[Cryptech Tech] STM32F4 <-> Novena Bridge

Peter Stuge peter at stuge.se
Thu Jun 11 19:50:26 UTC 2015


Fredrik Thulin wrote:
> > I've started this task, see the attached picture for tentative board size
> > (10x5 cm) and placements.
> 
> Finished routing of the FMC signals. New picture attached.

Sorry that this comes only after they've been routed - is there a
reason for the STM32 to be on top instead on bottom?

It looks all right, but moving vias around a little so that a trace
which goes straight on top and wants to go left on bottom does not
first go diagonally right to a via on top and then straight left on
bottom, but instead goes diagonally left to a via on top, and then
continues left on bottom. Does that make sense? It's a minor
optimization to reduce reflections and make things neater.


> the FMC traces are between 29 and 62 mm
..
> (Pavel) helped me do some back-of-the-envelope calculations in
> Stockholm that said the traces should not differ by more than 150
> mm (!) for IIRC 100 MHz signals so maybe it's not worth any more
> effort at all?

--8<-- Chapter 10: Time and Distance, Right The First Time
Electrical signals travel as electromagnetic energy. This EM energy
travels at the speed of light ... roughly one foot or 30 centimeters
in a billionth of a second or a nanosecond (10 -9 second). When this
electromagnetic energy must travel in a dielectric or insulator, such
as in a PCB, it is slowed down by the dielectric. In a PCB, the
dielectric slows the signal down to roughly half the speed of light
in a vacuum. As a result, a signal will travel roughly 6” or 15
centimeters in one nanosecond along a trace in a PCB.
-->8--

When trace lengths vary less than 150 mm then signals can be said
to be synchronized to within less than 1 ns. 100 MHz has a 10 ns
period and 10:1 is a safe margin.

When trace lengths vary less than 33 mm, as on your board, then
signals can be said to be synchronized to within less than 220 ps.

With 10:1 margin signalling would still be successful on your board
up to 450 MHz, as far as trace lengths are concerned.


//Peter


More information about the Tech mailing list