<wpwrak>
lekernel: just not working or with transient errors ?
<lekernel>
works (but not very reliably, especially on the second port of the board) when no video signal is present, fails reliably 100% when the computer is sending a video signal
<wpwrak>
crosstalk in the cable ?
<lekernel>
I tried an expensive HDMI cable that says "double shielded", no change
<wpwrak>
so marketing does not prevent the issue :)
<Fallenou>
=)
<wpwrak>
add a scope and see what the signal looks like ?
<wpwrak>
or does that make the trouble go away ?
<lekernel>
and I'd think 100pF would remove the >25+ MHz crosstalk from the video
<lekernel>
I don't have a DSO
<lekernel>
with the saelae logic analyzer, the waveform looks ok, except sometimes the fpga doesn't ack the i2c address
<lekernel>
seems the fpga always fails to ack when there is video
<wpwrak>
good luck debugging what may be signal integrity with an LA :)
<wpwrak>
you have no scope at all ? even an analog one may be able to show you what's going on
<lekernel>
if there is a simple way to get my GPU to send a permanently toggling pattern on the DDC...
<wpwrak>
can't you just generate on in the M1 ?
<lekernel>
it's a HDMI receiver
<lekernel>
not transmitted
<lekernel>
transmitter
<larsc>
if you use linux and it registers a i2c device for the bus and you have CONFIG_I2C_DEV enabled you can send i2c messages from userspace
<lekernel>
the i2c level translator is bidirectional though
<lekernel>
I said a simple way, spending hours figuring out some ioctl interface isn't
<larsc>
it's simple
<wpwrak>
there are tools ...
<larsc>
the package is called i2c-tools
<wpwrak>
i2c-tools
<lekernel>
and will the clock be permanently toggling?
<wpwrak>
larsc: you win :)
<lekernel>
ie can I get a simple, uninterrupted oscillation?
<larsc>
as long as you keep sending messages, although there will probably be small gaps
<larsc>
no
<lekernel>
and will that work even when x-windows is running and sending a video signal?
<larsc>
yes
<larsc>
the nouveau driver does bitbanging for the i2c bus
<larsc>
you could modify it and let it toggle scl in a regular interval
<wpwrak>
hmm, on my desktop: i2cdetect -l shows nothing. that's two video cards (nV and ati)
<lekernel>
regular is incompatible with PC-based bitbanging, there is always a lot of jitter
<larsc>
well thats all the hardware can do
<lekernel>
unless you install some realtime kernel
<larsc>
well rip your cable appart and connect a regular clock source
<lekernel>
hmm, I could also loopback the two channels and use one as transmitter
<wpwrak>
you shouldn't really need a periodic signal to check integrity. most likely, a problem affects everything
<lekernel>
except I won't get video
<wpwrak>
mmh ? are you saying I2C first works, then video starts, killing I2C, and then all goes dead ?
<lekernel>
yes
<lekernel>
but i2c doesn't even work reliably in the first place
<lekernel>
on the first port it works 90% of the time
<lekernel>
on the second port maybe 5%
<lekernel>
and it's the same on the two boards I built, so it's unlikely to be a solder problem
<wpwrak>
well, then you could just look at the non-video case first. maybe this already provides enough of a clue.
<wpwrak>
go you have any scope at all ?
<wpwrak>
s/go/do
<lekernel>
yes, an analog scope older than me
<wpwrak>
if the phosphorus is still good, that ought to do nicely :)
<Fallenou>
see you !
<GitHub120>
[mibuild] sbourdeauducq pushed 1 new commit to master: http://git.io/e0RpYw