Minutes, IBIS Quality Committee 16 February 2010 11-12 AM EST (8-9 AM PST) ROLL CALL Adam Tambone * Anders Ekholm, Ericsson Barry Katz, SiSoft Benny Lazer Benjamin P Silva Bob Cox, Micron * Bob Ross, Teraspeed Consulting Group Brian Arsenault * Bruce Archambeault, IBM David Banas, Xilinx * Eckhard Lenski, Nokia Siemens Networks Eric Brock Guan Tao, Huawei Technologies Gregory R Edlund Hazem Hegazy Huang Chunxing, Huawei Technologies John Figueroa John Angulo, Mentor Graphics Katja Koller, Nokia Siemens Networks Kevin Fisher Kim Helliwell, LSI Logic * Lance Wang, IOMethodology Lijun, Huawei Lynne Green, Green Streak Programs * Mike LaBonte, Cisco Systems Mike Mayer, SiSoft * Moshiul Haque, Micron Technology Muniswarareddy Vorugu, ARM Ltd Pavani Jella, TI Peter LaFlamme Randy Wolff, Micron Technology Radovan Vuletic, Qimonda Robert Haller, Enterasys Roy Leventhal, Leventhal Design & Communications Sherif Hammad, Mentor Graphics Tim Coyle, Signal Consulting Group Todd Westerhoff, SiSoft Tom Dagostino, Teraspeed Consulting Group Kazuyoshi Shoji, Hitachi Sadahiro Nonoyama Everyone in attendance marked by * NOTE: "AR" = Action Required. -----------------------MINUTES --------------------------- Mike LaBonte conducted the meeting. - The usual beginning formalities were skipped New items: Bruce Archambeault gave a presentation on "Feature Selective Validation (FSV) For Comparison and Validation of Data Sets" - Anders: On slide 21 the last and fourth from last items have the same characterization "very low to very high" Lance: On slide 9 what does the 0.1 value represent? - Bruce: The values are from a Fourier transform - The low frequency half is the ADM - The high frequency half is the FDM - Lance: On slide 21, low numbers are good Mike: The frequency plots could show the threshold lines - Areas under the curve could be shaded by threshold crossing - Bruce: It may be possible to get the Matlab code that created these - Alistair Duffy in the UK invented FSV - He and Antonio Orlandi may be of help Bob: How are the threshold values determined? - Bruce: To some degree they mimic human judgment Bob: How is the dividing line between ADM and FDM frequency determined? - Bruce: There is usually an obvious (to humans) change - Simulation timestep affects top frequency, which changes the scale - If the timestep is smaller than necessary there will be no energy at the high frequency end Bob: Is this compatible with non-uniform time points? - Bruce: That may throw off the software Anders: The reference and test data must be properly time shifted too - Mike: One technique is to pre-shift for minimum difference area Bob: Is this an approved IEEE standard? - Bruce: Yes, IEEE 1597.1 is published - Draft versions were sent due to copyright restrictions - A best practices document IEEE 1597.2 may be approved soon - They might like to have examples of IBIS FSV use for their website Bruce: FSV can also be used in 2D - For example, for image processing Bob: The description of FSV here differs from we have seen before - David Banas and Roy Leventhal describe measuring specific features - Mike: It was similar to JEDEC slew rate measurements, for example - But using a Fourier transform gives an overview of the entire waveform Bruce: It may be good to use phase as well as magnitude - Tools should be able to look at both - Mike: That makes sense since we are often measuring timing Moshiul: Measured data may be between fast and slow, but not on any corner - Mike: FSV seems appropriate as an overlay metric - Usually an envelope metric is applied to bench measurements - Bob: Sometimes measurements can be scaled to fit AR: Mike post Bruce Archambeault presentation on website Next meeting will be Feb 23 Meeting ended at 12:20 PM Eastern Time.