Current Server Time: 09:04 (Central Europe)

#apertus IRC Channel Logs

2021/01/04

Timezone: UTC


01:39
davidak[m]
left the channel
02:48
Bertl_oO
off to bed now ... have a good one everyone!
02:48
Bertl_oO
changed nick to: Bertl_zZ
03:18
anuejn
se6ast1an: could you create a collection of different raw frames from the beta somewhere
03:18
anuejn
I need one to tweak the parameters of the wavelet compressor
03:19
anuejn
best would be one image per scene as the compression scheme does not do inter frame compression
03:19
anuejn
off to bed now
03:19
anuejn
good night everyone :)
03:53
futarisIRCcloud
joined the channel
06:17
BAndiT1983|away
changed nick to: BAndiT1983
06:29
eppisai
joined the channel
07:14
RexOrCine
left the channel
07:15
RexOrCine1
joined the channel
07:17
RexOrCine1
left the channel
07:17
RexOrCine
joined the channel
07:59
eppisai_
joined the channel
08:01
eppisai
left the channel
08:01
eppisai_
changed nick to: eppisai
08:39
se6ast1an
anuejn: did you see https://wiki.apertus.org/index.php/AXIOM_Beta/Video ?
08:45
BAndiT1983
se6ast1an: but where are the RAW files? seeing only MP4, MOV and DNG
08:45
se6ast1an
Timelapse footage from AXIOM Belgium - Nextcloud | RAW12 and HD 1080 MP4.
08:46
se6ast1an
but we have more :)
08:46
se6ast1an
anuejn: see http://files.apertus.org/AXIOM-Beta/snapshots/
08:46
BAndiT1983
ah, now i see, thanks
08:47
BAndiT1983
this one is also interesting -> http://files.apertus.org/AXIOM-Beta/samples/reschart/
09:16
se6ast1an
hi eppisai!
09:16
eppisai
hello!
09:17
se6ast1an
I added a comment to https://lab.apertus.org/T1202 and would like to talk to you if you could take over that task as it also evolves around your checkbox icon
09:18
eppisai
yes, Ofcourse!
09:18
se6ast1an
great, please take a look
09:18
eppisai
I'll look into it. Thanks!
09:18
se6ast1an
if anything is unclear let me know then we can discuss it
09:19
se6ast1an
I used the home icon for testing https://lab.apertus.org/file/data/57jqrnw3ipuoc4o4o36e/PHID-FILE-qo5uf5lqryypqdwdrxt5/image.png
09:19
BAndiT1983
hi eppisai, please also check the task https://lab.apertus.org/T1193 before that, as the icon needs slight adjustment
09:20
eppisai
ok, I will see it!ðï¸
09:34
Bertl_zZ
changed nick to: Bertl
09:34
Bertl
morning folks!
09:34
BAndiT1983
hi
09:36
mumptai
joined the channel
09:36
se6ast1an
good day!
09:47
LordVan
joined the channel
10:38
eppisai
left the channel
11:37
anuejn
se6ast1an: thanks
12:35
eppisai
joined the channel
12:59
eppisai
left the channel
13:44
EmilJ
hi all
13:48
anuejn
hi
13:48
anuejn
just as a reference re earlier: I collected interesting frames here: inverse_multi_stage_wavelet2d(transformed, 3)
13:48
EmilJ
anuejn: I'm looking into hooking up an EDID reading setup from the HDMI DDC, I'm seeing some traces of bit-banged I2C in camera_usb3.py but not the software side
13:49
anuejn
whoops wrong clipboard
13:49
anuejn
https://files.niemo.de/axiom_sample_dng/
13:49
anuejn
EmilJ: the software side is the linux kernel
13:50
EmilJ
ohhh cool
13:50
anuejn
we load kernel modules with devicetree overlays
13:50
anuejn
see https://github.com/apertus-open-source-cinema/nmigen-gateware/blob/master/src/lib/peripherals/mmio_gpio.py and https://github.com/apertus-open-source-cinema/nmigen-gateware/blob/master/src/lib/peripherals/i2c/bitbang_i2c.py
13:51
anuejn
the overlay generation is - eh - a bit ad hoc ;)
13:51
anuejn
basically we load a gpio driver and then plug the i2c driver into that
14:03
anuejn
for the overlays to work you need some things
14:03
anuejn
we have a patched kernel (https://github.com/apertus-open-source-cinema/axiom-firmware/blob/master/patches/linux/overlay.patch)
14:27
EmilJ
thanks, I'll try to replicate that
14:27
EmilJ
oh no, a kernel patch
14:29
EmilJ
well the pynq has its own machinery, I'll see what I can do
14:31
xthenode
joined the channel
14:33
xthenode
happy Monday...
14:34
anuejn
EmilJ: you can try to just use the axiom-firmware images
14:34
anuejn
maybe that is easier to get started with
14:34
xthenode
what cad system is used for axiom hardware?
14:35
xthenode
camera body etc...
14:37
BAndiT1983
onshape
14:56
xthenode
thanks, ill check it out...
15:11
LordVan
left the channel
15:23
EmilJ
hmmm, can't figure out what framebuffer *driver* does axiom-firmware use
15:24
EmilJ
my devicetree is set, the kernel config of pynq definitely is configured for framebuffer work
15:24
EmilJ
but I don't get `/dev/fb0`
15:26
Bertl
anything in dmesg?
15:29
xthenode
left the channel
15:37
vup2
is uses the simple-framebuffer driver
15:41
vup2
EmilJ: how is your devicetree looking?
15:47
eppisai
joined the channel
15:54
EmilJ
Bertl: that was a good idea, apparently it is refusing to create the overlay. The error I'm getting (create_overlay: Failed to create overlay (err=-22)) typically is due to problems with devicetree symbols
15:55
EmilJ
either caused by the overlay lacking symbols, or the device tree lacking them
15:55
EmilJ
I'm in neither of those situations though
15:55
EmilJ
yeah I think it's time to switch to axiom-firmware
15:57
vup2
EmilJ: you need to compile your devicetree with the '-@' option to get symbols
15:57
vup2
both the overlay and the one loaded at boot time
15:58
EmilJ
I did that
15:59
vup2
that is ... curious
15:59
EmilJ
well, the overlay, at least, but I verified the `__symbols__` directory exists in the device tree directory
15:59
vup2
hmm
15:59
vup2
ahh
16:00
vup2
did you create the overlay yourself or did you use the one created by the gateware?
16:00
EmilJ
I used the emitted overlay
16:00
vup2
I see
16:01
vup2
do you have the source for your devicetree available?
16:01
EmilJ
well, yes, split into these .dtsi source fiels
16:02
EmilJ
it's not super transparent since xilinx assumes you're using their tools and abstracting everything, probably
16:02
vup2
because the overlay created by the gateware wants a part named `amba`
16:02
EmilJ
I do have that
16:02
vup2
maybe its not called amba for you
16:02
vup2
hmm
16:04
vup2
maybe its time to use axiom-firmware :P
16:26
EmilJ
Welp, that didn't boot
16:26
EmilJ
no UART either
16:27
vup2
well what did you do?
16:27
EmilJ
flashed a release for the micro onto a microSD card using um
16:29
EmilJ
gnome-disks
16:30
vup2
well
16:32
vup2
you probably need to use a proper u-boot and a proper devicetree
16:33
vup2
ie you need a pnyq folder that contains files similar to these: https://github.com/apertus-open-source-cinema/axiom-firmware/tree/master/boot/axiom-micro
16:33
vup2
(fsbl.elf is not needed anymore, bitstream.bit is not needed aswell)
16:34
vup2
if u-boot does not already support your board you need to add support for it using a patch aswell: https://github.com/apertus-open-source-cinema/axiom-firmware/blob/master/patches/u-boot/z_turn_lite_support.patch
16:34
vup2
the ps7_init_gpl.c thing can be extracted from vivado if you create a fsbl project targeting your board
16:55
Shohei
joined the channel
17:00
se6ast1an
MEETING TIM!
17:00
se6ast1an
who is here?
17:00
metal_dent[m]
is present
17:00
se6ast1an
*TIME :)
17:00
bluez
* bluez is here
17:00
Shohei
*Shohei is her
17:00
Shohei
e
17:00
se6ast1an
my keyboard tends to eat letters or double them recently...
17:01
se6ast1an
great, metal_dent[m] would you like to start the reporting?
17:01
metal_dent[m]
yeah sure!
17:01
metal_dent[m]
last week i worked on the image button PR (haven't pushed the changes yet)
17:02
metal_dent[m]
we also discussed about the desaturation shader task (T1215)
17:03
metal_dent[m]
today i was going through the code and learning more about FBO and shaders
17:03
anuejn
is here
17:03
vup2
is here aswell :)
17:03
Bertl
too ...
17:04
metal_dent[m]
other work was related to the USB code but i haven't tested that yet so waiting for the remote to come ':D
17:04
metal_dent[m]
i think that's it from me!
17:04
se6ast1an
great metal_dent[m], many thanks
17:05
se6ast1an
lets coordinate about the remaining imagebutton steps as I also started to work again on that area
17:05
metal_dent[m]
yes, sure
17:05
se6ast1an
bluez: your turn
17:05
bluez
yep!
17:06
bluez
so last week... as final task of my 'vhdl training', I translated an existing implementation of axi lite slave in verilog to vhdl
17:07
bluez
https://github.com/Swaraj1998/axi-lite-slave
17:07
bluez
code can be found here ^
17:08
bluez
i used the axi verification ip for verification using simple reads and writes
17:08
bluez
hit quite a few bumps.. mostly because of sparce documentation about some parts of vivado and all.. but did learn a lot
17:09
Bertl
would be nice to have the full project with the verification IP as well
17:09
bluez
i'll now run it in remote beta and do furter tests
17:09
Bertl
(online I mean :)
17:09
bluez
yep yep sure
17:10
bluez
i just did the verification today.. will add the tcl part on the repo so it can be reproduced
17:10
Bertl
perfect!
17:10
bluez
thats all from me i guess!
17:11
bluez
:)
17:11
se6ast1an
many thanks bluez!
17:11
se6ast1an
Shohei: your turn!
17:11
Shohei
Thank you :) I bought a MicroZed board, and tried I tried serial communication from my Linux laptop, and read this link (https://wiki.apertus.org/index.php/AXIOM_Beta/Enclosures) up to around "5."
17:11
se6ast1an
great
17:11
Shohei
I will do my best to catch up.
17:12
se6ast1an
point 6 is were it becomes really intersting :)
17:12
Shohei
Thank you! I will read it carefully : )
17:12
se6ast1an
many thanks for the report!
17:12
se6ast1an
vup2: your turn!
17:13
vup2
Ok sure, first of all, I did a quick fix for the axiom-firmware CI, for some reason it did not like the opennic dns server anymore, so I switched the dns server and the CI build of the firmware are working again :)
17:13
se6ast1an
nice!
17:14
vup2
Now furthermore, anuejn and I have been working on the wavelet compression some more. We have finally gotten lossless forward and backward transformation to work!
17:14
vup2
(in the python implementation)
17:14
vup2
you can find that here: https://github.com/apertus-open-source-cinema/nmigen-gateware/blob/3540c01/src/lib/video/wavelet/py_wavelet.py
17:15
vup2
now we are starting to experiment with different quantization parameters for the different wavelet parts to find a sweet spot of image size and image quality
17:16
vup2
If there are no questions, thats it from the this week
17:17
se6ast1an
what is the issue with image size vs speed vs quality ?
17:17
se6ast1an
how does it affect the process I mean
17:17
vup2
so the compression works by throwing away some bits of the high frequency components of the image
17:18
vup2
throwing away more of them makes the image compress better
17:18
vup2
but obviously lowers the image quality
17:19
se6ast1an
but it does not affect compression speed if I understood correctly?
17:19
anuejn
yup it doesnt
17:19
se6ast1an
image size would I assume (number of total pixels to compress)
17:20
anuejn
that does affect the runtime obviously
17:20
anuejn
almost linear
17:20
se6ast1an
would an FPGA implementation be able to work with arbitraray resolutions or would that make it hard to parallelize the process?
17:20
BAndiT1983
changed nick to: BAndiT1983|away
17:21
anuejn
the paralelization is not a big problem there
17:21
anuejn
(the algorithm is very well paralelizable)
17:22
anuejn
(currently there is no paralelization in the fpga implementation)
17:22
anuejn
the main problem with arbitrary resolutions is that you cant allocate buffers dynamically in an fpga
17:22
anuejn
so we can set an upper bound during synthesis and then support everythnig that is lower than that
17:23
anuejn
however dynamic resolutions are also not yet implemented
17:23
se6ast1an
great, that should do, as the image sensor max resolution is "static" but reading smaller windows would increase FPS
17:23
se6ast1an
but less pixels of a smaller window would then also compress faster
17:24
anuejn
yup
17:24
se6ast1an
very nice
17:24
se6ast1an
are the next steps already the step onto the FPGA?
17:24
anuejn
hm... not yet really
17:25
anuejn
the next step would probably be to find good parameters for the quantization and good codebooks for rle and huffman
17:25
se6ast1an
codebooks?
17:25
vup2
tables used for the huffman compression
17:26
vup2
we need to figure out which values occour often and which ones don't
17:26
se6ast1an
right
17:26
se6ast1an
so you need lots of image samples I assume?
17:26
vup2
yes
17:26
anuejn
yup that is what i started with earlier this day
17:26
se6ast1an
should we seek to collect more or are the ones on the server sufficient?
17:27
vup2
those are probably sufficient for now
17:27
se6ast1an
great
17:27
se6ast1an
you need samples with high image frequencies in particular I assume (pictures of fine details)?
17:27
anuejn
but if there is anything you definitely dont want to compress badly, send it now ;)
17:28
se6ast1an
fine pattern, textiles for example would be interesting then
17:28
vup2
yep
17:28
vup2
that would be good
17:28
se6ast1an
dont have a beta currently with me but will keep it in mind
17:28
se6ast1an
many thanks for the report! very exciting!
17:28
vup2
(we were using a image that had lower optical resolution that the image resolution and it showed :)
17:29
se6ast1an
please share before/after images if you collect them
17:29
vup2
sure
17:29
se6ast1an
would be interesting to see
17:30
se6ast1an
with acompanying quality "ratio"
17:30
se6ast1an
ok quick updates from me:
17:30
se6ast1an
the nextpcb order has advanced a few steps further
17:30
se6ast1an
and again at least one step back
17:31
se6ast1an
we had to upgrade the copper thickness as they quote the thcikness not per layer but per top and bottom layers combined
17:31
se6ast1an
the stackup looks good and we gave green light
17:31
se6ast1an
today we received again "production files" for verification
17:31
se6ast1an
which contained a completely different stackup again...
17:31
se6ast1an
maybe just a mix up with another project...
17:32
se6ast1an
vup2: tinkered with a wifi usb dongle driver on the beta until we concluded that this particular model is not going to work
17:32
se6ast1an
https://wiki.apertus.org/index.php/AXIOM_Beta/Manual#WiFi_Access_Point_Setup
17:33
se6ast1an
Edimax EW-7811UN Realtek RTL8188CUS
17:33
se6ast1an
I started diving into the AXIOM Remote code again
17:33
se6ast1an
and updated the documentation with lots of images and examples of the GUI so far: https://github.com/apertus-open-source-cinema/AXIOM-Remote/blob/dev/Firmware/README.md
17:33
se6ast1an
explaining the terms and classes better
17:34
se6ast1an
some fixes and improvements also pushed
17:34
se6ast1an
and BAndiT1983 merged the visualizer adjustments upstream
17:34
se6ast1an
so we now have the great looking blender render of the actual remote with illuminated button in the visualizer now
17:35
Bertl
yay!
17:35
se6ast1an
for the ESD foam we want to ship hardware in there is slight progress
17:35
se6ast1an
two chinese companies have agreed to ship free samples if we cover shipping
17:36
se6ast1an
those foams on paper would fit our requirements and are slightly and the other supplier significantly cheaper than what we researched before
17:36
se6ast1an
samples will tell if they are of useable quality
17:36
se6ast1an
$0.95$/dm³
17:36
se6ast1an
and $0.39$/dm³
17:37
se6ast1an
vs â¬2.18â¬/dm³ that we got quoted before in Europe
17:37
se6ast1an
shipping still needs to be calculated of course
17:37
se6ast1an
for any volume
17:38
se6ast1an
next steps now that Tele is back from vacation are to prepare anything left before we can start the assembly run
17:38
se6ast1an
ordering solder paste
17:38
se6ast1an
discussion who buys which components from the bom, etc.
17:39
se6ast1an
pick and place files, etc
17:39
se6ast1an
I guess thats it from my side
17:39
se6ast1an
Bertl: your closing words? :)
17:39
Bertl
thanks!
17:40
Bertl
well, besides the usual stuff (helping out here and there) I mostly spent my time with celebrating the new year, and after that, recovering from the celebrations :)
17:40
se6ast1an
very intense!
17:41
Bertl
yes, indeed, as every year :)
17:42
Bertl
so nothing really to report from my side this week
17:42
se6ast1an
right, what are the next steps for you?
17:43
Bertl
next steps are to get the SATA plugin working so we can do some tests there
17:43
Bertl
as well as further work on the power board firmware
17:44
se6ast1an
great!
17:44
Bertl
if there is some time left, I'd like to spend it on the HDMI digitizer
17:44
se6ast1an
anything else for todays meeting, or does anyone else have topics or things to discuss/share?
17:45
eppisai
left the channel
17:47
se6ast1an
eppisai fixed the checkbox icon on the remote just today for example: https://lab.apertus.org/T1193
17:47
se6ast1an
right then
17:47
eppisai
joined the channel
17:47
se6ast1an
many thanks everyone for attending/sharing
17:47
se6ast1an
lets see what 2021 will bring for the project
17:47
se6ast1an
MEETING CONCLUDED!
17:50
EmilJ
since it seems like the pynq setup is a pain to extract information from, since it's all linked together with petalinux/yocto config files, I'm delving deeper into the devicetree overlay I'm trying to load. There isn't really a good reason for it not to work
17:51
vup2
well to me it all seems like the original devicetree does not have symbols, or does not have a amba symbol
17:52
Bertl
the overlays need to match the device tree labels
17:52
vup2
yeah
17:56
Shohei
left the channel
18:03
EmilJ
the original devicetree does have `__symbol__` nodes but for some reason it acts like it doesn't
18:03
EmilJ
compiling without symbols and adding it stops the dmesg error from appearing
18:04
EmilJ
I think I should implement this
18:04
EmilJ
> Using the non-phandle based target method allows one to use a base DT which does not contain a __symbols__ node
18:05
EmilJ
right now I got `target = <&amba>;` and it resolves into 0xFFFFFFFF *exactly* the same as in the raspberry pi docs for dt: https://www.raspberrypi.org/documentation/configuration/device-tree.md
18:17
EmilJ
I arrived at the conclusion that simplefb is outright not shipped with the pynq image
18:17
EmilJ
judging from the existence of this repo https://github.com/ciniml/pynqz1fb
18:20
vup2
hmm that repo is something else
18:20
vup2
can you get the kernel config from /proc/config.gz ?
18:37
EmilJ
https://gitlab.com/-/snippets/2056747
18:46
BAndiT1983|away
changed nick to: BAndiT1983
19:06
Bertl
off for now ... bbl
19:06
Bertl
changed nick to: Bertl_oO
19:30
vup2
EmilJ: hmm yeah, the simplefb module seems to be disabled
19:30
vup2
can you get the linux source tree that pynq uses?
19:44
eppisai
left the channel
19:46
EmilJ
I got the correct release of the source tree and am figuring out how to bulid a single in-tree module
19:50
vup2
I think `make M=drivers/video/fbdev/simplefb` should work
19:59
EmilJ
that would require simplefb to have its own Makefile
19:59
EmilJ
I guess I'll find what Makefile builds it
19:59
EmilJ
and just build all in that Makefile
20:03
EmilJ
it builds the .o but doesn't link it into .ko, that's frustrating
20:16
EmilJ
oh, I set the config to y instead of m for module, that's on me :D
22:01
BAndiT1983
changed nick to: BAndiT1983|away