HDMI HPD disconnect does not disable HDMI Data and Clock

Technical Support
, , , ,
1

Hi,

We have a problem with our display solution with Verdin (iMX8MP Q 4GB IT V1.1A) / Kirkstone. We are using a transmitter IC to convert the HDMI1 signal to DisplayPort and wrap it into USB-C alternate standard to connect a USB-C monitor supporting DP-alternate mode.

The monitor is always working after startup. But when the display is disconnected and connected again, the monitor stays sometimes dark. When the same Adapter is connected to Windows 10 (HDMI / USB) it is always working.

Our analysis has shown, that after startup, without connected HDMI, no signal is present at the HDMI Data and Clk line (as it should be).
After connecting an HDMI monitor, the HPD Pin of the HDMI goes high and Data and Clk starts, the display is working.

Verdin HDMI connect
Verdin HDMI connect1280×800 92.7 KB

Yellow = Clk / Mangenta = Data / Green = HPD

When now the HPD Pin is pulled low (disconnect) there is still signal on Clk and random signal on Data. This shouldn’t be. It should go back into the original configuration before the initial connection of a HDMI monitor.

Verdin HDMI connected HPD to Low
Verdin HDMI connected HPD to Low1279×799 80.3 KB

Yellow = Clk / Mangenta = Data / Green = HPD

When instead of pulling low of the HPD pin the HDMI Connector is disconnected, the Clk and Data signals are idle again.

Verdin HDMI disconect
Verdin HDMI disconect1280×800 68.7 KB

Yellow = Clk / Mangenta = Data / Green = HPD

We have strong reason to believe, that this Clk and random noise on the Data line is disturbing the synchronization with our transmitter IC.

This behavior is easy reproduced on the Dahlia Carrier Board (Rev. 1.1). Just pull HDMI1 HPD to GND to simulate a disconnect with connected monitor. All Verdin Pictures were done on the Dahlia Carrier Board.

In Windows the Clk and Data Line are set to idle when the HPD is pulled low.

Windows HDMI connected HPD to Low
Windows HDMI connected HPD to Low1280×800 63.4 KB

Yellow = Clk / Mangenta = Data / Green = HPD

Thank you in advance for your help.

Best regards

Christian

2

Thank you for the documentation,

We are looking into it.

Best Regards,

Matthias

3

Hello,
do you see the same behavior on the other HDMI port? The one with the DSI to HDMI converter?

Best Regards,

Matthias

4

Yes, with the DSI to HDMI converter the Clk keeps on running with HPD pulled low.

Best regards,
Christian

5

Hello @C_Mueller,

which BSP or image version are you using?

Best Regards,

Matthias

6

we flashed Verdin-iMX8MP_Reference-Multimedia-Image-Tezi_5.7.2+build.21 (https://artifacts.toradex.com/artifactory/tdxref-oe-prod-frankfurt/dunfell-5.x.y/release/21/verdin-imx8mp/tdx-xwayland/tdx-reference-multimedia-image/oedeploy/Verdin-iMX8MP_Reference-Multimedia-Image-Tezi_5.7.2+build.21.tar)
We observe the same behavior: clock is still running when pulling HPD low.

7

Are there any news? We would like to proceed in our project and can’t until this problem is fixed.

Best Regards
Christian

8

we have to do some tests by ourselves.
On which point at the schematic did you measure the Hot plug pin?
Can you try put the pin low directly on the SOM ? before the ESD chip?

Best Regards,

Matthias Gohlke

9

For ease of access, I used a breakout board in between two HDMI cables for all measurements (Dahlia Out - breakout board - display). But I just tested when I pull HPD after the 22R series resistor (RA11B) low. CLK was still running. The behavior is the same.

Best Regards
Christian

10

Hello @C_Mueller,

just to make sure that there is no influence of the ESD chip or cable on the other side. Did you try to pull the pin before the ESD chip on the Dahlia?

Screenshot from 2023-11-22 12-17-50
Screenshot from 2023-11-22 12-17-501385×886 120 KB

Best Regards,

Matthias Gohlke

11

Hi Matthias

In the first tests I pulled it in the cable. This morning I tested it at the Verdin with the same results.

image
image878×467 34.9 KB

Also in our solution is no ESD chip. We measure with the same behavior.

Best Regards,
Christian

12

Hello @C_Mueller,

and on the other HDMI interface? (DSI to HDMI) Is it the same behavior there?

Best Regards,

Matthias

13

You asked that 22d ago and my answer was yes. Or did I not understand you correctly?

14

hello @C_Mueller,

I just wanted to reassure myself.
so we just tested it on the native HDMI of the Verdin Development board by unplugging the HPD jumper and it behaved as expected. To see if this is related to some driver issue. I keep you posted.

Best Regards,

Matthias.

15

Hello @matthias.tx

I would like to know if there was any progress in the analysis of this problem.

Best Regards,
Christian

16

Hello

I would like to know if there was any progress in the analysis of this problem and the timeline. As already said, the project is not progressing without solving this problem.

Best Regards,
Christian

17

Hi @C_Mueller,

we are still investigating this in our software team.

Best Regards,

Matthias

18

hello

we made some measurements on the DSI->HDMI converter. It also always turns off the clock, but the time it takes to do that varies a lot. It can even take ~550ms for the clock to shut down in some cases:

SDS00002
SDS00002800×480 10.9 KB

SDS00003
SDS00003800×480 8.7 KB

Best Regards,

Matthias Gohlke

19

Hello
Did you just pull HPD low or disconnect the cable? And why do you measure on the DSI->HDMI Adapter? We use the native HDMI.

Best Regards,
Christian

20

@C_Mueller,

Well, there has been some confusion about the use case in the team that was investigating it on the HDMI driver side.
So now we did both just pulling the HPD pin low and also unplugging. Your use case with the HDMI to EDP to USB_C was not clear to everyone. Please attach a schematic for clarification since sometimes different teams are involved.
We tested with unplugging and with just moving the HPD pin low.
When Unplugging we can confirm your results that the Clock and data go low. When leaving the HDMI plugged in and just pulling the HPD pin low we see the same conflict of the clock and data line still showing signals. But the attached display goes off and on when we release the HPD pin. So this is why it went unnoticed.

We checked the native and the DSI to HDMI interface with the same results.
In the driver, we only see the HPD interrupt. Nothing else would hint at that behavior.
So there seems to be a mode on the phy level that probes the data bus to see if there is something connected independent of the HPD signal. This would also explain why this is happening with the two HDMI interfaces even if the software implementation is different. We just don’t know yet if and where we can disable that behavior.

We like to set up a call with you.
Can you please send the schematic with the convert section to us @ support@toradex.com and also your preferred time for a call?

Best Regards,

Matthias Gohlke