jlaitine opened a new pull request, #17411:
URL: https://github.com/apache/nuttx/pull/17411

   ## Summary
   
   This improves the error detection logic for MPFS DDR training code. The 
original error checks were too restrictive, sometimes rejecting good training 
results. The issue was found on a larger production batch of devices, where 
some devices started failing to boot in extreme temperatures.
   
   1) The CA training status vector don't need to be separated by 5 taps; this 
condition was copied originally from Microchip reference code (HSS), but also 
that has since changed. The values are *typically* separated like that, and it 
is an indication of good signal itegrity, but it is not a valid condition to 
reject the training results.
   
   2) Add a check that the window offset + width doesn't exceed the whole 
range. If that occurs, either the training has failed completely, or the window 
is not properly centered. In the latter case, if it is consistent, one can  
apply offset to the window centering (RPC156) instead.
   
   ## Impact
   
   Affects MPFS builds which use NuttX as the bootloader (i.e. don't use HSS). 
This doesn't change the training procedure as such, only affects the pass/fail 
criteria for the extra sanity checks (additional checks to the controller's 
own).
   
   ## Testing
   
   Tested on custom MPFS design, with a batch of > 100 devices tested in 
temperature chamber. Observed some devices getting stuck at early boot in DDR 
training in temperatures below -20C, whereas after changes all work ok.
   


-- 
This is an automated message from the Apache Git Service.
To respond to the message, please log on to GitHub and use the
URL above to go to the specific comment.

To unsubscribe, e-mail: [email protected]

For queries about this service, please contact Infrastructure at:
[email protected]

Reply via email to