The first question was if there is really a need for a new model for each data rate. Eckhard answered that as the results he made have shown, this is really the case. The second question was, whether Eckhard had already compared the behavior of a real device showing pre-emphasis in comparison to the model. Eckhard had not, and he figured out that the problem of [Driver Schedule] is the modeling of normally two different models, one for the main buffer and one for the boost buffer, and both buffers switching to their specific load. Later on in the simulation, the tool will have to put these two buffer behaviors together to get a switching behavior. For the static curves the voltage levels can be calculated manually, but the transient behavior is up to the tool. Dr. Miersch stated that the pre-emphasis in accordance to the application would change, so that not only the data rate, but also e.g. the line length will afford a new model.
IQC - IBIS QUALITY CHECKER
Manfred Maurer* and Christian Sporrer**, *Siemens AG, and **Infineon Technologies AG, Germany
Manfred noted that the idea for this IBIS checker came up about two years ago in the Siemens IBIS Group (SIG). The intention was to improve IBIS model quality. On the SIG web page there are explanations of IBIS, details of what is expected from IC vendors concerning model quality and also the reasons why. Furthermore, on this web page hints and examples about good IBIS modeling can be found. He pointed out that regardless of the simulation tool used, the quality of the models are the main factor for good simulation results.
He continued by mentioning that the first attempt in this direction was the Excel sheet from the IBIS Quality task group. This Excel sheet contains about 80 lines of quality issues (for each model), so for manual checking this will be time consuming. This might be also the reason for the low acceptance until today.
Manfred has the impression (which he got from talks with many vendors and users) that an automation of these checks is necessary. The advantages of the checklist are an increased confidence in the models and better comparability between models. At the moment there are disadvantages for the vendor because it is very time consuming, and for the user there are only restricted consistency checks possible with no possibility of proving the quality info.
He mentioned that almost every company has its own scripts for testing the models, and that it might be better if the same quality checks would be available for all. Furthermore, the quality would have to be improved only once (by the vendor). He proposed one possible proceeding, which is based on the Excel sheet summary of the IBIS Quality task group and on the needs and requests from the SIG.
The concept should set priorities according to the Pareto principle (80/20). The structure should be a modular one, and later on this program could be transferred to the Quality group. He then showed an overview of the structure of the IQC. There will be a configuration file as well as a parameter file. The configuration file will contain less restrictive values for checking parameters, while the parameter file might contain more specific/restrictive values. The output will contain a message file in both ASCII format and graphic format and also a result file with all the accomplished checks. There will be consistency checks as well as curve evaluations and curve checks, but also information about the number of present models and pins will be there.