DRAM PCB-Layout, works only up to 60 MHz

[lpcware]

Content originally posted in LPCWare by movax on Tue Jan 22 02:47:15 MST 2013
Hi,
I made a double Layer board with an LPC4330 and a 16-bit DRAM. Its working properly up to 60 MHz. Even without series resistors there is no ringing visible on the scope. But the short slopes dont reach the right levels at higher frequencies.
The same code is working on an evaluation board with same DRAM.
Do the traces have too much capacitance? Are the meanders for matching wrong? Is it because of the missing ground layer? Dont know.
I have no experience in routing HF-Circuits and would like to ask for help.

[lpcware]

Content originally posted in LPCWare by torpebaun on Tue Dec 03 15:52:29 MST 2013

Hi movax,

Were you able to find the problem? Could you please share what you found?

One time I used gsm development board, and I had to download the code frequently. When I reset and download once, it would work  as planned for 5 - 6 times and it would stop working after that. Then I would have to reset and download the code again. I got dissappointed and gave up at one point. A development board cost so much and when they don't work according to your plans, it is frustrating. My guess was that the maker of the board is really cheap and they might have used cheap solder and cheap parts.

<a href="http://www.7pcbassembly.com/prototype-PCB-assembly.php">prototype assembly</a>

[lpcware]

Content originally posted in LPCWare by Pacman on Sat Mar 09 09:56:36 MST 2013
What's your status on this; did you notch it up a bit?
If you did, what helped the most ? :)
I'm especially interested in the difference that the resistors make.

[lpcware]

Content originally posted in LPCWare by peufeu on Tue Feb 26 09:48:14 MST 2013
Without a ground plane and proper decoupling, I'm quite surprised that it even works at 60 MHz.
This must emit lots of RF pollution.

[lpcware]

Content originally posted in LPCWare by suckfish on Thu Jan 24 21:59:03 MST 2013
Re:

> But the short slopes dont reach the right levels at higher frequencies

Have you set the I/O pins to high-speed in the relevant pin configuration registers?

If you haven't done it already, try setting bits 5 (EHS, high slew rate) on outputs and bit 7 (EZI, disable input deglitching) on inputs.

Cheers,
Ralph.

[lpcware]

Content originally posted in LPCWare by movax on Wed Jan 23 04:07:18 MST 2013
Thank you very much for the answers. Its very helpful, I will try this.

[lpcware]

Content originally posted in LPCWare by Dave on Tue Jan 22 11:42:57 MST 2013
Here is a good article on routing: http://cache.freescale.com/files/32bit/doc/app_note/AN2536.pdf

[lpcware]

Content originally posted in LPCWare by wmues on Tue Jan 22 06:44:05 MST 2013
1) Forget about using meanders. For one SDRAM, you will not need them.
   Instead, place the SDRAM close to the CPU.
2) Use a series resistor in the clock line, close to the CPU. Try 47-68 Ohm.
3) Separate 3 groups of nets so that the capacitive coupling is minimal:
   a) clock
   b) address & command lines
   c) data lines
4) Use a solid ground plane under CPU, SDRAM and under the traces.
   So all traces will be on the top layer, and ground plane on the bootom layer.

Item 4) is the most urgent. You will have ground bouncing without.

Your oscilloscope might have too high input capacitance and too low bandwith. Try a decent 500Mhz Scope.

I am running a LPC1785 + SDRAM 78 Mhz in Production @ 2 Layers. It works.