EMI, fail at SDRAM

[Boki]

Hi All,
Not familiar with EMI.

My baord encountered EMI problem in SDRAM clock.

1. I am using 4 level board.
2. The problem comes from SDRAM clock of core chip.
3. It not all fail, it fails on some high ( multiplier ) frequencies of
SDRAM clock.
4. I had added iron...
5. The layout of other circuits is very compact.

I had add some 0 ohm resistor / RC filter...

I think my major problem is the SDRAM clock.

Thank you very much for your comment.

Best regards,
Boki.

 

[Larry Brasfield]

Hi All,
Not familiar with EMI.

My baord encountered EMI problem in SDRAM clock.

1. I am using 4 level board.

Is it surface mount? How many plane layers?
Is the ground plane pretty contiguous or is it
broken up a lot for routing use?

2. The problem comes from SDRAM clock of core chip.

On what evidence do you conclude that?

3. It not all fail, it fails on some high ( multiplier ) frequencies of SDRAM clock.

What frequencies are giving you trouble?
What is the clock frequency?

4. I had added iron...

Where? Is the circuit in an enclosure? What is
that made of? How are the signals entering or
leaving the enclosure treated? Is your board
connected to the enclosure near the openings
that signals enter or leave?

5. The layout of other circuits is very compact.

I would expect the address and data lines to be about
as spread out as the RAM clock.

I had add some 0 ohm resistor / RC filter...

That might help with high order harmonics. Have you
tried increasing the resistor value? What kind of
termination do you have on the clock and other
high speed signals?

I think my major problem is the SDRAM clock.

With a spectrum analyzer you should be able to
tell whether it is the constant frequency clock or
other, less constant transitions whose spectra
are around the clock but more spread out. The
address and data lines could easily produce the
kind of problem you have described.

Do you have a means of evaluating incremental
changes in your circuit's radiation at the frequencies
causing the problem in your own lab? Or must you
rely on test results from an EMC testing house?

Thank you very much for your comment.

You are welcome.

 

[PeteS]

In addition to what Larry said (and I completely agree with all the
comments), other considerations are :

Was the layout of the signals set for impedance controlled tracks?
(this implies a ground plane somewhere, and if you get the impedance
right - which can be forced with the correct terminations - helps
enormusly with EMI issues)

If you know the source of the emissions is the core device, can you use
a spread spectrum oscillator to run it? That helps with EMI issues (by
spreading the energy across a wider band, thus reducing the emissions
at any one frequncy, at least for long period type issues).

What is the clock running at? If it's high, you may be able to slow it
down and that may very well help too.

Cheers

PeteS

 

[Joerg]

Hi Pete,

If you know the source of the emissions is the core device, can you use
a spread spectrum oscillator to run it? That helps with EMI issues (by
spreading the energy across a wider band, thus reducing the emissions
at any one frequncy, at least for long period type issues).

In some countries it is called "clock dithering". There are or at least
used to be chips that could even dith a clock after the fact. If I am
not mistaken Cypress is one of the companies.

Regards, Joerg