1.0 Executive Summary
1. Results of both thermal model simulations performed by Qsigma Associates and thermal testing performed at the client’s site have proven the Xilinx and DSP devices can be conduction cooled to the outer cover to meet manufacturer’s recommended maximum operating temperature specifications while operating at approximately 6.1Watts inside a Alpha Pack horizontal plastic case at room temperature.
2. Thermal management improvements directly translated into a gain in the product’s transceiver performance of 3 to 5 db. This improvement in performance and reliability can be achieved at a reasonable cost considering it is a retrofit solution to an existing hardware.
3. No external vents, heat pipes, or other external modifications to the ABS plastic case are required to meet the performance specifications.
4. Modifications to the Alpha Pack ABS plastic case were made by epoxy bonding a thin copper sheet heat spreaders to both inner surfaces of the plastic case using thermally conductive adhesive. A compressed thermally conductive gap pad material was attached both above and below the Xilinx and DSP devices. Device heat is conduction transferred to the inner surface of the plastic case where it is then transferred by convection and radiation to the external ambient temperature.
5. A baseline thermal model was created and checked out as a fair representation of the measured unmodified system temperatures. A series of suggested design improvements were tested on the model before actually being tested on hardware in the lab. Minor iterations were made to obtain a correct component stack up to compress the thermal pad until a final acceptable solutions was achieved. Measured case temperature of the Xilinx XVC1600E
Virtex BGA was reduced from an initial value of (85 -90 C) to a final value of 63 C for the same room ambient temperature conditions or a minimum drop in operating temperature of 22 C.
The client’s prototype Alpha pack design consists of 6” x 4’ x 0.8 ” ABS plastic case having a material thickness of 0.070 inches. Mounted on stand-offs in the center of the case was a 3” x 5 1/2” standard G-10 circuit card having both digital and RF sections. Circuit board heat dissipation was approximately 6 Watts with the majority of the heat (2.1 Watts) dissipated in a Xilinx fine pitch BGA. ABS plastic is a very poor heat conductor therefore the circuit card, devices, and surrounding internal air temperature reaches excessive temperatures in less than one hour of operation. The Alpha Pack also contained a battery back up section that produces some heating but was not considered for this analysis.
The purpose of this study was to determine if a thermal design modification could be made in a one week time frame and also cost effectively meet system transmission performance goals.
A baseline thermal model was created in the ICEPAK CFD software package to predict the plastic case, circuit board, and component temperatures during typical room temperature operation.
Figures 1 through 4 describe an ICEPAK thermal model created for the baseline alpha prototype and proposed thermal management modifications.
Figure 1. Alpha Baseline Thermal Model
Figure 2. Model for Baseline Alpha Pack
Figure 3. Baseline Model Shown Modified to Include a Copper Sheet Heat Spreader
Figure 4. Final Model Includes Two Copper Spreaders and Two Interface Pads
3.0 Analysis Results
Thermal model temperature contours for the baseline alpha are shown in Figure 5 below. The baseline model predicted an internal air temperature inside the case of 52 C and a Xilinx case temperature of 63 C. The final model simulation with design modifications is shown in Figure 6. The final model configuration predicted an internal case air temperature of 41 C and a Xilinx case temperature of 48 C for a lowering in the case temperature of 15 C. Actual measured drop in case temperature was 22 C and internal air temperature in the case was 39 C for a temperature drop of 13 C. Although at first this improvement may seem minor the modifications did produce remarkable improvements in the system operating performance. Several thermal tests to obtain statistical validity of the modifications suggested by Qsigma Associates have been performed and client test results are shown in graphs 1-4 that follow. Qsigma’s suggested hardware modifications are now successfully being implemented by the client.
Figure 5. Baseline Thermal Model
Figure 6. Final Model Predictions with Two Copper Spreaders and Interface Pads
This PDF file from the Xilinx web site recommends heat spreaders and passive heat sinks when operating the Xilinx BGA in the (2 to 6 Watts) range.