Hello Eric, Keep in mind that DDR4 controllers should support DQ bit training. This means that they should be able to delay each DQ by an amount appropriate to center it with respect to the DQS. So you'd think that you have the whole entire UI to mess around with ISI, Noise, and other frivolous whatnots, right? (of course not!) As you mentioned, the UI is divided up into the chip, package, board and DRAM. Now, the DDR controller and DRAM industries are far more concentrated than the PCB industry. The haggling over where the timing budget is allocated is, well, very democratic - two wolves and a lamb vote on what's for dinner. In your +/-178ps example, that makes me wonder about the 90ps allocated for the PCB. 90mil, maybe. (If you verify 90ps, please do pass on a link - it could be a good controller) Usually, since the DRAM side is fixed by JEDEC, and PCB guys having little voice since they're dispersed and come late into the game, the PCB requirement numbers usually come from whoever you choose as your controller. You pay mucho $ for your controller, you're likely to get a big fat PCB timing margin (and a free dinner from the friendly sales rep). If you're in a low cost consumer biz, and you use the controller your boss got a sweet deal on from the guy he met in NY who also tried to sell your boss the Brooklyn Bridge, well, there goes your PCB timing budget along with that month for debug. Either way, you should probably understand the capabilities and limitations of your controller since that'll be a big variable in your PCB timing budget. Of course, before you fab your board, you can simulate it with your favorite simulator vendor (Mentor Graphics has a fantabulous HyperLynx DDR Wizard which now supports DDR4) to give you that month of your life back, but you still should know the capabilities and limitations of your controller to know whether what you're seeing is good enough or whether the margins are too tight. Those controller timing numbers will likely be the key to explain your PCB requirements. If you want to know the details of why exactly the controller eats up so much of the budget, well your getting a good answer depends on the amount of sales you give the controller vendor. Regards, -Nitin -----Original Message----- From: si-list-bounce@xxxxxxxxxxxxx [mailto:si-list-bounce@xxxxxxxxxxxxx] On Behalf Of eric silist Sent: Friday, August 01, 2014 12:50 PM To: si-list@xxxxxxxxxxxxx Subject: [SI-LIST] source of DDR 4 timing recommendations Can anyone help me understand the source of the pcb routing recommendations I see for DDR4. For instance DQ to DQS timing for a byte lane? I see some papers quoting 90ps or 50mil, but I don't know where those numbers come from. At a basic level I understand that DQS is the clock for the data byte on DQ, And that the total timing budget must be spread across chip, package, board, and dram. If I look at say DDR4-2800 that's a clock of 1.4Ghz with a period of 714ps. So at double data rate I figure each byte is 357ps wide. Then that should mean I have +/- 178ps of total margin that the DQS can move from it's ideal position of being in the center of the. I think :) Maybe that's where the 90ps for the board and 90ps for chip come from. Anyway I'm just having trouble finding a good explanation to help me understand how all that timing margin is calculated for each piece in the chain. I did try to look up DQ-DQS timing in the JEDEC standard and it just says TBD. I've found several docs that give me an idea of what values to use, but none so far that explain how they are calculated. I feel like this is still an SI related question since I'll use this information to setup constraints for my simulations but sorry if it's off topic. Thank you. ------------------------------------------------------------------ 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 forum is accessible at: http://tech.groups.yahoo.com/group/si-list List archives are viewable at: //www.freelists.org/archives/si-list 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 forum is accessible at: http://tech.groups.yahoo.com/group/si-list List archives are viewable at: //www.freelists.org/archives/si-list Old (prior to June 6, 2001) list archives are viewable at: http://www.qsl.net/wb6tpu