USE DISCOUNT CODEEXPERT30TO SAVE $30 USD
Velocity difference between microstrip and stripline
michalb , 03-03-2017, 09:06 AM
I was wondering do you guys take velocity difference between microstrip and stripline into account when doing length matching (or delay compensation to be more precise)? I was looking through Robert's projects (imx6 rex and openrex) and his sdclocks are the same length as address/control lines despite differences in routing topology. Clock lines are routed as micrstrip for majority of length while other signals are striplines. Shouldn't velocity difference be calculated and clock lines lengthen accordingly?
robertferanec , 03-06-2017, 02:40 PM
When I route a group of signals which need to be length matched, I use the same topology (so all the signals in the group go through same layers in same order). In some signals (e.g. within important differential pairs) I keep the length same also on particular layers. For most signals in the same group I try to keep the length routed on top and bottom layers approximately the same, but there may be some differences (depends on signal group requirements).
I do not length match signals based on delay. I explained it in some other posts of this forum (basically I have had bad experience with it and the design length matched based on delay didn't work properly and length matching based on delay also limited PCB stackup adjustments). Hope this helps.
I do not length match signals based on delay. I explained it in some other posts of this forum (basically I have had bad experience with it and the design length matched based on delay didn't work properly and length matching based on delay also limited PCB stackup adjustments). Hope this helps.
michalb , 03-06-2017, 03:58 PM
Thank you Robert for your input.
When I mentioned your projects I was particularly thinking about your imx6 rex project and how sdram clock was the same length as address/cmd lines even though routing topologies were different.
But you saying that delay-based length tuning doesn't always work is good enough for me. Thank you.
When I mentioned your projects I was particularly thinking about your imx6 rex project and how sdram clock was the same length as address/cmd lines even though routing topologies were different.
But you saying that delay-based length tuning doesn't always work is good enough for me. Thank you.
robertferanec , 03-07-2017, 08:41 AM
It works oki. These memories are not so fast (533MHz).
PS: Many times the rules in design guides include some tolerance or they include some other considerations. If we would like to go into details, we may need to consider also the things like difference in signal speed between single ended and differential signals. But in some designs it may not be necessary - the influence for specific frequency and length may not play significant role.
PS: Many times the rules in design guides include some tolerance or they include some other considerations. If we would like to go into details, we may need to consider also the things like difference in signal speed between single ended and differential signals. But in some designs it may not be necessary - the influence for specific frequency and length may not play significant role.
michalb , 03-08-2017, 04:57 AM
thank you Robert, helpful as always!
Use our interactive Discord forum to reply or ask new questions.