In the PYNQ-Z1 video processing flow, the video cannot be displayed on the external monitor

[ars_M]

I am currently working with Vivado 2023.2 and the PYNQ-Z1 development board.
The HDMI input is connected to my laptop, and the HDMI output is connected to an external monitor.
Although I am able to load the overlay in Jupyter Notebook, the external monitor does not display any video output.

[marioruiz]

Hi @ars_M,

Welcome to the PYNQ community.

This is an issue we have recently noticed for Vivado versions that we have not verified. The verified Vivado version is 2022.1. We are addressing this for the new release.

Please, check all the AXI4-Stream IP involved in the pipeline have tkeep, tlast and tuser enabled.

Mario

[ars_M]

Thank you for your response, @marioruiz. I have not found the tkeep signal option in the IP configuration settings. Could you please advise on how to enable or access it?

Mars

[marioruiz]

Have you checked in the IP that consume/produce the AXI4-Stream for this hierarchy?

[ars_M]

I’m not sure if my ILA is connected correctly. With the current connection setup, I cannot detect any ILA cores in the Hardware Manager.

[ars_M]

I previously tested this video processing flow successfully, but after adding the video in/out IPs, it no longer works.

[marioruiz]

Hi @ars_M,

I suggest you use the PYNQ-Z1 base overlay as starting point.
Unfortunately, this topic is a bit outside of the scope of this forum.

Mario

[ars_M]

Thank you for your suggestions, but I would still like to resolve this issue thoroughly.

I inserted an ILA between the v_vid_in_axi4s and v_axi4s_vid_out to monitor the AXI4-Stream video data path. From the waveform, I can see that tdata is continuously active with valid pixel values, tready is asserted (high), tvalid toggles between 0 and 1, and tlast also toggles appropriately. However, tuser remains constantly low (0), which is unexpected since I would typically expect a high (1) on tuser to indicate the start of a frame (SOF).

Subsequently, I inserted another ILA after the v_axi4s_vid_out block to observe the vid_io_out interface (Video Out side). All outputs including vid_data, vid_active_video, vid_hsync, and vid_vsync are stuck at zero, indicating that no valid video timing or data is being generated.

At this point, I’m unsure whether the root cause is related to the missing tuser pulse (SOF) in the AXI4-Stream input, or a configuration or timing issue within v_axi4s_vid_out that’s preventing it from producing valid video output.

Any insights on whether I should prioritize debugging the AXI4-Stream tuser behavior or investigate the Video Out timing and reset configurations would be greatly appreciated.

[dikshant]

hi @ars_M
Can You tell me if u done figured out the problem.
I am also stuck at same place.

[ars_M]

hi @dikshant
I still haven’t resolved the problem. Recently, I updated dvi2rgb and other related IPs to the latest versions in Vivado 2023.2, but the issue persists — the external monitor still shows no video output.

This problem has blocked my progress for over two months, and I’ve tried many approaches without success.

If anyone has encountered a similar issue, or is willing to exchange thoughts and debug this together, I would really appreciate it. I’m open to collaboration and any possible suggestions. Thank you in advance.

[ars_M]

@dikshant @marioruiz
I finally found the root cause of the issue!

In my design, the Video Timing Controller (VTC) was configured to use both Generation and Detection modes simultaneously. This caused the VTC to attempt checking the timing signals from the video in interface. Since some expected signals were missing or inconsistent, the detection logic interrupted the generation process, which resulted in no video output to the external monitor.

After I disabled Detection mode and manually configured the resolution and video timing parameters, the video dataflow started working correctly, and the external monitor displayed the image as expected.

This problem cost me over two months of debugging, and I suspect this may be a regression or unexpected behavior introduced in Vivado 2023.2, since I don’t recall encountering this issue in earlier versions.

Hope this helps anyone facing similar issues. If you’re also seeing strange behavior from the VTC in Vivado 2023.2, I’d be happy to exchange thoughts.

[marioruiz]

Thank you for contributing back the solution.