Jump to content

v.5.17 and DVI monitor


Recommended Posts

Hello, I was able to complete custom (with USB gadget support) build of current Armbian version (5.17) for desktop using tip about workaround for u-boot compiling problem. But when I run h3disp in order to allow DVI monitor connection (-d), after the restart I still get no output to DVI monitor, while remote Putty connection works. I understand that this is version-in-development, but is there any way to fix (workaround) the problem?

Link to comment
Share on other sites

Thank you for answer! I work with Orange Pi PC card and use same video mode I used to use with no problem in older versions:

 

h3disp -m 9 -d

 

Could you provide me a tip how to compile older version, for example, 5.16 with Ubuntu xenial desktop (and kernel 3.4, of course, as usual)?

Link to comment
Share on other sites

Latest tagged build source is https://github.com/igorpecovnik/lib/releases so you can try to compile with: LIB_TAG="v2016.06" while there is no warranty that compilation will succeed. There are (too) many "variables".

 

Ubuntu Xenial desktop is officially unsupported while it should work. We focus to bring things together on Debian Jessie.

Link to comment
Share on other sites

I tried to build Debian Jessie 5.17 and DVI doesn't work.

 

When I used the tag as you suggested:

...

LIB_TAG="v2016.06"            # empty for latest version,
                    # one of listed here: https://github.com/igorpecovnik/lib/tagsfor stable versions,
                    # or commit hash

...

I got error message:

 

error: pathspec 'v2016.06' did not match any file(s) known to git.
 

Tried to download tar.gz file tagged v.2016.06 and got compilation error.

 

About the v.517:

 

fbset program shows that the video mode has switched to 1920x1080 after running of h3disp.

 

So is the HDMI monitor my only option to advance?

Link to comment
Share on other sites

h3disp

 

...is a pretty simple shell script that tries to patch the fex file (converted on the fly from script.bin) to adjust some settings, then convert the fex file back to script.bin so that after the next reboot hopefully different display settings are active.

 

So instead of trying to build Armbian from sources I would enter debug mode for this script (add " -x" to the first line) or simply edit the fex file by hand. Please check or better provide the output of the following two commands:

readlink /boot/script.bin
bin2fex /boot/script.bin | egrep "disp_|screen0_|hdcp|hdmi"
Link to comment
Share on other sites

Hello, I have to compile the system, because I develop USB Gadget based on GadgetFS and the default system doesn't include it. I managed to work with older system (5.16) and DVI, but compiled it with debug information. Now I work about release version, meanwhile sources have changed to 5.17, and now DVI doesn't work. I have a HDMI TV set at home, but not in the office. This is the output of two commands you asked about:

 

I did the command

 

root@orangepipc:~# h3disp -m 1080p50 -d
 

then restarted and then

 

root@orangepipc:~# readlink /boot/script.bin
bin/orangepipc.bin
root@orangepipc:~# bin2fex /boot/script.bin | egrep "disp|screen0_|hdcp|hdmi"
fexc-bin: /boot/script.bin: version: 0.1.2
fexc-bin: /boot/script.bin: size: 36412 (81 sections)
[boot_disp]
advert_disp = 0
hdmi_channel = 0
hdmi_mode = 4
hdmi_mode_check = 1
[disp_init]
screen0_output_type = 3
screen0_output_mode = 9
disp_init_enable = 1
disp_mode = 0
[hdmi_para]
hdcp_enable = 0
hdmi_cts_compatibility = 1
hdmi_used = 1
hdmi_power = "vcc-hdmi-18"
hdmi_used = 1
root@orangepipc:~#
 

They look correct for me:

 

hdcp_enable = 0

hdmi_cts_compatibility = 1

 

But DVI monitor still doesn't work.

 

Please advise.

Link to comment
Share on other sites

They look correct for me:

 

hdcp_enable = 0

hdmi_cts_compatibility = 1

 

Yes, the settings look correct but the only known difference switching this is that DVI displays then get an RGB signal and not YPbPr (which is default with Allwinner's most recent BSP kernel for whatever reasons when no directly connected HDMI display is detected -- this different signaling is suited for HDMI-to-Composite converters). So in case you don't use 'h3disp -d' then you could also prevent your display from displaying wrong colors by switching input color space from RGB to YPbPr as can be seen in post #28 here.

 

Unfortunately no idea why your display seems to not work any more. But since I fooled myself so many times thinking too much and avoiding to search for simple reasons first I would double check or even change cabling/connectors and display settings before I would think about recompiling software. Reverting back to a previous version of our build system seems pretty useless to me since you get different display behaviour either by a newly introduced bug in h3disp (not true) or kernel (maybe but then you would've to check the contents of our kernel patches directory and manually remove patches).

Link to comment
Share on other sites

Thank you for answer! I have two workstations with DVI monitors, and the problem happens in both.I still have the older version (v.5.16) that I prepared with debug information and it works with DVI monitor. I have some other backup folders too.

My question is, if I can do compilation in off-line mode.Then I would be able to compile 5.16 without debug information. The USB connection in debug mode works too slow, because of a lot of debug messages I added to analyze a data exchange process.

And I would be interested to keep software up-to-date, when possible, The DVI support is not critical for final release, more for development time. And I believe I am not alone with DVI monitor.

Link to comment
Share on other sites

My question is, if I can do compilation in off-line mode.Then I would be able to compile 5.16 without debug information. The USB connection in debug mode works too slow, because of a lot of debug messages I added to analyze a data exchange process

 

Well, removing the debug stuff from kernel/modules is always possible, simply use our build system to only generate u-boot + kernel, then grab the .debs that appear below output/debs/ and install the kernel related ones (linux-image and headers) using 'dpkg -i' on your Armbian system.

 

I wanted to test out DVI now but this has to wait since I'm in the process to do some time consuming measurements first (with HDMI totally disabled for energy saving reasons). Just to be clear: You have more than one Orange Pi PC (this is one of the H3 devices without any HDMI ESC protection!) and all those do not work connected to various DVI monitors with 5.17 while the very same hardware when running an older Armbian image does. Or is the statement 'with 5.16 it works' a historical statement and should read 'a few days ago when I used 5.16 it worked'?

Link to comment
Share on other sites

I have two SD cards - one with 5.16 debug and another with 5.17. I can see 5.16 desktop in DVI monitor and can't in 5.17, both switched to DVI mode using the same h3disp -m 9 -d command. Will wait on your results, take your time.

Link to comment
Share on other sites

Hi, yesterday I've compiled new Armbian v.5.17 in a new folder in attempt to check current status of DVI problem. The "h3disp -m 9 -d" command still doesn't bring a good result with DVI monitor. But the new option - "h3disp -m 36 -d"  - let me see the Ubuntu desktop screen in my DVI monitor (1680x1050 is not native resolution, but finally it works). Hope this will let to identify the problem with 1080p resolution and DVI. I believe it is connected to some kernel patch as mentioned in the new version of h3disp.

Link to comment
Share on other sites

Hi, yesterday I've compiled new Armbian v.5.17 in a new folder in attempt to check current status of DVI problem. The "h3disp -m 9 -d" command still doesn't bring a good result with DVI monitor. But the new option - "h3disp -m 36 -d"  - let me see the Ubuntu desktop screen in my DVI monitor (1680x1050 is not native resolution, but finally it works). Hope this will let to identify the problem with 1080p resolution and DVI. I believe it is connected to some kernel patch as mentioned in the new version of h3disp.

We added several new screen resolution but they are not all tested and problems could exists. I only try modes, which are supported by my monitor.

h3 util patch:

https://github.com/igorpecovnik/lib/commit/8c76c504a4f058bfc3ca29452d21012790991e4f

kernel patch:

https://github.com/igorpecovnik/lib/commit/144f704c41e4adea0d8cc00670069cfed0383600

Link to comment
Share on other sites

Thank you for answer! The problem still exists with the old, previously worked option - 1080p50 and DVI. The problem started to happen during changes in v.5.17.I have one Orange Pi PC working well with v.5.17 compiled on 12th July and it works well with 1080p50 and DVI monitor. Hope this helps to identify the problem.

Link to comment
Share on other sites

Guest
This topic is now closed to further replies.
×
×
  • Create New...

Important Information

Terms of Use - Privacy Policy - Guidelines