Current Server Time: 17:50 (Central Europe)

#apertus IRC Channel Logs

2020/10/05

Timezone: UTC


00:25
intrac
left the channel
00:26
intrac
joined the channel
02:31
intrac
left the channel
02:31
intrac
joined the channel
03:16
Bertl_oO
off to bed now ... have a good one everyone!
03:16
Bertl_oO
changed nick to: Bertl_zZ
06:12
BAndiT1983|away
changed nick to: BAndiT1983
08:49
anuejn
good morning :)
10:08
Bertl_zZ
changed nick to: Bertl
10:08
Bertl
morning folks!
10:17
se6ast1an
good day
14:00
Bertl
off for now ... bbftm
14:00
Bertl
changed nick to: Bertl_oO
15:58
Bertl_oO
changed nick to: Bertl
16:00
se6ast1an
MEETING TIME!
16:00
se6ast1an
who is here?
16:03
Bertl
is here ...
16:03
se6ast1an
just us two it seems
16:04
Bertl
well, we probably can save the time then ...
16:05
se6ast1an
true
16:05
se6ast1an
I will publish the gsoc article then now
16:07
se6ast1an
https://www.apertus.org/google-summer-of-code-wrap-2020-article-october-2020
16:09
Bertl
nice
16:10
se6ast1an
BAndiT1983 you can share some news right?
16:10
BAndiT1983
a bit, was not that present lately because of my regular job
16:10
BAndiT1983
but this weekend i've finally sat down and experimented with the PCB paneliser
16:11
BAndiT1983
we had a problem with rotated data there, which was resulting in complains from the PCB fab
16:11
BAndiT1983
after some discussion with Bertl and se6ast1an the solution was found, so the gerber data was rotated by using gerbv command line exporter, then assembled and adjusted in the python scripts
16:12
BAndiT1983
final obstacle was the drill layer, but after some experiments with decimal settings it finally worked
16:12
se6ast1an
beautifully I must say :)
16:12
se6ast1an
panels have been sent to pcb fabs already for quotes
16:12
BAndiT1983
will align with se6ast1an on further tasks, as i'm a bit out of the loop
16:13
BAndiT1983
don't have a preview right now, otherwise i would show some nice images of 2d lines ;)
16:13
BAndiT1983
maybe visualiser, firmware or bootloader for axiom remote as next
16:13
BAndiT1983
that would be it
16:14
se6ast1an
great, many thanks
16:14
se6ast1an
now I feel motivated to give a brief insight into the pcb prober as well
16:15
se6ast1an
to identify the mechanical inaccuracies bertl created a test pattern
16:15
se6ast1an
https://github.com/apertus-open-source-cinema/pcb-prober/blob/master/calibration/testpattern_v2_150x150.pdf
16:15
se6ast1an
that I printed and put on the bed of the pcb prober
16:16
se6ast1an
with the attached microscope camera I then ran an automated optical center detection of around 4000 points and did each measurement 4 times
16:16
BAndiT1983
not bad, but i experience the effect where the small dots start to turn gray individually, don't remember the name at the moment
16:16
se6ast1an
the results were visualized by Bertl
16:16
se6ast1an
https://github.com/apertus-open-source-cinema/pcb-prober/blob/master/calibration/result%20visualization%2001.svg
16:16
se6ast1an
https://github.com/apertus-open-source-cinema/pcb-prober/blob/master/calibration/result%20visualization%2002.svg
16:17
se6ast1an
you can clearly see the errors have a pattern
16:17
BAndiT1983
github is crashing there, maybe too big, have to download
16:18
se6ast1an
yeah lots of lines
16:18
BAndiT1983
are the peaks just some random errors?
16:18
se6ast1an
we suspect its due to the wheels on the ender being slightly egg shapped
16:19
se6ast1an
possibly because the tension on them is too high or too low
16:19
Bertl
note that the 'error' is magnified by a factor of 10 (or probably more)
16:19
BAndiT1983
usually the belt tension is not that important, at least not to the extent people always think and propagate
16:20
BAndiT1983
is the probe head too light and vibrates too much?
16:20
se6ast1an
I also tweaked the belt tension but what I meant was the pressure on the wheels
16:21
Bertl
BAndiT1983: for vibration the pattern too much depends on the position
16:21
se6ast1an
anyway, we figured we wanna try probing now and see where we actually end up
16:21
BAndiT1983
ok, strange, haven't looked at the ender construction yet
16:22
se6ast1an
the probing routine has been added today, there are some things still to be tweaked or improved
16:23
se6ast1an
at least normal sized pads (0402+) should be hit properly
16:23
se6ast1an
for the smaller chip pads we will see what the best approach is
16:23
se6ast1an
maybe optical pad detection, bertl also suggested just trying to probe several points around the target
16:23
se6ast1an
shotgun approach
16:23
Bertl
I'm optimistic that we are good for 99% of all pads we have :)
16:23
se6ast1an
the idea is that the measurement will tell us if we hit immediately
16:24
BAndiT1983
optical approach is interesting
16:24
se6ast1an
the problem is that the movement between centering camera and moving the needle there will also produce an error
16:24
BAndiT1983
but what if you hit the pad next to it?
16:24
se6ast1an
but potentiall smaller error
16:25
BAndiT1983
^it was about "shotgun approach"
16:25
se6ast1an
yeah, I am curious if the measurement will show that
16:26
BAndiT1983
is the other probe on the GND?
16:26
se6ast1an
I mean if it will allow us to identify the neighbour pad
16:26
se6ast1an
there is only one probe
16:26
se6ast1an
but the GND plane is grounded yes
16:26
BAndiT1983
ok, am curious how we will recognize individual pads, if it hits the wrong one
16:27
BAndiT1983
resistance of the trace?
16:27
se6ast1an
bertl designed the probe to not just measure connected/not connected but resistance, capacitance, ...
16:27
se6ast1an
Bertl: can you elaborate
16:28
Bertl
I think I already did a few months ago ...
16:28
Bertl
basically we do a resistive as well as capacitive measurement
16:28
Bertl
this still needs to be fine tuned but for that we need real world measurements which haven't been available yet
16:29
Bertl
regarding 'hitting the wrong pad', if we are that far off, we have a bigger problem
16:29
BAndiT1983
i remember something like that, but had already quite many other things in-between, so it got lost somewhere in the abyss of the mind
16:30
Bertl
chances are, if we miss a pad, we a right on the edge of the pad
16:30
Bertl
this means that e.g. a circular motion around this point will either hit the pad or miss it as well
16:30
Bertl
(a sufficiently small circular motion that is :)
16:31
Bertl
so the question there becomes: can we tell whether we did hit a pad or not?
16:31
Bertl
and I'm somewhat confident that we can tell the difference there
16:31
BAndiT1983
just a theoretical question, as i wasn'T aware of wheels on the ender (my prusa doesn't have such stuff), would it work better with "belt only" printer base?
16:32
Bertl
for sure a more stable construction would work better, we just started with the ender 2 because it was available and cheap
16:32
se6ast1an
the ender 3 has a linear rail modification kit
16:32
se6ast1an
not too expensive
16:33
se6ast1an
but means we need to build a new unit from scratch
16:33
Bertl
so for now, let's see how well this one does ...
16:33
se6ast1an
exactly
16:34
se6ast1an
so much for the prober
16:34
se6ast1an
Bertl: do you have any news to share?
16:35
Bertl
guess so ...
16:35
Bertl
besides helping out with the prober and other stuff, I worked a little more on the cmv_snap* side
16:36
Bertl
and I did a SIMD (NEON) implementation for the line conversion
16:36
Bertl
this can now pull the image data from memory in about 140ms
16:37
Bertl
still if we run the cmv_snap over ssh, we get about 1.6-1.8 seconds
16:38
Bertl
but for local storage this has some advantages and if we find a clever way to eliminate the ssh overhead, we can easily transfer images in less than 500ms
16:38
se6ast1an
did you also benchmark writing to micro sd card ?
16:38
se6ast1an
ah, great
16:38
BAndiT1983
iss SSH mandatory or can we use (R)UDP?
16:39
Bertl
we can use whatever we like I guess?
16:40
Bertl
I also updated the storage system at the hub, added two disk drives there and we are currently spreading the data over the drives
16:41
Bertl
and I manually probed the new power board PCBs so that we can start assembly for testing
16:41
se6ast1an
two boards actually no?
16:41
Bertl
two boards were tested, yes, the third one is in your prober :)
16:42
se6ast1an
correct
16:42
se6ast1an
regarding snap
16:42
se6ast1an
could you test this script with it:
16:42
se6ast1an
https://github.com/apertus-open-source-cinema/axiom-firmware/blob/master/software/scripts/axiom_picture_snap.sh
16:42
se6ast1an
until line 21
16:42
se6ast1an
line 15 & 16 are a work around because snap hung with every econd run
16:42
se6ast1an
*second
16:43
Bertl
so what you want tested is a snap to SD card, right?
16:43
se6ast1an
yes
16:43
se6ast1an
the script doesnt do much
16:43
se6ast1an
disable hdmi
16:43
Bertl
i.e. we do not need the other stuff there
16:44
BAndiT1983
regarding UDP or whatever can be used, can't we just check wit socat if it will work? just as a proof of concept -> https://knplabs.com/en/blog/share-files-on-the-local-network-with-socat
16:44
se6ast1an
capture image to sd card with unique name derived from current date
16:44
se6ast1an
turn hdmi back on
16:44
se6ast1an
the rest is dng conversion, and raw development that doesnt matter for the test
16:44
Bertl
the problem I see here is that there is no sync
16:44
Bertl
so most of the SD writes will end up in cache instead
16:45
se6ast1an
good idea, added
16:45
Bertl
but I can design a test which actually writes to SD directly
16:45
Bertl
(because caching and then sync is a baad idea too :)
16:45
se6ast1an
do your magic :)
16:45
se6ast1an
and update the script then please
16:46
Bertl
I plan to spend some more time with the memory read/reorder/write anyway, so no problem there
16:47
se6ast1an
great
16:47
se6ast1an
vup messaged me: no news
16:47
se6ast1an
anuejn is currently occupied with a lot of other things
16:47
se6ast1an
he will let us know when his life has more structure again
16:49
bluez_[m]1
Hey guys! I can share something...
16:49
se6ast1an
great, please do!
16:50
bluez_[m]1
tx!.. so I tested my driver with the latest kernel in-tree (had some help from vup there), and it seems to work nicely
16:50
Bertl
yay!
16:51
bluez_[m]1
only I realized as soon as the driver is bound with the devicetree and probes...it automatically does some jtag stuff (queries status)
16:51
bluez_[m]1
so it screws up the state as the jtag is not turned on by then (which is done by a script now)
16:52
bluez_[m]1
reloading the driver works fine then
16:52
Bertl
so we should focus on that and fix this with a little help from the pic side
16:52
bluez_[m]1
but i think it would be better if we turn jtag on from within the driver...
16:52
bluez_[m]1
https://drive.google.com/file/d/109YyFnUghoU8Ajr8Et2xjYYEmsQQNuwB/view?usp=sharing
16:53
bluez_[m]1
Bertl: yep... have a look at this patch.. looks good?
16:54
Bertl
well, kind of ... let's discuss this in the next few days, I should have some time
16:55
bluez_[m]1
other than this... i have yet to add some kernel documentation, update the newer devm version of some functions, and some other clean up stuff... I guess it will be somewhat ready for mainline then...!
16:55
bluez_[m]1
Bertl: ah, alright alright
16:56
bluez_[m]1
we can make this a more generic command rather than just for jtag as you mentioned...
16:56
Bertl
yes, that's what we should aim at
16:56
bluez_[m]1
anyway... i guess thats it from me now..
16:56
Bertl
like a command register to do 'stuff'
16:57
bluez_[m]1
Bertl: yeah yeah.. we can have a switch inside that function basically that does different stuff based on the argument..
16:58
bluez_[m]1
but yea we should discuss that as you are familiar with what other 'stuff' we can include here..
17:06
se6ast1an
ok, anyone else here otherwise we can conclude this weeks meeting
17:09
se6ast1an
MEETING CONCLUDED
17:09
se6ast1an
many thanks everyone!
19:34
comradekingu
left the channel
19:36
comradekingu
joined the channel
19:54
Bertl
off to bed now ... have a good one everyone!
19:54
Bertl
changed nick to: Bertl_zZ
20:15
RexOrCine
joined the channel
20:18
aombk2
joined the channel
20:25
matthew[m]3
left the channel
20:25
aombk
left the channel
20:28
matthew[m]3
joined the channel
21:41
BAndiT1983
changed nick to: BAndiT1983|away
22:27
lexano
left the channel
22:52
lexano
joined the channel