[Arm-netbook] HDMI High-Frequency Layout: Timing

[Richard Wilbur]

On Tue, Aug 1, 2017 at 8:25 AM, mike.valk at gmail.com <mike.valk at gmail.com> wrote:
> 2017-08-01 16:11 GMT+02:00 Luke Kenneth Casson Leighton <lkcl at lkcl.net>:
>>  *sigh* unfortunately symmetry is not completely achievable with the
>> drastically-reduced amount of space available.  the HDMI signals come
>> out right at the bottom of the board.
>
> That might not be to problematic. I've search to the net for talk
> about running tracks on top of each other. It keeps hunting me. And
> found a knowledable awnser.
>
> http://www.sigcon.com/Pubs/news/2_30.htm

Thanks for the interesting read.  It is an intriguing geometry and
looks like it uses vertical space more than horizontal.  I suppose we
could make use of a 2-D EM field solver to figure out design
parameters such as trace width, which layers to use (how much
dielectric thickness between traces), and horizontal offset from other
traces.

> On of the things mentioned is that the differential signals might not
> be quite aligned to begin with. So achieving symmetry might look nice
> on but can only give limited help in minimizing emission and pickup.

I guess to get better than that we would have to characterize the
differential signal sources and, if they have a repeatable output
skew, then design the traces with that initial skew from the source.
I have a suspicion that the important goals for this project include:
1.  operational HDMI v1.4 interface supporting all operating modes of
which the A20 chip is capable (clock up to 340MHz, data up to 3.4GHz)
2.  EMI radiation below regulatory limits for all concerned agencies:
FCC, CSA, EU, etc.
3.  EMI sensitivity small enough to avoid disrupting proper operation
of all systems of the EOMA68-A20.

Hopefully the drivers on the chip are aligned well enough to fit into
the skew budget for HDMI operation and we are able to execute a PCB
trace geometry to support all three goals.  (I'm willing to bet on
it!)