Monday, 2014-07-07

*** tpb has joined #timvideos00:00
*** rohitksingh has quit IRC01:30
shenkimithro: hello02:29
mithroandi-m: time plans?02:35
mithroshenki: heyo02:35
shenkimithro: are you going to submit a talk about the project to lca?02:35
mithroshenki: Probably not02:35
shenkimithro: nothing to talk about? or can't be bothered?02:36
mithronot enough time02:36
shenkiwould you like to do one together?02:36
JoelwIsn't it a 45 minute slot?02:37
shenkiJoelw: yes02:38
*** tija has joined #timvideos02:54
mithroshenki / tija: I wrote a triple buffer arbiter over the weekend - when I finally figured out the correct approach it turned out to be pretty easy03:05
tijamitrho: in VHDL?03:05
tijamirthro: If my understanding of triple buffer is correct then if we implement triple buffer, we will get frame rate of 40 frames per second03:07
shenkitija: you mentioned you got 30 fps on friday03:10
shenkitija: can you explain to us how you achived that?03:10
tijashenki: yes, I pipelined read and write. So now instead of two only one frame is dropped03:11
tijashenki: encoder fires as soon as 8 lines are written into DDR03:11
shenkitija: do you need a triple-buffer scheme? Or will two buffers be enough?03:17
tijashenki: triple buffer ensures that there is one frame to read from always. So it will give us a frame rate of 40 fps which is the max fps of encoder.03:19
shenkitija: okay. i do not understand what the third buffer provides us, can you explain why we need three and not just two?03:20
tijashenki: in double buffer, one buffer is written to and one buffer is read from. If reading is done and writing is not complete, encoder has to wait. But triple buffer has one buffer always ready to be read from.03:21
shenkiokay. but you only need to wait for 8 lines to be ready, you don't need to wait for the entire buffer to be done03:32
tijashenki: oh yeah, in that case double buffer should suffice03:41
shenkitija: yep, you read my mind03:47
shenkii'll be on in a sec03:47
shenkitija: ok, im in03:48
*** tija has quit IRC04:03
shenkimithro: hey, what kind of fitbit do you have?04:26
*** tija has joined #timvideos04:29
tijamithro: I have given you edit permission for this doc.04:31
tpbTitle: Debug Design - Google Docs (at
tijashenki: CAn you give me video samples of the kind of video we will be streaming using HDMI2USB. It will be useful to test the size of encoded frames.04:41
shenkitija: good idea04:41
shenkitija: there are two kinds of input, one would be presentation slides, the other is an image of a person standing at the front of a room04:42
shenkitija: so for the presentation, go to slideshare and find a slide deck that looks interesting04:43
shenkifor the person, that's a bit harder. we can't just take an existing conference recording, as they have been scaled to a very low resolution04:43
tijashenki: okay. So the presentation and the video of person will be muxed or input via the two HDMI ports?04:46
tijashenki: I mean we will use two hdmi input ports or one?04:47
shenkitija: that's a good question for mithro. the way i understood it, for now, we would have two different HDMI2USB devices - one for each input stream04:47
shenkijust he one, afaik04:47
tpb<> (at
shenkimmm hot pink diagrams04:49
mithroat the front, the two inputs could be used (IE a presenter laptop and a demo reel)04:53
andi-mmithro: i watched one of your talks, where you said you plan to create own pcb's some day04:55
mithroandi-m: yes, the first round of PCBs have been produced and are currently being assembled for bring up (arrived last week)04:56
andi-mis there some additional information available about this pcb's some where in the web?05:00
shenkiandi-m: mithro could answer that in detail, but in short the design is similar to the digilent dev board we have; DDR, Spartan 6 FPGA, Cypress FX2 USB, and HDMI i/o05:04
mithroandi-m: will hopefully have more information in the near future05:05
shenkiin addition, there are things like a power monitoring IC that weren't present on the dev board that we've added05:05
shenkimithro: hows the bringup going?05:05
mithroandi-m: there will be two versions, a budget one with cust reduced as much as possible, and a high end one which has extra functionality to make the device better for conference usage05:05
mithroshenki: they were waiting on parts to get through customs last I heard.05:06
mithrobasic should be roughly - 2 x HDMI input, 2 x HDMI output - Spartan 6, DDR Mem, Cypress FX2 USB05:07
mithroadvanced - 2 x DVI input, 2 x DVI output - Spartan 6 with GTP, DDR Mem, Cypress FX2, Atlys compatible VHDCI, GTP in/out on a DVI connector (possibility of 1080p and other high speed signals), PMIC for power monitoring + RS422 and RS232 serial ports + IR remote control, power from -48V<->+48V05:09
andi-mwe currently search a better alternative for the ethernet framegrabber we use at the moment05:21
mithroandi-m: do you have an Atlys board?05:25
andi-mno, not yet.05:25
mithroandi-m: I'd recommend getting one if you are interested in helping to improve this space05:25
mithrothe big difference we have is that we can actually improve debugability05:28
shenkiandi-m: what frame rates and resolutions are you grabbing?05:30
andi-mit allways depends... I prefer something that can handle 1080p6005:37
andi-mit doesn't matter if every frame gets transfered to the recording, but the signal shouldn't confuse the device05:38
mithro[email protected] is 3,981Mbits raw05:39
mithroandi-m: have you found any issues with things being marginal at that resolution?05:39
andi-mi don't understand this questions05:40
mithroandi-m: The higher the resolution, the better quality things like cables, senders and receivers have to be05:40
mithroandi-m: I've found that dropping to 720p increases the reliability without decreasing any noticable quality05:41
mithroandi-m: specially in things like conference venues where you don't control anything really05:41
andi-mthe problem is that the projector production resolution is often 1080p, and we can't controll that05:42
andi-mthe cables where dvi over fiber adapters, last year05:43
mithroandi-m: btw - are you C3 or some other group?05:43
andi-mbut of cause our smaller events, don't all have such requirements05:43
andi-mmithro: ccc, yes05:43
mithroDVI over fiber is not something we can afford :)05:44
shenkiit's just sand?05:44
andi-mthe building operator (cch hamburg) just included them in last year05:45
mithronor something I'd be willing to trust to running around a room which didn't have it permanently installed. Too easy to accidently crush.05:45
shenkii think there's a decent amount of development between where we're at and [email protected] fps05:45
shenkiwith recent improvements [email protected] fps should be doable, from a DSP point of view05:45
shenkiim not sure if we're there on the analog side though, are we mithro?05:46
andi-mthere are also some quite nice dvi over 2x cat6 network cabel adapters05:46
mithroandi-m: what type of price are you talking about? and at that level why not use SDI?05:46
andi-mfor our camera connections whe use sdi05:46
mithroha :)05:46
andi-mmithro: I don't know how expensive this cables are, the building operator just lent them for us05:47
mithroandi-m: I meant the converters05:47
andi-mstarting from 30 euros05:47
mithroandi-m: and you've found the reliable?05:48
andi-mit depends... ;-)05:48
andi-mthe problem is allways if you allow the speakers to bring there own notebooks...05:48
andi-mthat the resolutions changes and so on.05:49
mithroyes, we work in places where they always bring their own notebooks05:49
andi-mbut for connectiong a camera, or after a scaler or seamless switcher they should work quite reliable05:49
andi-mback to your fpga board:05:51
andi-mdo you redirect the edid data from the projector to the presenter notebook?05:52
mithroandi-m: we have complete control over the EDID05:52
mithroandi-m: we can present whatever EDID you want05:52
andi-malso at the projector side?05:52
shenkion both sides05:53
andi-mah, great.05:53
mithroandi-m: we can read the EDID data from the sink and then present whatever we want to the source05:53
andi-mwe currently can only controll the direction from the framegrabber to the presenter notebook05:53
andi-mbut this means you have to create indvidual edid files for nearly every projector05:54
mithroandi-m: When things are working correctly we present just 1 resolution [email protected] in DVI mode (or [email protected] if in HDMI mode)05:54
andi-mespacially you want correct aspect ration for 16:9 slides05:54
shenkimithro: do we do filtering, or just display a hardcoded edid?05:58
mithroandi-m: If the projector is projecting [email protected]  as 16:9 then there isn't much we can do about it because operating systems don't reliably support non-square pixels.05:58
mithroshenki: We originally did filtering, but it turns out to be a bad idea because it means your EDID data changes when you plug in a monitor (plus what do you do with two monitors with different modes)05:59
andi-mi used the 720p ony edid profile last time and got some complaints that some people were not able to activate display mirroring, as their internal display was not able to do 720...05:59
mithroandi-m: Macs?06:00
andi-mno, linux06:00
mithroandi-m: I've not seen that problem, but then I try and avoid Nvidia chipsets which have stupid restrictions06:01
andi-mso, i will go to downtown now. its now 8 a.m. in germany... ;-)06:01
andi-msee you later06:01
shenkisee ya06:02
shenkihrm, i wonder what the correct thing to do when the device doesn't do 720p is06:04
shenkithe sledgehammer would be to fallback to 4:306:04
shenkibut that's not awesome for people with wide slides06:05
shenkii wonder what the resolution of those devices was06:05
mithroshenki: there are a bunch of simple algorithms for scaling / fitting06:06
shenkimithro: mmm, but what do you pick?06:07
mithrothat is the hard part :)06:08
mithroblack-borders is the safest choice06:08
tpbTitle: Framegrabber Alternativen [CCC VOC] (at
mithroWe need a project to harden the EDID controller too06:11
shenkiin our system?06:11
shenkiperhaps we should have our EDID device fuzz the presenters laptop06:12
shenkionly if it doesn't crash are they allowed to continue presenting06:12
tpbTitle: Debug Design - Google Docs (at
shenkiAnonymous Wolf? Anonymous Tiger?06:15
shenkiwhat are the docs team smoking?06:15
mithroone of them will be tpb bot from above06:15
shenkihrm, so did we decide to put a micro into the system?06:16
mithroshenki: no, but I think we will shortly06:16
mithroshenki: I don't really know enough VHDL / Verilog to know how you'd implement modules "publishing" their debug info to a common management module06:19
mithrobut I think we'll need some type of event log / circular buffer06:23
shenkimithro: hrm06:41
shenkimithro: the way i'd design it would be to have a debug register in each module that contains some state information06:41
shenkimithro: and then we use the micro to read out the values of those registers and construct a log06:41
andi-myou can simply give the presenter laptop multiple resolution via edid06:51
andi-mit takes the first which fits06:52
andi-mbut i havn't understood how to generate this edid files yet. till now i use the files from ephiphans ftp server06:54
mithroandi-m: that doesn't work because the presenter will most definately end up choosing the wrong resolution :P06:54
mithroandi-m: edid.tv06:54
tpb<> (at
mithroandi-m: a previous year's GSoC project06:55
mithroandi-m: we haven't really been very good and utilising it06:55
tpbTitle: timvideos/edid_grabber_c · GitHub (at
shenkiwhat? that thing has a makefile that downloads curl06:57
andi-mif you like, you can add the files from
tpbTitle: Index of ./ (at
andi-mthe ephipan support also recommended for generating custom files06:58
tpbTitle: Advantiv™ EDID Editor | EngineerZone (at
mithroandi-m: can't seem to read that file...06:59
shenkimithro: does anyone maintain that tool?07:00
mithroshenki: nope07:01
mithroahh, I had to remove the table header / footer07:01
shenkimithro: okay. do we plan on using it?07:02
shenki{"error": "Submittion failed!"}07:02
mithroshenki: I'd like to use it07:02
shenki$ ./bin/edid-grabber --help07:02
shenki{"error": "Submittion failed!"}07:03
mithroshenki: but its been a low priority07:03
shenkii will fix it between kernel compiles :)07:03
andi-mthis is eg. a custom file generated by the support, with three 16:9 resolutions:
tpbTitle: Details - (at
*** tija has quit IRC07:07
andi-mwhere is detailed timings 2?07:08
mithroandi-m: good question :P07:09
shenkimithro: ok, the tool failed completely. it should be a python script on linux; no non-builtins needed and no need to force a user to build something07:12
shenkimithro: but pasting into the website worked07:12
mithroshenki: it's suppose to be a single binary you download/run07:13
shenkimithro: my dell monitor over VGA was prased07:13
shenkimithro: yeah, i think a python script is much better suited to that07:13
mithroEDID Grabber "Coming soon"07:13
shenkimithro: both from a maintainability and from a security pov07:13
mithroshenki: it doesn't work very well for Windows or Mac07:13
shenkimithro: but my LVDS panel in my x230 failed07:13
shenkimithro: right. so those people can sort out their own solution, but we don't need to make it fail on linux just for the sake of them07:14
shenkiim sure windows and mac people are used to running random binaries on their system07:15
mithroyeah :)07:16
shenkimithro: should we get the same results no matter the OS?07:16
mithroshenki: in theory07:17
mithroshenki: the travis-ci from edid-grabber-c was suppose to compile the binary and upload it to the website07:17
shenkiah okay. didn't quite get there?07:17
mithronot sure07:17
mithrobeen a while07:18
tpbTitle: edid_grabber_c/.travis.yml at master · timvideos/edid_grabber_c · GitHub (at
mithrothat looks like it should all work....07:19
shenkiyeah, i jsut saw the same thing07:19
tpbTitle: edid_grabber_c/.travis.yml at master · timvideos/edid_grabber_c · GitHub (at
shenkiat the end of it does this: curl -F "platform=$PLATFORM" -F "commit=$TRAVIS_COMMIT" -F "api_key=$EDID_GRABBER_UPLOAD_API_KEY" -F "[email protected]$RELEASE_FILE"
mithro$ chmod +x release/ && ./release/release.sh07:20
mithroThis script runs only for timvideos/edid_grabber_c07:20
mithrooh, wait that is for sewar's repo07:21
tpbTitle: Travis CI - Free Hosted Continuous Integration Platform for the Open Source Community (at
mithroUpload release EDID.tv07:21
mithroRelease posted successfully.07:21
mithroRelease uploaded to EDID.tv07:21
shenkiok. so we just need to add a download link?07:22
mithroshenki: it's probably just not been deployed07:25
shenkiwe need to install it on a beaglebone or raspeberry pi07:26
shenkiand just walk around unis, workplaces and department stores plugging things in07:26
mithroha :)07:27
mithrowe should just do a PIC which pulls the EDID information into the flash07:27
shenkiyeah, but then you need something that scrapes it and uploads to the net07:29
mithroshenki: but you can pull the power from the VGA/HDMI port07:30
shenkiinteresting. how much?07:30
mithroBetween 30mA and 50mA - in theory you could do up to 150mA but most things don't really supply that much07:31
*** tija has joined #timvideos07:50
andi-mmithro: how did you remove the header/footer from the ephiphan files?08:07
*** tija has quit IRC08:13
*** rohitksingh has joined #timvideos10:33
*** rohitksingh1 has joined #timvideos11:01
*** rohitksingh has quit IRC11:02
*** tija has joined #timvideos11:14
*** Niharika has joined #timvideos12:01
*** tija has quit IRC12:32
*** dhanush has joined #timvideos13:22
*** dhanush has left #timvideos13:24
*** rohitksingh has joined #timvideos14:05
*** rohitksingh1 has quit IRC14:07
mithrovim :P14:15
mithroandi-m: should just make it understand that format14:15
mithroayush3504: ping?14:15
ayush3504mithro: yes14:15
mithroayush3504: got PCBs for me to review?14:19
ayush3504mithro: sure take a look,
tpbTitle: Dropbox - Untitled3.png (at
ayush3504mithro: the pic and isolators have been moved to back side14:20
mithrowhy are there 2 isolators on the last one?14:24
mithroit's not one isolator per board?14:24
ayush3504mithro: a pair of isolator get's shared by each slot14:26
ayush3504mithro: the last two chips are 3 channel instead of 6 channel14:27
mithroayush3504: why?14:27
ayush3504mithro: you should take a look at the data isolation schematic14:28
ayush3504mithro: it's kind of hard to explain14:28
mithroI'm not seeing anything pushed recently?14:45
ayush3504I can see that I pushed an update 23 minutes ago14:46
mithroayush3504: the data isolation schematic is not very clear14:55
mithroayush3504: why haven't you named them properly?14:56
mithroayush3504: it's unclear if your if your layout even makes any sense15:01
mithrowith the current naming15:02
*** mparuszewski has joined #timvideos15:50
*** slomo has joined #timvideos17:48
ayush3504mithro: sorry for replying late. If you see the hierarchical block for data isolation block it's just 15 + 15 unidirectional lines in both directions. A & B postfix tell about the voltage levels. Knowing that A & B postfixes in pin names denote voltage levels and all lines are unidirectional it should make sense.17:49
*** rohitksingh has quit IRC19:49
*** Niharika has left #timvideos20:06
*** Niharika has quit IRC20:07
*** CarlFK has quit IRC20:38
*** slomo has quit IRC21:40
*** CarlFK has joined #timvideos22:10
*** ChanServ sets mode: +v CarlFK22:10

Generated by 2.12.1 by Marius Gedminas - find it at!