# Longest path length from SDRAM controller to DRAM?

Discussion in 'Embedded' started by sendthis, Oct 16, 2007.

1. ### sendthisGuest

I'm trying to design a test setup to do radiation experiments on SDR
DRAM
for my thesis. The DRAM needs to be isolated from our control unit b/c
I'm
worried about contaminating our results by having our control
circuitry
in there too - b/c then how do we know if it's the DRAM or
"something"
else that fails.

So my question is, what's the longest length I can run a ribbon cable
(or other type of connection) and not having issues? Can I bring the
clock down? We are using 20-24 inches of cabling. We decided maybe to
use CAT6 cable?

Thanks,
Eric

sendthis, Oct 16, 2007

2. ### CBFalconerGuest

You can get a rough idea by the propagation time. 2 feet of cable
will have a one-way propagation time in the order of 2 nanosecs,
and a characteristic impedance in the order of 100 ohms. So
ringing will be in the order of 250 MHz. The cures are proper
matching and timing.

CBFalconer, Oct 16, 2007

3. ### Not Really MeGuest

Someone with strong current hardware skills is sure to answer this better,
but I think the length through ribbon cable is basically 0 inches. The
timing is very fast and will be very susceptible to cross talk. Many
current SDRAM designs even use trace length balancing to insure all the
signals are available at the same time.

You will most likely need another solution, possibly an RF shield can over
the other circuitry.

Scott

Not Really Me, Oct 16, 2007
4. ### Tim WescottGuest

So, two feet of ribbon cable. Take a velocity factor of 0.66 for grins,
which gives you an effective 3 foot free-space length. Then consider
that the signals have to go there and get back.

Hmm. That's 6ns, or just about 1/166MHz. And that's ignoring any nasty
impedance bumps you may have in your connector and cable, not to mention
cross-talk.

What was your clock rate again?

Did you check for a minimum clock rate on the data sheet? The last time
I was around SDRAM there was a minimum not-to-be-exceeded clock rate,
and it wasn't all that much lower than the maximum not-to-be-exceeded.

--

Tim Wescott
Wescott Design Services
http://www.wescottdesign.com

Do you need to implement control loops in software?
"Applied Control Theory for Embedded Systems" gives you just what it says.
See details at http://www.wescottdesign.com/actfes/actfes.html

Tim Wescott, Oct 16, 2007
5. ### Jim StewartGuest

A stupid question.

Are you exposing the SDRAM to radiation
or are you measuring EMI from the SDRAM?

Jim Stewart, Oct 16, 2007
6. ### Tom LucasGuest

That's a bit harsh, although the poor chap seems unable to spell
"because".
Either way the OP could probably get reasonable results by wrapping his
controller up in aluminium foil or, more professionally, a metal
enclosure with proper braided earthing. However, I believe they
mentioned that they were expecting the device to fail so presumably they
are testing for EMI succeptibility.

Tom Lucas, Oct 17, 2007
7. ### Jim StewartGuest

My apologies to the original poster and
the group. My meaning was "I have a
stupid question".

Jim Stewart, Oct 17, 2007
8. ### Hans-Bernhard BrökerGuest

Forgive me for being a physicist but: what kind of radiation would that be?
What kind of "isolation"?
Ultimately, you never do. That's why people do stability checks, a.k.a.
calibration runs, where they measure the effect of the experimental
environment on the measurement devices themselves.
I'll second the guess already voiced here before: zero. SDRAM signals
aren't designed to travel on pretty much anything else than short(!),
scrupulously designed traces on multi-layer PCBs.

Hans-Bernhard Bröker, Oct 17, 2007
9. ### sendthisGuest

Well I'm willing to provide ANY clock rate. I'm trying 20 Mhz because
that seemed like a reasonable clock speed to work with. The data sheet
didn't specify a minimum clock speed but I would assume it would have
to be at least fast enough to provide a refresh rate (~100kHz).

sendthis, Oct 26, 2007
10. ### sendthisGuest

I'm exposing it to radiation. And I knew what you meant by "A stupid
question" when I read it, but thanks for clarifying anyway.

sendthis, Oct 26, 2007
11. ### sendthisGuest

When I mean isolated, I mean, "away from". The DRAM must be 19 inches
That's what I originally thought too, but people I asked (who don't
necessarily have experience in the subject) said it seems reasonable
to assume it's okay. We ended up using CAT5e cable for the cabling...

sendthis, Oct 26, 2007
12. ### sendthisGuest

That seems reasonable. I'm using a 20 or 50Mhz clock but I can
certainly slow it down or speed it up fairly easily. The cabling is
made with twisted pair although they're not perfectly the same length.
There's also a short length of it (to be reduced) that's using ribbon
cable.

I guess I'll know soon enough if it works or not since we pressed
ahead with building a test fixture. I'll let everyone know how it
works out...

Thanks,

Eric

sendthis, Oct 26, 2007
13. ### sendthisGuest

Well we've been playing with the test fixture for a while now. It
doesn't appear to be working. Not sure if it's due to skew although
we're compensating that with a PLL. We're also getting some voltage
spikes.

We're going to reduce the wire lengths and see if that helps. If so,
then we'll need to redesign our test. If not, I don't know what to do.

sendthis, Nov 6, 2007
14. ### sendthisGuest

Actually we manage to get it working. There's still some jitter and
ringing but we hope to be able to clean that up with some tweaking.
Thanks for the help.
So if anyone ever needs to run DRAM over 20 inches of wire, use
twisted pair and run it at a 10 Mhz clock... we have had some success
over a 50Mhz clock...

sendthis, Nov 13, 2007
15. ### sendthisGuest

Without termination we got it to run perfectly (passed some basic
memory scans) on 4 inches of twisted pair at 66.66667 Mhz. It fails
somewhere between 66.67Mhz and 75Mhz but it wasn't important to find
exactly when it failed.

On the longer cable, ~24 inches, it passes the memory scan at 20 Mhz
without any type of termination.

sendthis, Nov 15, 2007