Advertisement

Article

Use CFD modeling to analyze and improve PCB thermal design in a multiple-phase voltage regulator

Due to an increase in required power levels and a reduction in available board space in multiple-phase voltage regulator applications, printed circuit board (PCB) trace design has become an important part of voltage-regulator thermal design. The PCB helps dissipate a large portion of heat generated by the voltage regulator (in some cases, it is the only means by which to dissipate the heat). A well-designed trace of PCB improves the thermal behavior of the PCB by enlarging the effective thermal conductivities around the board's MOSFET and integrated circuit (IC) components.

On the other hand, reducing unnecessary traces is also desirable for cost reduction. Therefore, at the design stage, the variation of PCB thermal conductivity around the voltage regulator, and its effect on the voltage regulator's thermal performance, should be evaluated and adjusted to meet the above objectives.

A common approach for PCB thermal analysis is to calculate the average of effective parallel and normal thermal conductivities of the entire PCB based on the number, thickness and fractional coverage of copper layers and total board thickness, then apply the average parallel and normal thermal conductivities to calculate the heat conduction of the PCB (References 1 and 2 ). However, this approach is not suitable when we need to consider localized variation of thermal conductivity in the board.

A thermal-modeling software package, Icepak, was used to study the localized variation of thermal conductivity in the PCB. In addition to its computational fluid dynamics (CFD) function, the software takes into account PCB traces and vias, and computes thermal conductivity distribution over the PCB. This feature makes the software suitable for the following study (Reference 3 ).

Initial PCB design and model validation
An Icepak model was built based on ECAD files from a 1U server application. The initial PCB trace and via information was imported into the model, Figure 1a .



Figure 1a: Imported traces of initial design.

(Click on image to enlarge)

To check the thermal-conductivity distribution, a boundary condition of a constant temperature of 45°C was assigned to the bottom-side of the imported PCB, while a uniform heat-flux boundary condition was assigned to the top side. The calculation results are shown in Figure 1b .



Figure 1b: Temperature of PCB top surface with uniform heat flux, initial design.

(Click on image to enlarge)

In the figure, the high temperature indicates a low thermal conductivity, while the low temperature indicates a high thermal conductivity. It can be seen that in areas without traces, the temperature is high, and in areas with more traces, the temperature is lower. In areas of big vias, the temperature is close to 45°C.

This shows that the thermal-conductivity distribution is consistent with the trace distribution in the initial design. To capture the localized effect of small vias, a small background grid size should be used (Reference 4 ).

In this study, the background grid size is 1 mm x 1 mm. Each grid contains a PCB element that has its own thermal conductivities in X, Y and Z coordinate directions, which usually have different values.

In the model, power losses for the voltage regulator's components and traces are shown in Table 1. These power loss values were verified in the previous test.



Table 1: Power loss of voltage-regulator components.

(Click on image to enlarge)

The 1U application model is shown in Figure 2a with airflow above the PCB. The ambient temperature is 25°C and inlet airflow at 400 LFM. The simulation results are given below. Figure 2b shows the temperature of the board's top surface and components: the components with high temperature being the MOSFETs of the voltage regulator.



Figure 2a: Airflow above PCB.

(Click on image to enlarge)



Figure 2b: Temperature of PCB top surface of initial model.

(Click on image to enlarge)

The simulated results are compared with the test in Table 2 for the maximum temperature of each of key component groups and good agreement was achieved.



Table 2: Component temperature comparison between CFD results and test data
(units: °C).

(Click on image to enlarge)

PCB with reduced traces
The initial PCB design had relatively large trace coverage with the intention of increasing heat spread in the board, to lower the voltage-regulator temperature. However, reducing trace coverage for a cost reduction without using heat sinks is sometimes desired. Therefore, the traces were modified and the above validated model was used to predict the voltage regulator temperature.

Figure 3a shows the distribution of PCB thermal conductivity with modified traces in the same way as in Figure1b. It can be seen that because the trace coverage was reduced, high thermal conductivity is limited in a smaller area around the voltage regulator's components.



Figure 3a: Temperature of PCB top surface with uniform heat flux, modified design.

(Click on image to enlarge)

As a result, heat spread in a relatively narrow region and the maximum MOSFET temperature was increased by 9 degrees. This PCB, with a higher ambient temperature of 45°C, was also simulated and will be used to compare with the example with a CPU assembly in the following table.

Modified PCB with CPU assembly
In this simulation example, the CPU and its heat sink assembly were added in the above model to present their airflow blockage as well as the heating effect from the CPU. To examine the worst case, the ambient temperature was set at 45°C.

Figure 3b shows the airflow distribution for this example. Figure 3c shows the temperature of the PCB's top surface.



Figure 3b: Airflow above PCB with CPU.

(Click on image to enlarge)



Figure 3c: Temperature of PCB top surface (with CPU assembly, Ta = 45°C).

(Click on image to enlarge)

Figure 3b shows that the CPU and its heat sink did not significantly block the airflow which passed over the voltage regulator's components. On the other hand, CPU heat dissipation did not greatly affect the voltage regulator because it took place downstream of it. Therefore, the maximum temperature of MOSFETs was increased by only 3°C, compared with the previous example of 45°C ambient temperature without CPU assembly.

Modified PCB with CPU assembly, reversed airflow
If the wind-tunnel airflow direction is reversed, the impact on the voltage regulator's thermal performance from the CPU will be much larger. Figure 4a and Figure 4b show airflow and temperature distribution for this example.



Figure 4a: Airflow above PCB with CPU assembly (hidden), airflow reversed.

(Click on image to enlarge)



Figure 4b: Temperature of PCB top surface (with CPU, Ta = 45°C, airflow reversed).

(Click on image to enlarge)

Because of blockage from CPU, airflow over the MOSFETs was greatly reduced. In addition, preheated air was brought to the voltage-regulator area. A portion of CPU dissipated heat was also conducted to this region through the PCB. The combination of these three factors raised the MOSFETs temperature by 16°C.

The maximum temperatures of critical components for all above examples are shown in Table 3 . The simulation results show that the modified traces can meet the thermal target of the voltage regulator application while reducing the material cost of traces. However, if the airflow direction is reversed, the MOSFETs' temperature will be a little too high.



Table 3: Maximum temperature of critical components of above cases.

(Click on image to enlarge)

Summary
A 1U server voltage regulator application was modeled using CFD software Icepak. The model was validated through comparison between the simulation results and wind tunnel test data. The voltage regulator application was then simulated with modified traces and added CPU assembly at different temperature levels. The simulation results show that the modified traces are thermally feasible for the intended application.

Acknowledgements
The authors would like to thank Darryil Galipeau for providing PCB ECAD files for the study, and Stephen F. Imms for carrying out the PCB wind-tunnel testing.

References
1. J.E. Graebner, “Thermal Conductivity of Printing Wiring Boards,” Technical Brief, Electronics Cooling Magazine , Vol. 1, No.2, October, 1995, p. 27.
2. K. Azar and J.E. Graebner, “Experimental Determination of Thermal Conductivity of Printed Wiring Boards,” Proceedings, SEMI-THERM XII Conference , March, 1996, pp. 169-182.
3. “Icepak Version 4.3.10 User's Guide,” Fluent, Ansys Inc., Lebanon, New Hampshire, November 2006.
4. Rabindra Paul, “PCB Modeling: Trace Import,” Icepak training presentation, Ansys Inc., Austin, Texas, 2006.

About the authors
Chong-Sheng Wang , Ph.D., is a Senior Thermal/Mechanical Application Engineer at International Rectifier, working on thermal analysis and design for electronic packages, PCBs and systems.

Steve (Xingsheng) Zhou , MSEE, is an Applications Engineering Manager at International Rectifier, Rhode Island Design Center, working on Multi-Phase Voltage Regulator design for desktop and server applications.

0 comments on “Use CFD modeling to analyze and improve PCB thermal design in a multiple-phase voltage regulator

Leave a Reply

This site uses Akismet to reduce spam. Learn how your comment data is processed.