0
\$\begingroup\$

I have a Raspberry Pi and a custom HDMI screen board stacked as shown in the photo here.

enter image description here

If I use a 'standard' HDMI lead then I have no problem. However, due to the form factor a lead does not suit my application so I created a HDMI connector PCB as shown below, and this does not work (the screen just remains backlit).

enter image description here

The first mistake I apparantly made was assuming point-to-point connections. I metered out a lead that worked and noticed that the GND connections are not through connections. I cut all these tracks, the problem persisted. Even though I have not been able to find any evidence for not having point to point connections on any pinout diagram

enter image description here

I then dissected a known working cable just to double check I hadn't missed anything obvious and couldn't see anything. So then I cut down this cable to make it so it could at least get me up and running in my application ... however, the cut-down version now does not work (blank screen).

enter image description here

The commonality between the two is that there is no shielding and/or twisted pairs ... I thought that this might have some impact on quality, but not cause it to not work!?

There are many 'PCB' adaptor boards online, so I am sure this is not just a case of wrong impedance/shielding ....

Any thoughts and suggestions would be very welcome.

Regards, Tom

\$\endgroup\$
  • \$\begingroup\$ I think it probably is wrong impedance/shielding, at that speed. Can you at least determine whether the DDC/CEC is working (try "read-edid"on the Pi) \$\endgroup\$ – pjc50 Nov 27 '17 at 10:27
  • \$\begingroup\$ DDC/CEC is very, very slow by comparison with the video data so should work even if nothing else does. Also check that you are seeing 5V on the HPD line. \$\endgroup\$ – Dan Mills Nov 27 '17 at 11:45
1
\$\begingroup\$

Horrible signal integrity!

The TMDS channels (and TMDS Clock) are FAST as in multi GHz edge rates. You cannot route them without respecting that nature.

The ground connections should be either thru or all commoned (Common is the way to go on a PCB because you need a ground plane on the board to define the pair impedance.

On the PCB you need all the TMDS lines to be run over a ground plane with a controlled impedance, this usually means you want either a buried ground plane or a thin PCB to get the trace widths reasonable at the target impedance value.

It is all in the HDMI electrical specifications, and is really not rocket science, but you need to think like an RF (and low band microwave) designer, and that has a somewhat undeserved reputation for being black magic.

A search on "Signal integrity" will bring up more then you ever wanted to know.

\$\endgroup\$
  • \$\begingroup\$ Appreciate the affirmation of what I feared :-) I have obeyed these design rules on my main board, but then looking at some of the 'commercially' available HDMI break out boards I relaxed maybe a bit too much on the adaptor board. \$\endgroup\$ – tomdertech Nov 27 '17 at 11:49

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.