Current Server Time: 02:56 (Central Europe)

#apertus IRC Channel Logs

2013/06/18

Timezone: UTC


08:49
jucar1
joined the channel
08:51
jucar
left the channel
09:39
se6astian
joined the channel
11:38
jucar
joined the channel
11:39
jucar1
left the channel
12:06
jucar1
joined the channel
12:07
jucar
left the channel
13:26
jucar
joined the channel
13:28
jucar1
left the channel
14:15
philippejadin
joined the channel
14:16
philippejadin
hello, fellow hardcore irc users !
14:16
se6astian
;)
14:19
se6astian
maybe we can have a soft revival of IRC
14:19
Bertl
hey philippejadin!
14:19
se6astian
now that the NSA is managing our inboxes and logging our hangouts ;)
14:20
se6astian
May I introduce, Philippe from Belgium and Bertl from yesterday evening ;)
14:20
Bertl
I'm the one to blame for the sudden interest in irc :)
14:20
se6astian
yes, its all his fault
14:21
Bertl
I've heard that folks around the world are interested in axiom related development, and I thought it might be beneficial to virtually come together somewhere ...
14:22
Bertl
and as I hang out on IRC anyway ...
14:36
philippejadin
Good idea
14:36
philippejadin
hello !
14:37
philippejadin
it reminds me when I was on the worldforge.org project
14:37
philippejadin
I spent hours on irc
14:37
philippejadin
huge time sucker tool
14:37
philippejadin
but great fun
14:46
philippejadin
So, Bertl, what do you intend to work on ?
14:51
dmj_nova
joined the channel
14:51
dmj_nova
hello all
14:53
philippejadin
hello !
14:55
se6astian
hurray, more people, hi david!
14:56
Bertl
welcome dmj_nova!
14:59
dmj_nova
se6astian: I'm checking on the possibility to get the zedboard with academic pricing
14:59
se6astian
perfect
14:59
philippejadin1
joined the channel
14:59
philippejadin1
left the channel
14:59
se6astian
here it was no problem at all, they didnt even need any proof after I sent them a mail from my .edu address
15:00
dmj_nova
from diligent's site?
15:00
philippejadin
the .edu address is alreadya good proof :-)
15:01
se6astian
I found an austrian distributor
15:07
dmj_nova
Hmm...I suppose such a board could be used as an embedded computer and control system for nanomoco boards
15:08
dmj_nova
And therefore be "required" for my independent study
15:08
se6astian
definitely!
15:08
se6astian
:)
15:13
dmj_nova
zedboard can run ubuntu, right?
15:14
Bertl
philippejadin: well, I'll be investigating our FPGA options, and I'm currently trying to collect information about the components and adapter PCB
15:14
se6astian
Bertl, this could be usefull later: http://blog.aeste.my/2013/05/09/zedboard-on-ubuntu-12-04-lts/
15:14
philippejadin
dmj_nova: yes
15:14
philippejadin
but you might prefer a lighter distribution
15:15
dmj_nova
philippejadin: yep, then I can definitely use it for my independent study
15:15
philippejadin
(or realtime)
15:15
Bertl
yeah, while ubuntu is probably nice for desktop, it's overkill for embedded systems
15:15
dmj_nova
philippejadin: mostly care since it avoids having to port tons of dependencies
15:15
Bertl
(which can be easily seen on raspberry pi :)
15:16
Bertl
but probably a good starting point anyway
15:16
dmj_nova
and if I want to do fun things with realtime collaboration
15:16
philippejadin
I hav eno experience with realtime linux, but for our use it might be interesting
15:16
dmj_nova
*cough*dmedia integration*cough*
15:17
Bertl
philippejadin: there is hard realtime and soft realtime ... basic realtime support is already present in almost all recent linux kernels
15:18
Bertl
i.e. you can utilize the realtime sheduler for example
15:18
philippejadin
the idea anyway is to do the hard processing on the fpga and use linux for not so realtime menu overlay. I hav eno idea when we'll have to write files somewhere how we'll do that :-)
15:19
dmj_nova
I expect that we'll want to do all the stuff that needs to be hard realtime in the FPGA
15:19
Bertl
most likely, at least to the extend possible
15:20
Bertl
philippejadin: you mean the actual data from the sensor or what?
15:20
dmj_nova
the FPGA could likely output frames to memory that's shared with Linux and the Linux can read and store it as appropriate
15:21
Bertl
ala dual ported ram?
15:21
dmj_nova
assuming our hardware can do that
15:22
Bertl
we'll see, no idea atm ...
15:22
philippejadin
Wild guess : image processing -> fpga / writing images to disk -> linux (as long as it can be done fast enough)
15:22
dmj_nova
What is everyone's FPGA experience?
15:23
Bertl
I've started recently with FPGAs mainly because I got my hands on a bunch of really cheap lattice chips. I've the theory behind it and I've already managed to write a bunch of VHDL programs to play with
15:24
dmj_nova
I programmed the game Spacewar! for an FPGA in a class, and I designed a simulated 16-bit CPU with L2 cache and basic pipelining in VHDL.
15:24
Bertl
I have a lot of experience with embedded systems in general and I use Linux for almost everything, including design and development
15:25
Bertl
dmj_nova: nice!
15:28
dmj_nova
I also use Linux pretty much always. I'm part of the team making Novacut.
15:30
Bertl
I probably should mention that I have no experience whatsoever with making movies or filming in general, but I've done a lot of image processing (technical) and know the mathematical models
15:30
dmj_nova
I have some experience with (very) indie filmmaking
15:31
dmj_nova
and quite a few contacts with filmmakers
15:31
dmj_nova
so we should be able to get feedback when the time comes
15:31
dmj_nova
I do have some gstreamer experience
15:33
dmj_nova
Some of this may be a bit obvious but a few things to keep in mind since we're a distributed team on a hardware project
15:34
dmj_nova
Keeping things modular with well defined interfaces should reduce the amount of breakage/stepping on other people's work
15:34
dmj_nova
Bertl: have you ever worked with other people on your projects?
15:35
philippejadi
has to go, see you later guys, I'm very happy to see you on board !
15:36
Bertl
dmj_nova: yes, I'm maintaining an open source project since 2001
15:36
dmj_nova
Bertl: like a hardware kind?
15:36
philippejadin
left the channel
15:36
Bertl
software (Linux-VServer)
15:37
dmj_nova
okay, hardware (from my experience) tends to be a bit harder to collaborate
15:37
Bertl
quite understandable
15:37
dmj_nova
So modularity and not changing things at the interfaces is a pretty big deal
15:38
Bertl
very similar to the linux kernel interface, actually :)
15:38
dmj_nova
since (depending on your tools) merging can be a major problem
15:40
dmj_nova
I was also talking to se6astian about the possibility of having an automated test setup to verify that our commits don't break functionality
15:41
Bertl
automated as in physical hardware running whatever was commited?
15:41
dmj_nova
yes
15:41
Bertl
sounds good
15:41
dmj_nova
and then reporting back any issues
15:41
Bertl
I'm all for automation
15:41
Bertl
I hate doing things twice, I'd rather script it :)
15:41
dmj_nova
Novacut has a test suite that runs on launchpad for every build
15:42
dmj_nova
it's helped us catch problems and keep trunk releaseable
15:43
Bertl
yep, we have made similar experiences with automated kernel building and testing
15:43
Bertl
so I'm all on-board with that
15:44
Bertl
especially as I prefer to have an automated build process to get from source to binary
16:53
ApertusWeb0
joined the channel
17:02
ApertusWeb0
left the channel
18:04
dmj_nova
So it looks like we will have to be very space conscious with the zedboard
18:05
dmj_nova
since it only has 85k logic cells
18:07
dmj_nova
se6astian: The actual device will be incorporating the Zynq SoC on a custom PCB, yes?
18:11
se6astian
yes
18:13
se6astian
do you think our logic will require more gates than we have to our disposal?
18:17
se6astian
Elphel camera just had 19K logic cells
18:20
dmj_nova
elphel was only HDish, right?
18:20
dmj_nova
My guess is that the zedboard should do for development for the time being.
18:21
dmj_nova
However, additional logic cells will be highly useful for building in more advanced functionality
18:22
dmj_nova
as in, having a certain amount of surplus logic cells makes the system more extensible
18:24
se6astian
what do you mean with HDish?
18:25
se6astian
the FPGA tasks inside the Elphel camera are pretty broad, they even do jpeg compression in there
18:25
se6astian
and if I remember correctly they still had plenty of cells left
18:25
se6astian
and we have 4 times more available already
18:50
Bertl
IMHO it's always good to develop on lower end hardware ... it keeps designs efficient and performant :)
19:11
dmj_nova
Bertl: Yeah, I think the zedboard will do fine for development
19:38
se6astian
all parts routed!
19:38
se6astian
https://github.com/apertus-open-source-cinema/alpha-hardware/tree/master/SFE-PCB
19:48
dmj_nova
se6astian: nice
20:06
se6astian
I had the idea that we could try to cut a hole into the image sensor socket and PCB in the middle of the image sensor
20:06
se6astian
and try to attach a cooling element to the sensor
20:07
se6astian
from behind
20:15
dmj_nova
se6astian: hmm
20:15
dmj_nova
any idea how easy that would be to fabricate?
20:15
se6astian
cutting 2 holes?
20:18
dmj_nova
So we would attach a metal plate to the back of the sensor and feed 2 heat pipes into it from behind?
20:21
Bertl
does it require additional cooling?
20:22
dmj_nova
I know that other cameras require careful consideration of cooling needs
20:22
se6astian
not "require" - but noise increases with temperature
20:22
dmj_nova
hmm...We might want to put quite some effort into cooling then
20:23
dmj_nova
especially since our "competitors" are putting forth some very impressive devices noise-wise
20:24
se6astian
well its just a prototype
20:24
se6astian
but my idea was that some tests couldnt hurt
20:24
dmj_nova
well, the prototype noise is *less* important
20:25
se6astian
since the SFE goes directly on top of the zedboard we dont have much space between them
20:25
dmj_nova
but testing is always good
20:25
dmj_nova
and the final model should be designed to be usable in hot conditions for extended periods
20:30
Bertl
we do not even know how the socket looks like (yet), so it might or might not be possible to do direct cooling
20:31
Bertl
in general, I think it would be a lot better to use extra thick copper for the final board and vias to cool from behind
20:32
Bertl
(if the socket allows for that)
20:33
Bertl
alternatively, to reduce dark current and silicon noise, we could use a peltier element
20:33
dmj_nova
so the board becomes effectively a heat spreader?
20:33
Bertl
the area below the sensor, yes
20:33
dmj_nova
yeah, I was thinking peltier might be a good (optional) element to use
20:33
dmj_nova
at least when lots of power is available
20:34
dmj_nova
but you still have to dissipate the heat outside the box one way or another
20:34
Bertl
the question is how much heat the sensor itself will produce
20:35
Bertl
if there isn't much heat to dissipate, a low power peltier element could keep it at very low temperatures
20:36
Bertl
but I think, atm, we do not have enough information available
20:36
dmj_nova
hmm...does the sensor have a temperature sensor?
20:37
Bertl
yes
20:37
Bertl
16bit, digital
20:37
dmj_nova
that should help in tuning things if there is a peltier or other active cooling system
20:37
Bertl
yup
20:38
dmj_nova
and compensate somewhat for lower battery performance in cold weather
20:38
dmj_nova
not having to spend as much power on cooling
20:38
Bertl
the CMV12000 has a total consumption of 1900mW
20:38
dmj_nova
hmm
20:38
Bertl
so roughly 2W max
20:39
dmj_nova
That's not a huge amount
20:39
dmj_nova
not trivial though
20:39
Bertl
not sure how much is dissipated as heat though
20:39
dmj_nova
all of it
20:40
dmj_nova
seeing as it's not a light bulb
20:40
dmj_nova
or a radio transmitter
20:40
dmj_nova
or a speaker
20:41
Bertl
yeah, but it drives outputs
20:42
dmj_nova
true, though I expect that's minimal compared to what it consumes for its own operations
20:43
dmj_nova
I could be wrong though
20:43
dmj_nova
Bertl: do you have a zedboard yet?
20:46
Bertl
yes, I got one, I'm currently preparing my development environment
20:47
dmj_nova
cool
20:47
dmj_nova
btw, is the software windows-only or does it support better systems?
20:47
Bertl
I'm on linux, so it will be better than windows :)
20:48
dmj_nova
cool, good to know it works on linux :)
20:48
dmj_nova
much less annoying
20:48
Bertl
definitely ... I avoid windows and microsoft products completely
20:48
dmj_nova
well, as long as it doesn't suck as much as mentor graphics
21:01
dmj_nova
I'm pretty sure that my windows box is both 6 months out of date and very secure against online hacks
21:04
dmj_nova
but that's just cause it's not been turned on in that long :P
22:55
se6astian
left the channel