All,
IBIS leaves to the EDA tool the analysis of the Impulse Response output of
AMI_Init and the waveform output of AMI_GetWave (when clock ticks are not
generated). One of the complications is that a single Tx or Rx I/O buffer may
support multiple standards which have different eye quality metrics, data
encoding techniques, …
At SiSoft/MathWorks, we associate analysis rules with the channel (we call
channels with the same I/O buffers and rules “Transfer Nets”).
I would like to propose a new AMI Reserved Parameter “Analysis_Method”:
(Analysis_Method (Type String) (Usage Info) (Value|List <one or more string
parameter>)
(Description “<sufficient description on how each
Analysis_Method should be analyzed by the EDA tool”))
It would be preferred that the Analysis Method be a pointer to some industry
standard (e.g. 802.3bj). The point is that a user should be able to configure
his EDA tool to analyze a channel in accordance to the Analysis Method
requested. There is no reason why we cannot have some reserved Analysis Method
names.
One of the focused issues in this paper is where to sample the data in a
statistical eye. This can be included in the Analysis_Method above, or we can
opt to have another reserved parameter Clock_Sample_Time:
(Clock_Sample_Time (Type String) (Usage Info) (List “Center_e-3_Contour”
“Center_DDR4_Mask” “Center_Zero_Crossings”)
(Description “Method used to determine location of clock sample
time in statistical analysis”))
This is a BIRD that I could write or support.
I disagree with the suggestion on slide 21 is to have the AMI model provide the
results directly. Many of our customers use IBIA-AMI simulations to select Rx
Buffer IP. How can we compare the methods that two different IP vendors
implement in their DLLs.
Walter
Walter Katz
Work 508.647-7633
Cell 720.417-3762
[Description: Description: Visit MathWorks.com]
Attachment:
bermensolo.pdf
Description: bermensolo.pdf