[eagle] CAN-Do EMI - Let's Get Going on This!

Juan Rivera juan-rivera at sbcglobal.net
Fri Jun 29 19:25:01 PDT 2007


Bdale,

I'm sorry to pick on you, but you seem like a good guy to complain to...

I seem to be having a problem stimulating a discussion.  After saying that I
though the CAN-Do power supply needed to be completely scrapped and replaced
with one running at around 1 MHz I expected to be buried in email, but I've
only received one message referring to this so far, and it wasn't from
anyone working on CAN-Do.

The CAN-Do module is unique in that it is going to be an integral component
INSIDE of every payload, so any deficiencies it may have are going to have a
large impact.  

Let me restate my finding so far:

There are 4 categories of EMI and the CAN-Do module / 70 cm Receiver
combination is experiencing all four.  They break down into radiated and
conducted emissions and susceptibility, and they are generally referred to
by a two-letter designation - RE, RS, CE, and CS.

RE and RS go together - the CAN-Do module's switching power supply inductor
radiates the 5 kHz switching noise out the back directly towards the
receiver (RE.)  The Receiver's VCO's are both very sensitive to EMI and are
impacted by the CAN-Do module if they are within 4-1/2 inches of the
inductor (RS.)  I've had to move the CAN-Do module off of the receiver PCB
and interconnect it with a ribbon cable to deal with this problem.  The good
news is that I am fairly confident that It can be fixed by going to a
two-compartment chassis, with a bulkhead separating the CAN-Do module from
the analog Receiver to provide shielding.  The bad news is that I think this
means we need a milled module chassis.

CE and CS also go together, and this is the real problem I see since you
can't fix conducted EMI with a shielded enclosure.  It requires filtering.
The CAN-Do module is trashing the DC input from the power source and also
feeding noisy power to the Receiver.  The outgoing noise is the bigger
concern because it will add to the CS problems for everything connected to
the power source.  In the other direction, the switched power from the
CAN-Do module shows up in the IF output as 5 kHz spurs.  Moving the CAN-Do
module physically away from the Receiver only dealt with the RE/RS issue.  I
had to bypass the CAN-Do module and run clean power directly from the lab
bench supply to deal with the CE/CS problem.  This means that there is no
current monitoring and no power control.

The 5 kHz switching frequency is bad for two reasons - it makes filtering
this noise a much larger problem than it needs to be, and the impact is more
severe since it is putting spurs all over the passband of the IF at 5 kHz
intervals.

If you sit back and think about the impact of a dozen noisy power supplies
all feeding EMI back to the common power source where they all mix together
and make their way back to each payload, it starts to look nasty.  All these
supplies will be drifting around and beating with each other to produce sum
and difference noise on the power bus.  5 kHz noise is hard enough to get
rid of but what if there are difference components at a few hundred Hz?  How
can you design a filter when you won't know what to expect until you hook
everything up and turn it on?  And by then you're out of time.

I'm not making this stuff up.  People I know have run into this exact
problem before and the result was very bad.

A while ago Howard Long made a suggestion that I think has great potential.
Here's what he had to say:

...in the original SDX PSU design I had in San Francisco last year is an
SMPS using the LM2672 device. These can be fitted with an AC coupled SYNC
signal to override the internal default SMPS frequency. I selected 375kHz
for my unit (6MHz divided by 16) to ensure its harmonics were outside the
10.7MHz IF passband. If the external SYNC fails the internal SMPS oscillator
takes over.

My Suggestions:

1) Revise the CAN-Do module to move the switching frequency up as far as
possible to move spurs out of the passband of sensitive analog circuitry,
and to ease the burden on EMI filtering.
2) The power and grounds must be filtered in both directions to minimize CE
back to the power source and to the payload.
3) The switching inductor should be a shielded to reduce RE inside the
module chassis.
4) A power supply capable of synching to a master oscillator should be
strongly considered.

This topic needs to be elevated to the top of the queue.  The EMI
environment surrounding the CAN-Do module impacts the design of the next
revision of the 70 cm Receiver, and also directly impacts the chassis
design.  What do we need to do to get going on this?

73,

Juan - WA6HTP  







More information about the Eagle mailing list