 40 nm SoCs
rtp has quit [Read error: Connection reset by peer]
xes_ has joined #linux-sunxi
xes has quit [Ping timeout: 240 seconds]
<jernej>
smaeul: so you do some kind of 32768 Hz calibration in Arisc?
<jernej>
I mean AR100
<jernej>
anyway, that's still of the table for H3
<jernej>
s/of/off/
<smaeul>
well I don't care about 32768Hz directly. I care about IOSC so I can calculate the divider for R_UART baud rate
<jernej>
true, but same thing can be achieved
<smaeul>
25% error is way out of UART spec :)
<jernej>
I know that issue very well from work
<jernej>
we kept usually in +/- 2% range
<jernej>
but they were cortex m4 mcus
<smaeul>
jernej: TIMER0/TIMER1 can run from the same "about 32KHz" clock source
<smaeul>
for calibration, count the number of 24MHz cycles in a "~32kHz" cycle, then multiply by 32*INTOSC_CLK_PRESCAL to get IOSC frequency
<smaeul>
you can average across as many cycles as you need
<jernej>
that's interesting, thanks!
<jernej>
sadly both timers use same source
vagrantc has joined #linux-sunxi
<smaeul>
jernej: you mean TMR0_CLK_SRC and TMR1_CLK_SRC aren't independent?
<jernej>
oh, I misinterpret diagram
<smaeul>
where during boot do you plan to do this calibration?
<jernej>
good question, I thought to make it cyclic in Linux due to temperature drift
<jernej>
but I'm not sure it's worth it
<jernej>
but maybe I should try with U-Boot first
<smaeul>
hmm, I could measure temperature effects, but I think random variation might be too large for it to matter
<smaeul>
oh, s/multiply/divide/ above
<smaeul>
you can reduce the rounding error (and thus the time needed) by setting the divider to 1 before calibration, but only if nothing else is using the clock
<smaeul>
the general procedure would be 1) set up both timers 2) start slow timer 3) start fast timer 4) spin until slow timer reaches some value 5) read fast timer 6) do math
<jernej>
I know the theory :)
elros1 has quit [Read error: Connection reset by peer]
gaston1980 has joined #linux-sunxi
dev1990 has quit [Quit: Konversation terminated!]
lurchi_ is now known as lurchi__
<smaeul>
did some quick testing with the calibration code in crust: http://ix.io/2wkQ if the divider being off by one matters, then temperature correction is needed
<jernej>
which frequency is that? RC oscillator is about 16 MHz
<smaeul>
jernej: that is the RC oscillator. on every A64 device I have, it's 12-13 MHz
<jernej>
smaeul: Are you sure? User manual says it's 16 MHz and currently mainline has it set to 16 MHz too (probably doesn't matter)
vorian45620 has quit [Remote host closed the connection]
<smaeul>
jernej: yes, I'm sure. that's why I mention 25% UART baud error. I originally hardcoded 16MHz, and R_UART didn't work. It works up to (at least) 9600 baud with my calculated frequency