Ravinder, What was your measurement technique for determining the various timing measurements? I'm interested in knowing what type of probes were used, where, and did you de-skew your probes? Without knowing the background it's hard to spot what the exact cause is. Thanks, Bill > -----Original Message----- > From: si-list-bounce@xxxxxxxxxxxxx > [mailto:si-list-bounce@xxxxxxxxxxxxx]On Behalf Of Ravinder Ajmani > Sent: Monday, May 06, 2002 6:46 PM > To: si-list@xxxxxxxxxxxxx > Subject: [SI-LIST] DDR DRAM > > > > Hi SI Experts, > I am using DDR DRAMs for the first time in our design. The > design uses a > single x32 DRAM in point-to-point topology. Due to space > constraints, I > have used only source terminations on data bits at the DRAM end. My > simulations show that DRAM drivers are stronger than my ASIC > drivers, and I > am able to get reasonable signals in this configuration. > > The first version of the card failed to work with DRAM from > Manufacturer > "X". We found three data bits failing when they were > switching high to > low, and their adjacent bits were switching low to high. We > tried DRAM > from Manufacturer "Y" and it seemed to work fine. This card > had long DRAM > data lines (about 4" long), so I thought that the board level > crosstalk was > an issue. Another card, which had DRAM lines less than 2" > also showed same > problem with DRAM "X", but worked fine with DRAM "Y". A > further analysis > of ASIC input timings and DRAM output timings show the following: > > 1) Our ASIC has very little timing margin for the three failing bits. > > 2) The worst case Data Strobe to Data delay for DRAM "X" > varies from 36 ps > to 382 ps for different data bits (measured at the DRAM > output), which is > still within DRAM vendor specs. However, at the ASIC end, > the same delay > increases to 300 ps to 670 ps for worst case switching. > > 4) Interestingly, DRAM "Y" appears to be putting Data out BEFORE Data > Strobe, ranging from 100 ps to 400 ps. > > Placing parallel termination at ASIC end (680 ohms each to > 2.5V and Ground) > seems to help the working of DRAM "X". I have further > separated the DRAM > data lines to the extent possible. The length of Data Strobe > and Data bits > is matched to within 10% (it is very difficult to improve > upon this due to > tight routing space). The simulations on the new card layout > show very > little crosstalk (less than 50 mV) with worst case switching. > > Our intention is to use both DRAM vendors. While we are > working to improve > the ASIC routing (in next revision) to improve timing margin for these > failing bits, I will appreciate your suggestions for making > improvements to > board layout. > > Regards, Ravinder > PCB Development and Design Department > IBM Corporation > Email: ajmani@xxxxxxxxxx > ************************************************************** > ************* > Always do right. This will gratify some people and astonish the rest. > .... Mark Twain > > > ------------------------------------------------------------------ > To unsubscribe from si-list: > si-list-request@xxxxxxxxxxxxx with 'unsubscribe' in the Subject field > > or to administer your membership from a web page, go to: > //www.freelists.org/webpage/si-list > > For help: > si-list-request@xxxxxxxxxxxxx with 'help' in the Subject field > > List archives are viewable at: > //www.freelists.org/archives/si-list > or at our remote archives: > http://groups.yahoo.com/group/si-list/messages > Old (prior to June 6, 2001) list archives are viewable at: > http://www.qsl.net/wb6tpu > > ------------------------------------------------------------------ To unsubscribe from si-list: si-list-request@xxxxxxxxxxxxx with 'unsubscribe' in the Subject field or to administer your membership from a web page, go to: //www.freelists.org/webpage/si-list For help: si-list-request@xxxxxxxxxxxxx with 'help' in the Subject field List archives are viewable at: //www.freelists.org/archives/si-list or at our remote archives: http://groups.yahoo.com/group/si-list/messages Old (prior to June 6, 2001) list archives are viewable at: http://www.qsl.net/wb6tpu