Current Server Time: 06:43 (Central Europe)

#apertus IRC Channel Logs

2020/10/19

Timezone: UTC


00:07
mumptai
left the channel
02:36
futarisIRCcloud
joined the channel
05:43
BAndiT1983|away
changed nick to: BAndiT1983
06:42
Bertl_zZ
changed nick to: Bertl
06:42
Bertl
morning folks!
07:36
mumptai
joined the channel
07:53
se6ast1an
good day
09:06
Bertl
off for now ... bbl
09:06
Bertl
changed nick to: Bertl_oO
09:22
illwieckz
left the channel
09:36
illwieckz
joined the channel
09:40
illwieckz
left the channel
10:04
illwieckz
joined the channel
10:18
EmilJ
Got frustrated figuring out u-boot, I'll do that another day when I actually have basic understanding of the OS and boot procedure
10:18
EmilJ
I moved the frame buffer to the top of RAM
10:18
EmilJ
which is a *total* hack but this is just a demo, really
10:19
EmilJ
I'm memsetting the frame buffer at 115 MB/s = 920Mbps (1050Mbps is specified speed for my DDR3 RAM on the Pynq Z2)
10:20
EmilJ
...however, without regard to which of the buffers is in use. Anyway, I've figured out that *only the lowest byte* of the words is used :D
10:35
EmilJ
actually, no, that doesn't seem the case. Memsetting 0x000000FF to the frame buffer definitely does draw white on the screen... but writing 0x000000FE already makes visible single pixel vertical lines
10:52
EmilJ
oh I think I get it. The HdmiBufferReader is only used in camera.py and isn't a generic thing, but is probably designed for whatever cursed format comes out of the camera
10:53
EmilJ
https://github.com/apertus-open-source-cinema/nmigen-gateware/blob/b9cae2e3678aa96ff9efd392b2036d2ebf747cba/src/cores/hdmi/hdmi_buffer_reader.py#L83
10:53
illwieckz
left the channel
11:01
xerrox_
joined the channel
11:08
illwieckz
joined the channel
11:22
anuejn
EmilJ: yup currently HdmiBufferReader is only able to output grayscale (as you can see)
11:23
anuejn
moreover the decoding is somewhat messed up (now that i read it again)
11:23
anuejn
however you should be able to modify it so that it fits your needs :)
11:24
anuejn
this is a dataformat that somehow packs 4 12bit raw pixels into one 64bit memory word
11:44
intrac
left the channel
11:46
intrac
joined the channel
13:21
intrac_
joined the channel
13:23
intrac
left the channel
13:33
EmilJ
guess who forgot memset works on bytes and not words
13:34
EmilJ
^ this guy
13:38
EmilJ
would be nice if RAM-alocated buffers showed up in the `memorymap:` from `address_assignment_hook`. I might attempt to hack that together
13:57
comradekingu
left the channel
14:10
comradekingu
joined the channel
14:11
Bertl_oO
EmilJ: 'RAM-allocated buffers'?
14:40
intrac
joined the channel
14:42
intrac_
left the channel
14:54
intrac
left the channel
15:18
anuejn
EmilJ: sounds like a nice idea :)
15:27
vup
Bertl_oO: the RAM framebuffer
16:00
intrac
joined the channel
16:00
se6ast1an
Meeting time!
16:00
se6ast1an
who is here?
16:01
Bertl_oO
changed nick to: Bertl
16:01
Bertl
is here ..
16:02
vup
is here for another 15 minutes
16:03
se6ast1an
great, vup any news to report? like vivado ?
16:05
vup
well, recently I worked on starting to automate the gateware build process, as that is one of the remaining "binary blobs" that are not automatically built yet
16:06
vup
currently this is tested in the nmigen-gateware repository to check if the various gatewares we create there synthesis for their various configurations and it works quite well there so far
16:06
Oscqar
joined the channel
16:06
Oscqar
Hi!
16:07
vup
so now the plan is to split off the gateware from the axiom-firmware repository into its own and start to automatically create the required bitstreams when the gateware changes instead of just downloading prebuilt ones
16:08
vup
so yeah thats it from me this week :)
16:08
se6ast1an
great, news, many thanks!
16:08
se6ast1an
*great news
16:09
se6ast1an
hi oscar!
16:10
se6ast1an
how are things going?
16:11
Oscqar
Good. Just finished and sent in the quarterly tax return. (BTW quite a week with COVID test but it turned out negative)
16:11
Oscqar
How are things over there?
16:12
se6ast1an
good to hear!
16:12
se6ast1an
things here are busy, as always :)
16:13
Bertl
we have an all-time-high regarding covid-19 in Austria ...
16:14
Oscqar
Yes, I read about it. Same over here in Belgium. Cafes and restaurants just closed again.
16:15
se6ast1an
quick updates from me: the PCB panel collection of offers continues
16:15
se6ast1an
currently what we know: rogers material (far too expensive), TU 883 and similar (slightly too expensive), FR4 -> cheap
16:16
se6ast1an
while we are hoping that standard FR4 will work, acutal tests will show
16:16
se6ast1an
after Bertl finished assembling and testing of new PB we will order a set in FR408 and FR4 for automated assembly
16:16
se6ast1an
the search for a gigahertz oscilloscope for that purpose continues
16:17
se6ast1an
in other news: I brought the pcb prober back to Bertl today after I filmed it in action yesterday (for new TT most likely)
16:17
se6ast1an
we attended the virtual google mentor summit last week
16:18
se6ast1an
thats it from me
16:18
comradekingu
We have gigahertz oscilloscopes at the radio enthusiast club at the uni in Trondheim
16:18
se6ast1an
nice :)
16:19
comradekingu
Swing by before the coronas take over
16:19
Bertl
comradekingu: send them over! :)
16:19
se6ast1an
we are also trying to get in touch with local universities
16:19
se6ast1an
so far no luck
16:20
comradekingu
Might try the military guys and various corps have them
16:20
comradekingu
What city is it?
16:21
se6ast1an
We are in Vienna
16:21
comradekingu
Are you going full ted etching your own boards?
16:21
se6ast1an
yes we should try the austrian military :)
16:22
comradekingu
Each year in mid-December, officials from the 56 OSCE participating States gather in Vienna to exchange information on their armed forces, military organization, manpower and major weapon and equipment systems.
16:22
comradekingu
https://www.osce.org/fsc/74528
16:22
se6ast1an
https://image.shutterstock.com/image-photo/bad-ischl-austria-31-december-600w-1198717606.jpg
16:22
comradekingu
Guy in the middle there looks to have 0 GHz oscilloscopes
16:23
se6ast1an
Bertl: please finish us , as usually :)
16:23
Oscqar
:D
16:23
Bertl
tries some finishing moves ... but fails ...
16:24
Bertl
well, I tried to finalize the backup this week, but failed because the server is 'unstable'
16:25
Bertl
I also tried to update the Axiom Beta at the observatory with some neat features, but failed as well, problems with the SD card and similar
16:26
Bertl
on the positive side, I'm almost done with one side of the new power board prototype so we should be able to test that soon
16:26
Bertl
that's basically it from me for this week.
16:27
comradekingu
I only found https://www.semaf.at/ https://www.army-store.at/fahrzeuge
16:28
Bertl
yeah, both are not into multi GHz bandwidth as far as I know ...
16:29
se6ast1an
thanks Bertl
16:29
se6ast1an
anyone else with topics/reports for the meeting?
16:29
comradekingu
(help, I am in a meeting)
16:30
Bertl
just play dead, it's almost over :)
16:32
se6ast1an
its over!
16:32
se6ast1an
MEETING CONCLUDED
16:32
se6ast1an
many thanks everyone!
16:33
Oscqar
Great. See you soon!
16:34
Bertl
are you coming over?
16:35
Oscqar
Virtually I mean! I would like to though, but I'm afraid it has to wait a bit.
16:39
Oscqar
left the channel
16:53
EmilJ
I don't think you should limit yourself to military people in search of an oscilloscope to borrow for a weekend. I've seen 40GS/s Teledynes in Siemens in Prague, and those guys only do industrial automation and some train stuff
16:54
EmilJ
so that's 4GHz bandwidth
16:54
Bertl
would be perfect
16:58
EmilJ
is wondering if the axiom-firmware is flexible enough to work with the UART, ETH and USB conections on my zynq board, and how much work would it be to adapt it
16:59
Bertl
you very likely need a proper (hand crafted) device tree, but otherwise I think you should be fine with minimal modifications
17:00
EmilJ
actually, it's not clear to me from the website, what state are the dev kits in - this project in general seems to me like contributing to it is potential master's thesis material
17:01
EmilJ
so I may be interested on getting my hands on one
17:31
EmilJ
...okay, I'm looking at the wiki, and I realized how high-end the entire thing is, wow
19:03
mumptai
left the channel
19:30
se6ast1an
EmilJ: I think the beauty is that you can start with very high level modifications like creating a bash script but if you want to go all the way down to the fpga code
19:30
se6ast1an
you have that option
19:31
se6ast1an
I don't know any military source of an osci
19:31
se6ast1an
I only have conncetions to industrial and university folks, lets see where that takes us
19:58
mumptai
joined the channel
20:19
vup
EmilJ: u-boot config and devicetree should be enough to make axiom-firmware work
20:20
vup
With u-boot probably being the hard part
20:20
vup
We don't use the fsbl generated by vivado, but instead boot using u-boot spl
20:21
vup
Which needs a c file that inits the soc, which can be generated by vivado when you create a fsbl project with it
20:23
EmilJ
u-boot spi? Meaning axiom firmware is supposed to live in SPI connected flash?
20:23
vup
spl
20:23
vup
Second program loader, a alternative to fsbl
20:23
vup
https://github.com/apertus-open-source-cinema/axiom-firmware/blob/master/patches/u-boot/z_turn_lite_support.patch
20:23
vup
Thats how adding support to u-boot for a board looks like
21:11
Bertl
where you can probably keep the current u-boot for a start
21:11
Bertl
(i.e. the one you already have on your board)
21:13
xerrox_
left the channel
21:14
xerrox_
joined the channel
21:37
EmilJ
geez what a massive file
21:38
EmilJ
autogenerated register definitions, I see
21:54
EmilJ
update: I can write to the buffer, and it looks like both vsync and hsync are broken
21:54
EmilJ
https://irc.tywoniak.eu/file/1/NjV3rrRWgQwthPMR
21:54
EmilJ
wow, convos can auto host files. That's wonderful
21:55
EmilJ
this image is, in theory, me writing *one* dark pixel, to x=y=1
21:56
EmilJ
instead of that I get these silly broken lines and they *move* over time
22:00
Bertl
what gateware do you use at the moment?
22:10
BAndiT1983
changed nick to: BAndiT1983|away
22:11
EmilJ
Uh, nmigen gateware, with the HDMI buffer reader, which supposedly was never successfully tested to work :D
22:12
Bertl
ah, okay, yeah, then you have to check with vup/anuejn I guess
22:13
Bertl
but I presume it is somewhat modeled after the one used on the Beta at the moment
22:13
Bertl
does the movement repeat? is it slow or fast?
22:26
EmilJ
It's real fast and does not appear to repeat every buffer_umber of frames, no. What's interesting is that I've had vertical lines working without any issue when I was wildly memsetting.
22:31
Bertl
could be that the vertical lines provided enough information for the monitor to do the sync
22:32
Bertl
but double check sync inversion and timing, this is one of the sensitive points
22:32
Bertl
does the devel board have a direct HDMI connection?
22:38
Bertl
you are using the PYNQ-Z2 right?
22:41
Bertl
so there you have directly connected HDMI ports and you want/need to use the TMDS_33 IO standard on a 3V3 io-power rail
22:42
Bertl
at 1080p60 signal quality could be at the limit there
23:16
xerrox_
left the channel
23:30
mumptai
left the channel