Electronics > Beginners
HDMI output only works with a certain 5m cable
(1/3) > >>
nuno:
Hello All,

So I have this friend that designed a carrier board for a SOM (system on a module), which has an HDMI output.
He routed the HDMI pins from the SOM's connector to a mini HDMI connector on the carrier board. He followed best practices for the differential pairs, pair lines close to each other and little further apart from other pairs, always a ground plane underneath, same length on the pairs. The HDMI lines go through an ESD protection and voltage conversion chip (specific for HDMI); the differential pairs go directly to the mini HDMI connector, the chip just hangs on them.

Funny thing is, he says the HDMI only works if he connects the output to a monitor through a 5m long (apparently high quality) HDMI cable of an unamed brand. Trying even very short mini HDMI -> HDMI cables doesn't make it work (no image or constant signal loss). If he has the short cable that doesn't work and then adds the "high quality cable", it works.

So, what do you think is hapenning here? Is the long cable adding capacitance that slows down the edges? He says that any other cable he has works on everything else, from his PC to a Raspberry PI.

Thanks, any ideas will be appreciated!
AndyC_772:
Clock / data timing skew, perhaps? Have you checked carefully what the timing specs are for your interface, and in particular, where the clock transitions with respect to the data valid window?

Maybe this "high quality" cable actually has a mismatch between the lengths of the clock and data pairs, which has the effect of moving the clock within the data window, putting it into a position where the data actually is valid when the clock transitions?
soldar:

--- Quote from: nuno on May 15, 2019, 02:22:25 pm --- Thanks, any ideas will be appreciated!
--- End quote ---

I would start by making sure all cables are wired/connected alike. That is the first thing. I remember the old VGA cables where no two were alike and would give problems.

Next make sure the connectors connect alike. I have found Ethernet RJ45 where one particular cable has a problem with one particular connector but both work well with other cables/connectors. Mysteries of science.

I have gotten to the point where all my USB cables are assigned a certain phone/function and they all work in that assigned role but once I start changing them around I start having problems.

Only after all these simple connector and continuity issues have been discarded would I start looking for more complex causes.
nuno:
Thanks for the suggestions.

I met with my friend. It is a not a "connection alike" issue, we tried several different HDMI cables and cable sets (with adapters, sometimes 3 or 4 cables in series but not exceeding ~2m) with the board, a PC, monitors, a RaspberryPI. All cables / cable sets work perfectly (full HD) with all the devices except with the carrier board he designed; only the "magic 5m cable" works with it, it is enough to add the cable in series with whatever combination of cables we already have to make it work.

We checked trace length on the PCB. The biggest difference there is is almost 3mm from CLK- to DATA0+. We built a small adapter to correct a little for the length difference and to add common mode ferrites to the differential pairs which the carrier board didn't have, but there's no difference at all.

Then we put the adapter and a "non working" cable and I stuck a finger over the exposed nodes on the adapter board until I made it work, and I did. I pinned it down to a finger over the DATA2 differential pair. My finger makes it work flawlessly :D . As I touch the 2 pins (DATA2+ and DATA2-) it starts working (monitor recognizes the image) but with a set of blinky pixels, then I start pressing a little more and the image gets better until it is 100% perfect 100% of the time.

My friend has an evaluation board for the SOM and, although HDMI works without the magic cable, it also has some small issues with the image, there are some minor interferences sometimes (a few pixels blinking, an occasional horizontal line).

What do you think of this? We don't have how to measure these signals, we have a 50MHz digital scope and that's it.
lamabrew:
Do you have more than one of these SOMs?  So far from the description it sounds like what is wonky is on the SOM and trying to swap those might give some more ideas.  It could be as simple as a bad solder joint somewhere on that board.

Maybe make a second 'inverse' adapter and stick it between some other HDMI things you have on hand.

You don't mention what resolution(s) you've tried to see if changing the clock/datarate would make a difference.  At lower pixel counts HDMI doubles up the data so lowest resolution isn't the lowest clock rate (I've forgotten the details on where that cross over is).

Last thought is DDC (I2C) failure due to all sorts of funny issues with capacitance on the lines is possible (when the design itself isn't robust) but your finger test would seem to have ruled that out.    But at least that's something you can look at on your  scope.

Ultimately a scope with the high bandwidth differential probe is the only way to know you have the interface meeting the HDMI waveform requirements, even a  used setup for only basic 1080p HDMI speeds will run you US$ 30K plus so I feel your pain on the scope issue  (add in the required TDR and jitter tests and now it's 3x the scope cost).
Navigation
Message Index
Next page
There was an error while thanking
Thanking...

Go to full version
Powered by SMFPacks Advanced Attachments Uploader Mod