Data Input

Step input from scope

• Initially a single step response time record will be requried
• For more complex transmitter signals the use of a penrose matrix inversion can lead to multiple step responses representing the transmitter, however, I would suggest we leave this for now
• The support of simple step responses is already supported in the XML definition language of v4.2.2. The support of multiple steps is open.
• more details of this interface are discussed here

S-Parameters

• all the features from v4 (8×8, mapping, cascading, interpolation, ifft), currently all available in the GUI and XML
• wizard for SATA, clearly is a new feature, and some .net features have been tried out already by Edotronik
• investigation of padding requirements and bandwidth requirements for 25Gbps under investigation, needs additional option in XML and GUI

Language Support

• python suport for all necessary machines OS done
• use of numpy for all necessary math functionality done
• support of necessary graphical output using matplotlib almost complete

Jitter Definition

• Pulse Width Shrinkage support using time correlated edge jitter shall be initially defined in terms of a simple bimodal peak to peak jitter.
• Further definition of a time correlated jitter is under discussion with the Technische University München
• Mid band jitter shall be calculated as in v4. The engine shall accept as an input a histogram description of the jitter distribution. (Higher level functionality in the GUI or scope should be responsable for conversion of RJ/DJ to pdfs.

Coding and emphasis

• All combination are theorically possible, however, each combination needs a new state transition definition therefore a limited number of combinations should be implemented, exact combination to be define. GUI and XML should be changed to
• cover equalisation using a single object as opposed to a transmitter and receiver object (exact details under discussion, e.g. interpretation of old xml files)

• 8b10b; discussion still ongoing concerning exact state transition table
• nrz; no problems
• duobinary; needs pre-coder, however exact coder/decoder not clear
• pam4; including pre-coder clear, however, decision level for receiver not clear. No PAM-n.
• lte; adaption algorithm in v4, as per OIF, was very unsatisfactory, method should be done using convolution of the step response and adaption as below
• dfe; as per v4. Support for PAM-4 or duobinary currently unclear
• fir (de-emphasis); as per v4, however, not all combinations with other schemes clear
• current optimisation using simplex algorithm and eye opening may not be enough, therefore move to more theoritical zeroing of partial or full cursors in step response may be preferable. Would allow faster analysis, and maybe more stable results. Crosstalk ISI should be included using a peak distortion minimisation.
• all equalisation shall be implemented as analytic solution directly on the pulse/impuse response, i.e. not looping the stateye engine

Some additional requests

• padding for ifft should be function of the fbaud
• can the time interpolation be removed, as we now have such a large fmax?
• can fractional fir be supported
• for DFE/CDR analysis can the edge also be well analysed
• does FIR @ the Tx = FIR @ the Rx