How Accurate is Today’s Simulation? Print E-mail
Written by Patrick Carrier   
Friday, 03 May 2013 16:49

Results of an interconnect test characterized using 2-D and 3-D field solvers.

Simulation is a key part of any high-speed new product design. What was once exotic and necessary only for a tiny percentage of designs is now routine in most designs today, thanks to both decreased cost and much better usability. One question that often arises is, How accurate is simulation? Furthermore, to determine simulation accuracy, we must also ask, How accurate are measurements? Every measurement device has intrinsic errors, as well as errors introduced by its presence in the circuit being examined.

When is simulation appropriate? The nature of engineers is contrary to accepting claims on blind faith, so I set out with a colleague to illustrate the type of accuracy that is readily available in today’s tools. Before looking at the accuracy itself, however, it’s important to understand when simulation is appropriate, and when measurement is important.

Simulation is used early in the design process. Signal integrity and power integrity analyses often begin even before schematic entry is complete. To be valuable at this early stage, care and time need to be spent ensuring all components have been properly modeled. At multi-Gigabit data rates, quite literally, modeling everything from basic I/O down to microscopic characteristics of the copper on the printed circuit board becomes important.

The primary advantage to simulation is to determine potential problems with the design before any capital is invested building prototype software. The costs of changes at this early stage are minimal compared to further into the design process. In addition to finding problems, alternative design ideas can be tried in simulation.

Once the design has been built, actual measurements (where possible) are used to verify the performance and guaranteed operating characteristics of the product. At this level of sophistication, a good deal of experience to make accurate measurements.

Test setup. We used a test board supplied by a third party (BetheSignal.com) for simulation and measurements. Simulations were performed on the circuit design using Mentor 's HyperLynx software. The measurements were made using a Teledyne LeCroy digital oscilloscope. S-parameters were plotted using the output of the simulation. Then, the actual measured S-parameters were superimposed on that data to show the accuracy (Figure 1).

Up to 11GHz, the plots are virtually the same. At higher frequencies some difference arises, but up to 18.5GHz, none are greater than about 4dB. These results validate that simulation can provide extremely accurate correlation with measured values well above 10GHz.

Interconnect simulation. Next, eye diagrams were simulated and measured to determine how well they correlated. An eye diagram is a succession of bit patterns laid upon one another. As long as the eye is open to the required specifications, the circuit will work. If the eye closes, meaning that some bit patterns were not successfully transmitted within required signal parameters, the circuit may not work and should be modified to ensure an open eye.

Figure 2 shows the results of the simulation eye diagram on the left, and the measured eye as seen on the oscilloscope. The results are essentially identical.
In the simulation, the interconnect was characterized using a combination of 2-D and 3-D field solvers, as well as S-parameter models for the connectors between boards. The 2-D field solver was used to characterize the traces on the board; the 3-D field solver was used to characterize the vias. The matching results show that the simulated interconnect models were indeed accurate, as well as combined correctly in the simulation, eliminating interconnect modeling as a source of error in the correlation.


Is it accurate? While this was by no means an exhaustive study, it does illustrate that today’s simulation tools have ample accuracy to handle virtually any designs put to them. Keep in mind that each model introduces some finite but very small error into the simulation. Therefore, the greater the number of models, the greater the error that is introduced. However, today’s tools, models and accuracy are all on par to deliver dependable, accurate results.

Patrick Carrier is product manager for high-speed PCB analysis tools at Mentor Graphics (mentor.com); This e-mail address is being protected from spambots. You need JavaScript enabled to view it .

Last Updated on Friday, 03 May 2013 19:38
 

Search

Search

Login

CB Login

Language

Language

English French German Italian Portuguese Russian Spanish
 


Printed Circuit Design & Fab Magazine on Facebook