Hi Robert,
Your
Schematic & PCB Design Course mentioned the T branch topology and the fly by for the DDR3, I have some details as I have to route 2 DDR3 on board.
The PCB is 6 layers and one of my concerns is the propagation time for point to point lines as for my worst case I got 55ps difference between DQS and a data line, DQ routed on external and worst case on internal layers. (I used Saturn calc for propagation time on track & via)
+Is this time difference too much or the DDR3 can handle this difference? MT41K512M16HA-107 :A
For the length match for Address & control lines:
I first did it to the first DDR3.
then to the second DDR3 removing the stub to get the real distance from the Up to the second DDR3.
Then I’ll be removing the stubs from both DDR3 and length matching from the Up to the terminators.
+ I would like to have your comments about my procedure to length match the segments.
Still I need to verify the worst case for daisy chain signals (add,clk,etc), my guess is that will be very close to the 55ps also as I used internal layers form DDR3-A to DDR3-B.
I used Saturn to make the calculations.
The tech note said to daisy chain the DDR's.
I also know which are the best practices, but engineering bought an evaluation board which was routed in 6 layers.
I want to take everything in count to make it work in 6 layers.
I have the stack to have 50 ohms in each layer also 100 for diff pairs.
Thanks