LVDS and HDMI in iMX6 DualLite

[juliocruz]

Hello,

A custom board included an iMX6 DualLite with a LVDS and HDMI display split as below:

• LVDS: IPU1, display 0
• HDMI (optional): IPU1, display 1

The LVDS is working properly while the HDMI display is not working. The test criteria is based on the imx-test autorun-fb0.sh. 

I has been reviewing the different variables, as the HDMI display properties, hardware schematic and layout, u-boot, kernel drivers and discussions (here and there) regarding HDMI and dual display using the iMX6 series.

Due to the big quantity of variables that could be involved, I will share many of those through pastebin.com trying to keep the post short.

Configuration:

• Kernel 4.1.20-rt22
• u-boot 2016-03
• device tree
• u-boot board init
• u-boot config options
• kernel command line

Booting (log):

• [dmesg] system log (filtered by hdmi)
• [dmesg] system log (filtered by mxc_sdc_fb)
• [dmesg] system log (filtered by ldb)
• [dmesg] system log (filtered by ipu)

After boot (setup):

• frame buffer devices [ls /dev/fb*] 
  • LVDS: fb0, fb1 (overlay)
  • HDMI: fb2
• frame buffer configuration [fbset -fb /dev/fb]
• EDID info from the iMX board [find /sys/ -name edid -exec cat {} \;]

Testing:

• [criteria] autorun-fbX.sh. This script perform the following tasks:
  • Turn off blanking
  • Blank test
  • Color tests
  • Pan test
• [log] autorun-fbX.sh fb0 (LVDS). The results is PASS and the display show behavior is as expected. Check video here. 
• [log] autorun-fbX.sh fb2. The results is FAIL and the display show "not signal".
• [image] same display connected to Windows 10 @ 1280x720 showing NXP website

Display:

• Wacom Cintiq Companion 2
• EDID information in Windows 10 (Monitor Asset Manager)

Comments:

1. The EDID data seem OK and match with the info when the display is connected to a PC with Windows 10, as well an online EDID reader
2. The frame buffer devices are as expected, two [main and overlay] for the first device (LDVS) and one for the HDMI. Any difference (function or performance) if the HDMI is first with fb0 and fb1 and LVDS second with fb2?
3. The kernel detect when the HDMI cable is connected and disconnected (see similar log in HDMI test result log during the imx-tests, where the fb2 is blank)
4. The frame buffer setup (based on fbset) is as expected and similar to the PC in WIndows 10
5. During the testing with the HDMI display, the log show an error (multiple times) in the line #71 [echo 0,$i > /sys/class/graphics/$1/pan]: echo: write error: Invalid argument. Although the screen does not show anything, will it be worth checking this error?
6. During booting there is not sound card for HDMI registered. aplay -l command show only the sgtl5000 codec card and the log show mxc_hdmi 20e0000.hdmi_video: hdmi_enable_audio_clk related with audio & HDMI. Any thoughts regarding HDMI audio?
7. Once the test is performed in fb2, the test in fb0 not update the display anymore. After reboot, the fb0 is OK and the test result is as expected.
8. The clock frequency show in the system log [mxc_hdmi 20e0000.hdmi_video: final pixclk = 74250000] match the value obtained with a scope in the connector pin.

I appreciated any suggestion regarding this situation. I've run out of options or variables to look at.

Thanks

[juliocruz · Accepted solution]

Hello,

The issue is solved. The LVDS and HDMI are working together.

Below the answers to previous comments:

1. The EDID data is parse correctly. Sometimes a WARN_ON(true) show up from mxc_edid.c. This is something for other discussion. 
2. After fix the issue, there is not difference (apparently, I didn't perform specific testing).
3. ok
4. ok
5. The test criteria was incorrect. The script autorun-fb0.sh must be used only for fb0
6. After adding the appropriate configuration to the kernel, the HDMI sound card was listed
7. Due to incorrect test criteria, the test was messing with the frame buffers
8. ok

Below the final configuration.

• Kernel 4.1.20-rt22
• u-boot 2016-03
• Device tree - solution 
• u-boot board init (same as before)
• u-boot config options (same as before)
• kernel command line (solution)

The testing was performed using fb-test.

The display is the same Wacom Cintiq Companion 2.

After solve this situation, a new issue show up regarding the clock reconfiguration once the HDMI is connected. Basically, the LVDS clock change from 51MHz ro 20MHz causing visual defects. To solve the issue, the display parent clock (ipu1-di1) was changed to use a different clock, as show the Device tree.