Below are the questions asked during the event, along with their respective answers.
Q: How difficult is it to use HEEDs to drive a Flotherm analysis?
A: HEEDs supports many applications, including Simcenter Flotherm, through the use of portals specific to the application software.
Q: How does predicting PPM failure rate help with product quality?
A: In the Context of DFMEA, the ability to predict PPM level failure can help to modify a design to have a lower PPM, which will lower the Risk Priority Number.
Q: Do you have data that shows how much simulated value is close to the actual (experimental) value?
A: The model was not calibrated against experimental values. Ideally, the Simcenter Flotherm model would be calibrated against Simcenter T3STER measurement before quantifying the various uncertainties.
Q: What degree of defects or number of defects do you expect in a real 25 micron thick TIM?
A: I can’t provide a definitive answer. Comparing a model against Simcenter T3STER measurement structure functions would identify if TIM thermal resistance was performing below expected values.
Q: Are there simple ways to get Flotherm training to get a job using this with low cost? I have done thermal design by hand early in my career in companies that did not have this software. How do I best get this without prior experience?
A: I recommend checking out our online virtual lab, a web-based 30-day eval.
Q: How do you decide max operating junction temperature?
A: As a system integrator the maximum operating temperature would be provided on the datasheet.
Q: Did you try a defect shape other than a rectangle? Could you have varied it under HEEDS?
A: HEEDs can modify any input. Any restriction on topology would be due to the analytical tool. In our study, we only considered rectangular defects.
Q: Do you model void as air gap? Would radiation/convection play any role in void?
A: We did not scale our defect conductivity as low as an air void. While convection and radiation would occur in the void it was a heat transfer mode we didn’t consider.
Q: Regarding TIM1 thickness 25um, is there any concern about package level reliability? especially stress to die corner area due to package warpage?
A: There are concerns, especially for larger die sizes.
Q: Were the input ranges for each part determined using the tolerance values listed on the vendor datasheets? What if there is only partial information?
A: We did base our ranges on tolerances but not on specific datasheets. We based it on was is typically seen in the networking industry. Engineering judgment played a large role in our determination.
Q: Isn’t die curvature or deformation with temperature an issue? Why is this not a variable in the model?
A: It would be an issue. We bundled all thermally related variants in terms of effective thermal conductivity.
Q: How were the successful Simcenter Flotherm runs (154) determined? And (criteria wise) why were the other runs unsuccessful?
A: The simulations were tagged by HEEDs as unsuccessful when the void position and size violated our position constraint. This occurred when the void would be beyond the TIM boundary.
Q: To model the defect, did you use a cutout in Flotherm? Or a cuboid with an air property?
A: We used a cuboid with varying thermal conductivity.
Q: The FloTHERM software is based on Finite Element method?
A: Simcenter Flotherm is based on the Finite Volume Method.
Q: For CFD model to Conduction only model simplification, do you need to verify with different surface temperature cases? Because the convection heat transfer coefficient is temperature related?
A: The convection in our model was dominated by the forced airflow across the heatsink.
Q: How do you model the Printed Circet board? do you model it as multi sheets or one sheet with calculating thermal conductivity?
A: In this model, we considered the PCB as one object with orthotropic conductivity. The conduction to PCB, convection from PCB accounted for only a small portion of the heat input.
Q: Were the 1000 DOE experiments run in the Simcenter flotherm command center?
A: Command Center was not used for the DoE. HEEDs was used and used a DoE approach that provided the best surrogate of Die maximum temperature.
Q: Is there a method to automate varying parts of a design directly in Flotherm without using HEEDs (ex. TIM size)?
A: Every variation considered in HEEDs could have been done in Command Center. HEEDs was used for its robustness in managing 100s of simulations, surrogate, and Monte Carlo modeling.
Q: How would you calculate the thermal conductance or heat transfer coef. value between a chip and PCB
A: These values are direct outputs from Simcenter Flotherm. Generally, these sorts of values can be determined by surface temperatures, the amount of heat transfer, and geometric information.
Q: Can we assume a ~1.15% max error with the 2250 ppm since a conduction only method was used and not a CFD method?
A: Our errors were associated with using an uncalibrated model, descoping to conduction only, and using a Kriging surrogate that had a non-zero cross-validation error.
Q: How was the vapor chamber modeled? A cuboid with high conductivity, such as 1000 W/mk?
A: The based was modeled with an outer Copper layer, inner wick layer (15 W/mK), and vapor core (10,000 W/mK)
Q: Is it possible to model PCB with Gerber trace layers in Flowtherm?
A: Simcenter Flotherm supports ODB++ FLOEDA file for routed PCB representation. FLOEDA files are created by using a Simcenter Flotherm plug-in in the PCB layout tool.
Q: Can heatsink pressure, improve some conductivity of TIM1, as it is defect there?
A: Yes, increased pressure reduces the thermal resistance of many interface materials but it is important for the pressure to not exceed the package requirements.