As usual, the devil is in the details. Microwave engineers know well that
surface traces do have radiation loss. As with the case of right-angle
bends,
these effects are real, we just need to determine whether it creates a
problem
in the system on hand. And in case anyone might wonder: yes, we had EMI
issues
which were tracked down to surface traces over solid planes. The
clarification
though is important: those were associated with the clock of the
high-speed channel.
I do not recall any instance when we had radiated EMI issue tracked down to
random data. Data behaves like dithered or 'spread-spectrum' clock,
reducing
the signal power falling into any given test bandwidth. So early on we
adopted
the rule that having data traces on the surface was OK over solid
planes, but
the first choice for clock signals was always an inner layer. As it was
also pointed
out, the radiation is miniscule, a good enclosure should help to stop
it. While
this is true, not every system with high-speed data communications is
allowed
to have metal or metalized enclosure. Furthermore once the radiated wave
leaves the source, in a complex large system there might be structural
resonances
that can pick up this signal and transfer it outside the box. The
conclusion is that
it is always easier to stop the radiation at the source. And to the
question about
the serpentine routing: I do not recall having found any radiation issue
due to
serpentining data traces on outside layers.
Regards,
Istvan Novak
Samtec
On 6/4/2021 11:16 PM, Zachariah Peterson, NWES wrote:
Hi everyone,------------------------------------------------------------------
While I was conducting a design review today, my client expressed some
concern over serpentine routing that was used to enforce length matching
between an SDRAM chip running at 166 MHz (1 ns rise time) and an MCU.
The concern was that the serpentine routing structure would create an
EMI problem. The traces are routed according to specs in the datasheet
with a reference plane on the adjacent layer.
I have never heard of serpentine routing creating a new EMI problem or
making some existing EMI problem worse. I started looking into it and I
find mixed recommendations. This is one instance where I won't trust
application notes, they all give inconsistent guidance, and they show
poor images of the exact way serpentine routing should be implemented.
I looked into the literature and there are some studies on EMI, effects
on eye diagrams, and reflections from serpentine routing structures.
Here's one example:
https://www.researchgate.net/publication/321143234_Differential_Lines_Paired_with_Serpentine_as_Potential_EMI_Aggressors_in_Mobile_Electronic_Devices
If anyone can provide some perspective, or if someone can confirm this
is just another signal integrity myth, I'd appreciate it. Thanks in
advance.