Current Server Time: 17:59 (Central Europe)

#apertus IRC Channel Logs

2024/02/19

Timezone: UTC


02:06
bogdan
joined the channel
02:07
bogdan
changed nick to: BogdanXOR
02:14
BogdanXOR
left the channel
02:34
BogdanXOR
joined the channel
02:39
BogdanXOR
left the channel
03:19
BogdanXOR
joined the channel
03:23
BogdanXOR
left the channel
03:47
Bertl
left the channel
03:49
BogdanXOR
joined the channel
03:54
BogdanXOR
left the channel
04:19
BogdanXOR
joined the channel
04:23
BogdanXOR
left the channel
05:19
BogdanXOR
joined the channel
05:23
BogdanXOR
left the channel
05:49
BogdanXOR
joined the channel
05:53
BogdanXOR
left the channel
06:19
BogdanXOR
joined the channel
06:23
BogdanXOR
left the channel
06:49
BogdanXOR
joined the channel
06:53
BogdanXOR
left the channel
07:05
BogdanXOR
joined the channel
08:04
BogdanXOR
left the channel
08:18
BogdanXOR
joined the channel
08:20
aombk
left the channel
08:20
aombk
joined the channel
10:34
anuejn
Bertl: a few notes on the hdmi capture from my side (since I was on holiday last meeting)
10:35
anuejn
I have (or better had) a working setup for capturing hdmi signals and have hdl code to do so
10:36
anuejn
(cf. https://github.com/apertus-open-source-cinema/naps/tree/main/naps/cores/hdmi/rx)
10:36
anuejn
the hardware this is running on is not open source yet (because I didnt get around properly documenting it) but is mainly a raspi compute module and an ECP5
10:37
anuejn
I think your setup is actually more handy to get deeper insights into HDMI but this might be of interest for decoding
10:38
BogdanXOR
first time I'm seeing python as hdl :))
10:39
anuejn
there is also https://warmcat.com/2015/10/25/hdmi-capture-and-analysis-fpga-project-6.html (and the parts in the series before) that might be interesting
10:39
anuejn
BogdanXOR: thats amaranth HDL where you write a python script, that then generates the netlist
10:40
anuejn
I like it quite a lot because you can use all sorts of metaprogramming tricks and build yourself tools to make your life easier
10:41
BogdanXOR
can it also output the corresponding hdl itself (I see it uses Verilog) - or goes directly to synthesis?
10:47
anuejn
it can directly invoke the synthesis tool or it can output verilog
10:47
anuejn
however you like
10:54
anuejn
Bertl: looking at your linked files again
10:54
anuejn
that is really nice!
10:54
anuejn
🥳
10:54
anuejn
do you also have the code for generating those images?
13:48
mithro
left the channel
13:54
mithro
joined the channel
13:57
se6astian
Bertl is missing
13:58
se6astian
since 4 in the morning
13:58
se6astian
[04:47:21] <-- Bertl (*email address removed*) has quit (Ping timeout: 260 seconds)
14:08
BogdanXOR
left the channel
15:04
BogdanXOR
joined the channel
15:08
BogdanXOR
left the channel
15:09
BogdanXOR
joined the channel
15:31
Bertl
joined the channel
16:59
se6astian
and Bertl is back!
16:59
Bertl
indeed!
17:00
se6astian
for the MEETING TIME
17:00
se6astian
I just finished preparing the beta in the CP prototype enclosure for raw video capture
17:01
se6astian
the connectivity issues were resolved by buying an air compressor and waiting a few weeks :)
17:01
se6astian
before I started taking the pcbs appart today I quickly tested if training would fail again and it didnt :P
17:02
Bertl
I'm pretty sure, the air-compressor part was relevant!
17:04
se6astian
yes, the high pressure the connectors felt
17:06
se6astian
ok Bertl any news to report?
17:06
Bertl
yes, quite some news here ...
17:07
se6astian
cool!
17:08
Bertl
as I reported last time, I did some captures on the RPI regarding HDMI and I continued with other RPI formats (e.g. 800x600) and made all the decoding work properly
17:08
Bertl
I then switched to the AXIOM Beta and first updated the gateware to build with recent Xilinx tools (Vivado 2023.2)
17:10
Bertl
after that was sorted, I added a TMDS logic analyzer version of the HDMI output module to easily capture the 10bit code symbols directly from a PMOD debug module in one slow, while having the same output on the HDMI plugin in the other slot
17:11
Bertl
a bunch of sampling and decoding scripts later, I got a complete capture of our HDMI stream basically as it is sent over the cable for the first time
17:11
Bertl
together with the decoding already done on the RPI streams, I could quickly identify any differences between the streams
17:12
se6astian
very interesting!
17:12
Bertl
after some comparison, I quickly found the issue why we never had any success with the data islands (info packets)
17:13
Bertl
there was a small but tricky bug in the way the data was generated and after fixing this in the gateware, we are now able to send AVI info frames and other data over HDMI
17:14
Bertl
once that was done and tested, I also added two additional configuration bits to the scan registers which allow for easy inversion of the sync signals without messing with the scan configuration
17:15
Bertl
the fix to the data island generation can be found here: http://vserver.13thfloor.at/Stuff/AXIOM/BETA/delta-cmv_hdmi3_clean_raw-hdmi-di-fix.diff
17:15
Bertl
the add-on sync inversion here: http://vserver.13thfloor.at/Stuff/AXIOM/BETA/delta-cmv_hdmi3_clean_raw.2023.2-syncinv.diff
17:16
Bertl
while I was at it, I also improved the pblock layout and cell asignment to allow for faster (and more correct) building of the gateware
17:16
Bertl
the entire gateware including the prebuilt bitstreams can be found here: http://vserver.13thfloor.at/Stuff/AXIOM/BETA/cmv_hdmi3_clean_raw_fixed.2023.2/
17:17
se6astian
excellent, I see you talked to vup/anuejn already how to get the new gateware binaries upstream already
17:17
Bertl
well, it was a one-way discussion so far, but I'm optimistic :)
17:18
Bertl
here are the scripts to encode the data (on the Beta): http://vserver.13thfloor.at/Stuff/AXIOM/BETA/HDMIDI/
17:20
Bertl
and captures of our new and improved output can be found in the old location here: http://vserver.13thfloor.at/Stuff/AXIOM/HDMI/
17:20
se6astian
did you check if this makes a difference on BogdanXOR atomos ninja?
17:20
Bertl
I also tested this on the Ninja V+ from BogdanXOR and indeed activating the AVI info frame switches the color space
17:21
Bertl
and the vendor information is displayed as well
17:22
Bertl
unfortunately I was not able to record anything on the Ninja V+, but maybe I simply do not understand how to configure it correctly
17:23
Bertl
I also tested with the RPi and I could not record that input either, so it is definitely something wrong in the Ninja V+ configuration (or maybe it is not capable of recording 1080p60 in raw)
17:24
se6astian
cool, is there a recording media in the ninja (2.5" SSD caddy?)?
17:24
Bertl
my plan here is to get the Ninja V+ to you when we meet next time and you can test this
17:24
Bertl
yes, I have an SSD attached to the Ninja
17:26
se6astian
right, I can also try, but on the ninja itself all you need to record with a connected ssd is hit the red button on the touchscreen
17:26
Bertl
regarding firmware 2.x, the current gateware is a drop-in replacement for the existing one and it should behave exactly the same as long as you do not configure any info frames
17:26
se6astian
lower left corner here: https://i.pcmag.com/imagery/reviews/06lZ22NSdwy2FRr0qit24pb-9.fit_scale.size_1028x578.v1572399655.jpg
17:26
Bertl
well, if it just was that simple ... (regarding ninja)
17:27
Bertl
there are several configurations for HDMI and for the codec used
17:27
Bertl
and depending on the selection, I get either a nice picture, but a circle in the top left corner stating 'no raw input'
17:27
se6astian
yes, but codec options should just change the amount of remaining record time
17:28
Bertl
and the record button is greyed out
17:28
se6astian
well yes "raw" rings the alarm bells, what do you send from the beta currently?
17:28
Bertl
or alternatively, I get no picture at all (both RPi and Beta)
17:28
se6astian
not RGB24 in 1920x1080 ?
17:29
Bertl
we are sending 1080p60 24bit, plain and normal image
17:29
se6astian
right, definitely a fine tuning of metadata then I guess
17:29
Bertl
didn't bother activating any of the raw encoding yet, but it would still be a normal image
17:30
Bertl
what I read online is that the 'RAW' mode of the Ninja V+ is actually something special which only works with a small number of cameras
17:30
se6astian
yes it expects proprietary bayer data in an undisclosed format then
17:30
Bertl
and those seem to abuse the HDMI output to send raw data in some other form (maybe not even HDMI)
17:31
Bertl
so very likely this kind of 'RAW' data will not work, unless Atomos decides to open up some of those formats or add our own format
17:32
Bertl
but as I said, no Ninja expert here, but the HDMI output is now solid and we get a correct picture on the Ninja V+
17:33
se6astian
wonderful!
17:33
Bertl
btw, reading through the standards it seems that the Ninja is non-standard conform in assuming YCbCr when there are no AVI info frames
17:34
Bertl
as the to-be-assumed default seems to be RGB 444
17:35
Bertl
also, the gateware was tested on the latest firmware 2.x build
17:35
Bertl
so it definitely works fine there without any changes
17:36
se6astian
great!
17:36
Bertl
that's it from my side, if there are any questions regarding procedure or anything else, please let me know
17:37
se6astian
next steps: 1. integrate into automatic firmware builds on github (anuejn/vup)
17:37
se6astian
2. create documentation - could you take care of that bertl?
17:38
Bertl
there are no real changes to the register layout except for the two bits to invert the sync
17:38
se6astian
but to fill/write the data islands?
17:38
Bertl
that's the same as it was before
17:39
se6astian
but there is no documentation about it
17:39
Bertl
ah, okay, that's possible, where would/should that docu go?
17:40
se6astian
ultimately the wiki but since you hate the wiki and txteditor is fine and I will put it i the wiki
17:40
Bertl
okidoki
17:41
se6astian
great, anyone else around?
17:43
se6astian
ok then, MEETING CONCLUDED, thanks for the great progress Bertl!
17:43
Bertl
you're welcome! thanks for the moderation se6astian!
17:44
se6astian
my pleasure, now off to the supermarket!
20:42
BogdanXO1
joined the channel
20:42
BogdanXOR
left the channel
21:29
BogdanXO1
left the channel
21:34
BogdanXOR
joined the channel
21:38
BogdanXOR
left the channel
22:04
BogdanXOR
joined the channel
22:09
BogdanXOR
left the channel
22:34
BogdanXOR
joined the channel
22:39
BogdanXOR
left the channel
23:02
Spirit532
left the channel
23:02
Spirit532
joined the channel
23:04
BogdanXOR
joined the channel
23:09
BogdanXOR
left the channel
23:16
BogdanXOR
joined the channel