Sunday, 2016-01-24

*** tpb has joined #timvideos00:00
cr1901_modernsorry, mithro. sb0 needs the widget by like tonight, so I've been fixing bugs (read: rewriting) with it. I can still do the freq counter stuff after the fact00:01
CarlFK_florent_: I am holding a hv30.  according to xfxf the hdmi out is based on the menu setting for "rec/in setup, hd standard, hdv/dv"00:44
CarlFKsounds like yours is set to dv00:44
CarlFKthe "play/out setup, comp. out" is composite out, so not hdmi00:45
CarlFKoh wait... maybe it does00:47
*** CarlFK has quit IRC01:06
*** CarlFK has joined #timvideos01:27
*** ChanServ sets mode: +v CarlFK01:27
*** springermac_ has joined #timvideos01:37
*** springermac is now known as Guest5922101:37
*** springermac_ is now known as springermac01:38
*** Guest59221 has quit IRC01:39
mithro_florent_: ping me when you get up04:47
*** glasnt has joined #timvideos04:57
* glasnt pokes mithro with something sharp and pointy04:57
mithroglasnt: heyo!04:57
*** rohitksingh has joined #timvideos04:57
glasntoh hi04:57
glasntI've replied to your latest email.04:58
glasntI'll have a hack on the CSS later on, and send you a PR when I have something not-awful.04:58
mithroglasnt: that would be awesome04:58
mithroglasnt: feel free to adjust the language as you see fit, people doing things get to make the choices :)05:00
glasntwoo, design freedom :D05:02
*** Bertl_zZ is now known as Bertl05:39
mithroglasnt: if you think you can improve the website theme, would love to see that too ;)05:41
mithro_florent_: still running into that stupid ISE bug where it gets confused and takes 2 hours :(06:55
mithro_florent_: is there a reason you didn't merge the file line ending change too?06:56
*** Bertl is now known as Bertl_oO07:33
_florent_mithro: no, we can merge theh file line ending change07:47
mithro_florent_: okay great07:47
mithro_florent_: Can you check my cleanup of your encoder changes?07:47
_florent_where is it?07:48
tpbTitle: GitHub - mithro/HDMI2USB-misoc-firmware at florent: A version of the HDMI2USB firmware based around the misoc+migen tools produced by m-labs. (at
_florent_ok thanks07:51
_florent_have you been able to test hv30?07:53
mithro_florent_: Not yet, I'm still trying to verify and understand the encoder running at 30fps on the Opsis07:55
_florent_I change two things in fact07:56
_florent_- a problem in the fps limitation (just some lines of C code), with this I was able to get up to 24 fps on the opsis and the encoder running at 50Mhz07:57
_florent_- to have more performance, I changed clock domain for the encoder and made it run at 100MHz (now we are able to encode at up to 48 fps with simple images)07:58
mithro_florent_: Why does the encode speed depend on the image?08:01
_florent_it's not really the encode speed, the the limitation on the usb throughput08:02
_florent_if you have images  that are difficult to compress (ie larger at the output) the usb throughput limitation will limit your fps08:03
mithro_florent_: so the encoder is coupled to the speed at which the USB can shuffle the bytes away?08:05
mithro_florent_: okay, do you think it makes sense to just run the encoder and save the bytes to ram, then shuffle the bytes from ram onto the USB output at all?08:07
mithro_florent_: that would give the lm32 access to the output of the encoder, right?08:10
_florent_yes but I'm not sure we will have enough SDRAM throughput for that08:11
_florent_why do you want to do that?08:12
mithro_florent_: allows us to understand how fast the encoder can run, lets us run a couple of test images through the encoder and verify their output, etc?08:16
_florent_ok but this is a bit of work, it's probably better for now to concentrate on the issues we have to fix08:19
_florent_thanks for the cleanup on my fork, I merged things08:19
mithro_florent_: btw you should just be able to set PROG=impact in your environment rather than needing to edit the files08:21
_florent_yes I know, but I'm a bit lazy and just want to do make gateware / make load-gateware :)08:22
mithro_florent_: your on windows right?08:22
_florent_for this yes since it's more practical and faster08:24
mithro_florent_: then you can set the value on the system environment?08:29
_florent_yes maybe, but that's was really for me and not important, I'll cleanup things when I'll be ready to merge in official repo08:31
mithro_florent_: so what are you planning on working on now?08:38
_florent_I'd like to fix the problem with my new buffer so that we can integrate it in official repository09:11
_florent_then try to understand why the YCbCr convertion has effect on colors09:12
_florent_and then see why we have no picture with the hv3009:13
mithro_florent_: so it looked like your encoder stuff depended on the buffer thing?09:13
_florent_no it can also work with the old09:15
_florent_do you want I integrate that?09:15
_florent_(with the old buffer)09:15
mithro_florent_: I couldn't see an easy way to distangle the commits?09:18
_florent_I'm going to integrate the software fix09:19
_florent_with that you will be able to do up to 24fps09:19
mithro_florent_: so, why wasn't the software problem affecting the Atlys board?09:22
_florent_in fact the limitation was buggy09:22
_florent_if you were able to encode more than 30fps, it limited fps to 3009:23
_florent_but if the encoder was not able to encode at 30fps (29 for exemple), it limited fps to 15...09:23
*** springermac has quit IRC09:24
_florent_since the opsis encoder was running at 50Mhz and the Atlys at 75MHz, we were able to encode at more than 30fps with the atlys, so limitation was working correctly09:24
*** springermac has joined #timvideos09:25
mithro_florent_: the csc should be roundtrip able right? IE RGB2YCbCr2 -> YCbCr2RGB should give the same results right?10:02
mithro_florent_: we should be able to set up a test for that?10:03
_florent_I just did it10:03
tpbTitle: Output via HDMI2USB shows red/pink badly · Issue #165 · timvideos/HDMI2USB-misoc-firmware · GitHub (at
mithro_florent_: Yes, what I meant was having a test which prevents a bug like that creeping back in?10:05
_florent_that's a design issue10:05
_florent_I didn't saw the colors difference while designing10:06
_florent_but I was already able to see it in fact (just that it depends of the picture you choose and for some it's easier to see that there was a problem)10:07
mithro_florent_: tests which you return pass/fail are much easier then ones you have to look at things with :)10:08
_florent_yes, but you will still have rounding errors between the original and the results of your loopback10:08
_florent_but I agree with you, we could try to improve that, but not now :), that's not the priority10:09
mithro_florent_: that was what I was asking, if the result was exact or if there would be things like rounding errors - I was under the impression it was exact?10:12
_florent_no you will have some small differences10:14
mithroI'm going to head to bed and try and catch you on the other side10:37
mithro_florent_: the HV30 cameras should be sending an interlaced HDMI video signal10:38
mithro_florent_: a couple of things that might be going wrong are10:38
mithro_florent_: the camera is sending HDMI metadata packets, causing the hsync/vsync signals to look wrong10:38
mithro_florent_: I would start with looking at the hsync/vsync signals10:40
_florent_yes I m going to add some debug10:42
mithroMight want to take a look at
tpbTitle: Have frequency (and hsync/vsync) information on the inputs in the status message · Issue #94 · timvideos/HDMI2USB-misoc-firmware · GitHub (at
*** se6astian|away is now known as se6astian10:51
tpbTitle: Support short names or tab completion on the control interface · Issue #81 · timvideos/HDMI2USB-misoc-firmware · GitHub (at
*** se6astian is now known as se6astian|away13:41
*** rohitksingh has quit IRC13:41
*** rohitksingh has joined #timvideos13:55
*** se6astian|away is now known as se6astian15:26
*** sb0 has joined #timvideos15:30
*** sb0 has quit IRC16:16
*** sb0 has joined #timvideos16:21
*** sb0 has quit IRC16:38
*** sb0 has joined #timvideos16:40
*** sb0 has quit IRC16:45
*** rohitksingh has quit IRC17:14
*** ivodd_ is now known as ivodd19:45
*** springermac_ has joined #timvideos19:48
*** springermac is now known as Guest5741319:48
*** springermac_ is now known as springermac19:48
*** Guest57413 has quit IRC19:50
cr1901_modernIt's the afternoon o.0;20:14
glasntit's morning.20:15
cr1901_modernGeez, my sense of time must be horribly off20:15
glasntand/or you're in a different timezone.20:16
cr1901_modern(Yes. Very much so.)20:16
mithro_florent_: looks like you figured out the artifact issue with your buffer?20:17
_florent_the random artifact yes20:18
_florent_I'm now trying to fix the fixed issues (bad pixel on each 8x8 block)20:19
CarlFKmithro: morning20:23
mithro_florent_: did you get a chance to look at the hv30 at all?20:31
_florent_no, I wanted to fix the encoder before20:31
mithro_florent_: okay20:42
_florent_can you eventually look at it?20:43
_florent_ie try to understand what can be wrong?20:43
mithro_florent_: Yes, will try and look at that further today21:20
mithro_florent_: Would it be helpful to have a set of test patterns we send through the encoder / device?21:21
mithroactually let me rephrase, I'm creating a set of test patterns - would they be helpful for you?21:22
_florent_I just added a black/white line pattern to test  the problem with the encoder21:22
*** se6astian has left #timvideos21:47
mithro_florent_ / CarlFK:
tpbTitle: timvideos/test-patterns · GitHub (at
mithro_florent_: I don't have the white lines style patterns yet, I just generated a bunch from gstreamer23:27
mithro_florent_: and created one for testing frame rates23:27
_florent_I already have the white/black line pattern:23:28
_florent_for(i=0; i<h_active*m_active*2/4; i++) {23:28
_florent_framebuffer[i] = 0x801080ff;23:28
_florent_that's all :)23:28
mithro_florent_: I also run all the test patterns through jpg at different levels and then look at the output23:36
*** glasnt has left #timvideos23:43
mithro_florent_: I'm going to put together some patterns with text in them too23:47

Generated by 2.13.1 by Marius Gedminas - find it at!