Current Server Time: 19:35 (Central Europe)

#apertus IRC Channel Logs

2015/11/02

Timezone: UTC


01:49
fsteinel
left the channel
01:49
fsteinel_
joined the channel
03:41
intracube
changed nick to: intracube|away
06:47
Bertl_zZ
changed nick to: Bertl
06:47
Bertl
morning folks!
07:50
se6astian|away
changed nick to: se6astian
07:55
se6astian
good morning
07:57
Bertl
morning se6astian!
08:48
sebix
joined the channel
08:55
sebix
changed nick to: niemand
09:41
Bertl
off for now ... bbl
09:41
Bertl
changed nick to: Bertl_oO
09:44
tezburma
joined the channel
10:27
LordVan
joined the channel
11:32
plaes
left the channel
11:32
plaes
joined the channel
11:33
plaes
left the channel
11:33
plaes
joined the channel
11:35
arpu
joined the channel
12:59
aombk3
joined the channel
13:02
aombk2
left the channel
13:14
arpu
left the channel
14:01
Guest35178
left the channel
14:30
tezburma
left the channel
14:43
tezburma
joined the channel
14:53
fsteinel_
changed nick to: fsteinel
15:00
LordVan
left the channel
15:17
cbohnens
changed nick to: cbohnens|away
15:35
troy_s
left the channel
15:42
troy_s
joined the channel
15:58
se6astian
changed nick to: se6astian|away
16:01
dmjnova
left the channel
19:14
John_K
Beta PCBs arrived from OSHPark!
19:19
Bertl_oO
congrats!
19:20
Bertl_oO
maybe document the entire build process
19:20
Bertl_oO
changed nick to: Bertl
19:20
Bertl
so that we can later make an article from/about it?
19:22
Bertl
(for others to get inspired)
19:23
John_K
Yeah, I'm trying to put together a BOM atm
19:23
John_K
Difficult even with the Google sheets
19:24
Bertl
let me know if you need anything (i.e. if something is unclear)
19:24
John_K
Some parts are out of stock, etc... At least I ordered the ZIFs in advance
19:25
John_K
Will do, hoping to have them assembled later this week or early next week
19:25
Bertl
usually you can use similar parts (for passive components)
19:25
Bertl
and the active ones should be available from various distributors
19:26
John_K
True, I saw a few werenobly available from Farnell but not element14 America
19:26
John_K
Were only*
19:27
danieel
we have the opposite issue :) some are from america only and not at Farnell :))
19:28
Bertl
John_K: also note that not all parts need to be populated to get a working beta
19:29
Bertl
specifically on the power board and the sensor frontend
19:29
John_K
Other than the DNP passive?
19:30
Bertl
yes, for example you can drop the entire voltage check blocks from the sensor frontend (for now)
19:30
Bertl
as you are adjusting the power board manually to the desired voltages
19:30
Bertl
and there you only need to populate the circuits for the actual voltage rails in use
19:31
Bertl
it might bite you lateron when the boards get updated, but then it should be easy to add the missing parts and/or do a new board
19:32
Bertl
http://files.apertus.org/AXIOM-Beta/Beta%20Sensor%20Front%20End1.jpg
19:32
Bertl
here for example you can see the "minimal" configuration
19:33
Bertl
well, you could leave out the eeprom at the top left as well
19:36
Bertl
on the power board you can drop the FTDI circuit if you don't want the JTAG interface via USB (and use the one on the microzed)
19:36
Bertl
you can also leave out the micro SD circuit (if you don't need it)
19:37
Bertl
and here you can see the minimal config from the regulator side:
19:37
Bertl
https://wiki.apertus.org/images/f/f4/BetaPowerBoard_0.18_TOP.jpg
19:37
Bertl
i.e. WW/SW/SE/EE as well as MCN/MCS are not required for basic operation
19:40
danieel
Bertl: once i wanted to do a html/svg visualization of boards.. this required/optional/dnp could be shown nicely with such a gui..
19:41
Bertl
kind of on/off layer composition or so?
19:41
danieel
yes, and as components are tagged they can be filtered by that
19:42
danieel
handful for hand assembly (highlight all of same value)
19:42
danieel
does such tool exist?
19:45
Bertl
not that I know of
19:50
danieel
for my next birthday (april 1) we can rewrite the debayering article: http://stargazerslounge.com/topic/166334-debayering-a-dslrs-bayer-matrix/ :))
19:56
John_K
http://imgur.com/gallery/gGWlr
19:57
John_K
Purple PCB prob
19:57
John_K
Porn
19:57
John_K
Oops
19:57
Bertl
almost
19:58
John_K
Http://imgur.com/gGWlr
19:58
John_K
I'll do this from not mobile
20:03
John_K
Third times a charm http://m.imgur.com/gGWIr
20:03
intracube|away
changed nick to: intracube
20:05
John_K
Bertl: thanks for the pointer to minimal configs, may go with that for now
20:10
Bertl
@images: nice :)
20:11
Bertl
on the main board, you can leave out the shield connectors for now (they can easily be added lateron) and the piezo
20:17
irieger
danieel: That debayering looks nice!
20:17
irieger
Would like to have a spare cam and have it "debayered" :D
20:18
irieger
Makes demosaicing so much easier thought :D
20:19
danieel
definitely hardware based implementation of the algorithm :)
20:26
irieger
Yeah. Btw.
20:27
irieger
Bertl: I head a great idea. Would be nice to be able to just output raw 12 bit data via HDMI and would be recordable with existing hardware. My idea was: Pack raw data into 1920x1080x3colorsx12bit
20:28
danieel
you have 4 in bayer
20:28
Bertl
and 12bit requires deep color
20:28
irieger
should be possible to even get a 3672x3072 or so for 4K anamorphic recording with around 33 fps into a 1080p60 stream
20:29
irieger
Bertl: is that a problem with the HDMI controlers?
20:29
irieger
danieel: you mean 4 what?
20:30
irieger
just did a small calculation how this would fit into 1920x1080 while on the road cause I liked the idea and there is already hardware able to capture 1080p50 / 1080p60 at 12 bit 444
20:30
Bertl
it doubles the bandwidth or halves the frame rate
20:31
irieger
had a chance to shoot some test material with the lovely Zeiss Master Anamorphics on Friday and I just fell in love with anamorphic lenses btw ;-)
20:31
Bertl
(deep color that is)
20:31
irieger
So 60p is only possible with 8bit and 10/12 will require to go back to 30p?
20:32
Bertl
with the current direct from FPGA HDMI out, yes
20:32
Bertl
also note that most recorders cannot capture 12bit 444
20:33
irieger
I know. But there are some that can (Convergent Design). And some capture cards/thunderbolt capture modules can do so
20:33
Bertl
in raw or with some lossy codec?
20:33
irieger
Why must everything be so difficult ... Sounded so nice
20:34
irieger
Bertl: in 12bit 444 DPX so uncompressed "raw"
20:34
irieger
implementing this idea would mean the image would look like crap of course but with a small software it could easily be converted to cinemaDNG or so
20:36
Bertl
well, there are still some options left, I have some ideas which we will test in the near future
20:38
irieger
Top secret currently or can you give a hint? ;-)
20:39
irieger
btw. I recently played with color matrix creation with our toolset which I currently test and look for optimization possiblities. Here is a test image: http://content.irieger.net/bm_colorchecker_srgb_tun.jpg
20:40
irieger
looks quite good and was just the first step. Ignore some glitches in this image. I intentionally used a simple daybering algorithm and ignored some stuff like usable pixel area maybe ...
20:40
irieger
And on the linux laptop I'm currently on it looks quite worse than on my Macbook I have to say
20:41
danieel
irieger: when you get the table right, any real object will be strange colored :)
20:41
irieger
;-)
20:42
Bertl
top secret of course ... :)
20:43
irieger
Well, I will play with different optimizations. Our tools allow to throw in some parameter files to for example focus the optimization on skin tones etc.
20:44
Bertl
what we plan to do is to distribute the raw data in a way that it won't get mutilated too much by recording with cheap recorders
20:44
Bertl
(i.e. with compression in place)
20:44
irieger
Now I'm listening?!
20:46
Bertl
for example, assuming that there is 422 recording at 60FPS, we can break down the raw image (at 30FPS) into two images RG and GB with a proper distribution
20:47
Bertl
but we need to do some tests to figure out the optimal layout there
20:48
irieger
Bertl: well, most really cheap recorders can't do 60fps either ;-)
20:49
Bertl
indeed, but it is probably a good middle ground
20:50
irieger
Definitely.
20:50
Bertl
i.e. 60p pro-res (or similar) is more common than 30p 12bit 444 raw
20:50
irieger
Yes. Just fear what prores will do to the raw image. And it will lose 2 bits
20:51
Bertl
but note that 30p 12bit 444 at 1080p should be no problem with the current HDMI interfaces
20:52
Bertl
so if you have two recorders which can do that, then you might already have a raw recording solution with (almost) full size
20:52
John_K
How much over gig-e transmit rates are we with 1080p 60fps 12-bit?
20:52
irieger
Of course, that should work.
20:52
John_K
Seems like there should be plenty of space in the FPGA for some lossless compression
20:53
irieger
John_K: Full sensor is around 432 MByte/s I'd say
20:54
irieger
4096*3072*12/8 bytes per image
20:55
John_K
Right but for 1080 should be possible with minimal compression
20:55
John_K
1080p60 raw should be ~185MiB/s
20:55
John_K
So should be attainable with compression
20:57
John_K
Looks like 10GBE could be a possible expansion too as long as the PCB routing is compatible
20:58
irieger
John_K: The problem is: You have to do the bad stuff DSLRs do to get 1080p RAW or just take the inner 1080p frame of the sensor which is also kind of bad.
20:59
John_K
What would the problem be as long as you capture 16x pixels all around no?
21:00
John_K
Sorry ~16px extra on all sides
21:00
irieger
That the field of view of the lenses is totaly different then
21:01
irieger
Sorry, have to go afk, will look back later. (Note to myself: do a proper afk state like bertl/se6astian ...)
21:01
danieel
you can /nick
21:02
John_K
Hrm, true
21:03
John_K
Could be some interesting things you could do with the RAW though
21:03
John_K
Might be fun to play with at some point
21:09
Bertl
that's the beauty of the Axiom ... you can easily try out new ideas
21:09
niemand
left the channel
21:23
tezburma
left the channel
21:23
se6astian|away
changed nick to: se6astian
22:14
se6astian
time for bed
22:14
se6astian
gnight
22:14
se6astian
changed nick to: se6astian|away