Greg Olson wrote:

> Hi all..
> I've got a 2.5"x3.5" plug on daughterboard which contains my processor, memory, 
>flash, etc... The processor is a 20MHz AM186 driven by a 20MHz crystal. I just had 
>this system in for FCC EMC testing and found that the 20MHz crystal is bleeding its 
>harmonics (especially 80MHz) all over everything! It's getting onto the IO lines from 
>the daughterboard to the main board and from there onto just about every wire leaving 
>the enclosure!

Since the crystal is not running all by itself, but is clocking many other circuits, 
it is not absolutely positive
that the harmonics are coming from the oscillator only, but it could be many other 
signals, like memory
strobes that is causing it.  You might be able to lock the processor in reset and see 
if the 80 MHz emission
drops, to tell whether it is the oscillator alone.

It seems to me that square waves are supposed to be composed of ODD harmonics.

One fix I have heard of is to put a small series resistor on the output line of the 
oscillator.  Also,
switching to an oscillator with a lower output current may be possible.  But, it 
sounds like you have
a built-in oscillator in the chip, and are using a bare crystal?  You might be able to 
lower the
drive level to the crystal with a series resistor (this is necessary in 32 KHz 
crystals, which will
self destruct if given too much drive).  Do you have capacitors to ground to resonate 
the crystal?
A common circuit uses 2 caps of about 15 - 33 pF from each side of the crystal to 
The 'sense' side is fine, but on the 'drive' side, the output driver of the crystal 
oscillator is
now trying to pump a (5 V) square wave into a 33 pF cap to ground.  This will cause 
current pulses unless the output driver is specially designed to have 'weak' output.  
It may be
that placing a resistor of a couple hundred ohms in series from the drive pin to the 
junction will relieve the chip of that heavy capacitive load, and stop the problem.

As you note it is a 4-layer board, with power and ground planes, it should have pretty 
grounding of the power nets, so that shouldn't be a problem.  (I've frequently had 
EMI gremlins on 2-layer boards.)  But, you might check the power, ground and decoupling
layout on chips that are likely to sources of the 80 MHz emission, to see if the 
routing of
the tracks could be shortened.  Minimum track length (or, optimally, zero length by 
use of
thermal reliefs on the pin's through holes) of the power and gnd connections is 
as CMOS chips tend to draw power pulses right at clock edges.  Do you have DRAM on
this board?  CMOS static RAM can also pull large pulses on the power lines when address
lines change (sync SRAM can pull pulses when clocked).  But, DRAMs are phenomenal
EMI sources when the strobes are pulses, as they can switch 300+ mA PER CHIP!
(The newer stuff is usually a lot less than that.)

I hope this gives you some ideas of where to look, and a few tricks that may help
isolate the specific chip or region of the circuit that is the source.


* * * * * * * * * * * * * * * * * * * * * * * * * * * * * *
* To post a message: mailto:[EMAIL PROTECTED]
* To leave this list visit:
*                      - or email -
* mailto:[EMAIL PROTECTED]?body=leave%20proteledaforum
* Contact the list manager:
* Browse or Search previous postings:
* * * * * * * * * * * * * * * * * * * * * * * * * * * * * *

Reply via email to