Lpc2000

I have a board built with the LPC2214.  This board has 3 LDO's, a 
rectifier bridge and two offboard connectors and is also two layer.  
It is one of the cleanest micros I have ever worked with.  Without 
knowing more about the board (and other components) it would be very 
hard to say what will turn out to be the source of the interference. 
Throwing ferrites and cap's on the board can be counter productive 
without knowing the frequencies that are causing 
the problem. Do you have access to a spectrum analyzer? This is 
probably the most critical tool needed to diagnosis the problem.
 
Also be aware, that even though your crystal is a lower frequency, 
edges internal to the part(s)will occur much faster, as well as 
harmonics of the fundamental frequency.  One you have the 
frequencies and magnitude involved, then it's possible to formulate 
a strategy.
 
I believe most all the obvious fixes have been discussed already, 
but I did not see anyone ask about your trace lengths, width and 
spacing, as well as are there traces that run along the edge of the 
board.  Do you have any traces that change direction by more than 45 
degrees?  Are there any floating pins?  make sure you drive or 
pull/up down any pin without a built in pull up/down.  Are you using 
the A2D's?  If so, is your analog ground separated from your digital 
ground?
 
Just to illustrate how many different things that can go wrong in a 
complex system.  I bring up my own "nightmare" board (non LPC cpu).  
In the end we were able to get all into compliance. There were 
several small boards and one machine casting that turned out to be a 
great ground plane for picking up and transmitting RF.  All but one 
board was designed overseas, with parts purchased overseas.  The 
main board with the CPU we designed here in the USA for our 
customer.  The board drove several motors, Brushed-DC and a couple 
of Stepper Motors.  Between the Brushed-DC motor and the power 
supply and one fast CPLD.  
 
To fix, we ended up changing the CPLD for a slower,  competitive 
model.  The motors and power supply could not be changed do to cost 
considerations (consumer white goods).  The excess emissions were 
primarily broadband noise.  And not a multiple of the crystal.  
Turns out the diode rectifier was the primary source of the 
broadband noise.  Slight upgrade in parts here as well as a few well 
chosen capacitors in parallel and one ferrite bead solved most of 
the problems here.  The rest had to do with long wires coming off 
the board to sensors and to the motors.  Again choosing the right 
set of capacitors and inductors to filter the signals helped again.  
On a couple of stubborn traces, we changed the impedance.  All in 
all, I think we ended up adding about $2 to the BOM.
 
 
Tim Larson
Great Notions

BTW: Going to four layer did not help in this case... It cleaned up 
the lower frequencies, without much effect to the 700Mhz and beyond.
Of course, YMMV


> Yep, all those things we've done. Top of the board is 100% 
components
> and 99% signals and the bottom is 1% signal and 99% ground plane. 
Top
> of the board that is not used by traces is also a ground plane. 
Took a
> while to lay it out, but it's better then going multilayer. I also
> added CTS resistor arrays right next to the processor. They are 
great
> since they have 8 single resistors on 0.5mm centers. My capacitors 
on
> the other hand are at the connector locations which is about 3 
inches
> away (100pF).
> 
> I was just wondering if anyone of the LPC processors has EMI
> experience? We recently went throught this with the PXAG49 
processors
> from Philips and found that the control lines are very noisy at 
29.4xxMHz.
> 
> regards,
> Charles