Simulation techniques help cool the calibration head for the world’s fastest real-time oscilloscope.
Oscilloscopes display and measure the wave shape of an electrical signal. High-performance oscilloscopes, which are capable of measuring signals at very high frequencies, are primarily used in high-speed serial communications, radio frequency/radar/aerospace and high-speed physics applications. Agilent Technologies, Santa Clara, Calif., recently released the Infiniium 90000 Q-Series oscilloscope, which is the world’s highest bandwidth commercially available real-time oscilloscope and the first to reach the 60-GHz barrier.
As part of the design of the Infiniium 90000 Q-Series, Agilent Technologies determined they needed to develop a new electrical calibration source to ensure the oscilloscope could set a new standard for measurement accuracy. This design of the calibration source (Agilent N2806A) involved challenging electrical, mechanical and thermal requirements. Agilent Technologies utilized ANSYS CFX computational fluid dynamics (CFD) software to model the environment and produce a design that could exceed all of the requirements. Agilent was able to deliver a first-pass success on the design. However, a number of challenges were faced in the process.
The major obstacle in designing the calibration head package was cooling. The circuit for the head is packaged in a machined aluminum base, which includes a heat sink and cavities for mounting the itegrated circuits, and needs to dissipate 3.2 W with 5,250-mm2 surface area. Because of the small box size, the energy released by the circuits produces a considerable temperature increase, which can adversely affect handling comfort. To address this cooling challenge, the engineers tried a cross-flow heat-exchanger approach and turned to ANSYS CFX to model and simulate this configuration. They first performed an airflow-only analysis to determine pressure drop, then followed with a heat-transfer analysis to determine temperature rise. The design created with the aid of simulation worked perfectly when production parts were built.
Evaluating alternative cooling designs
Before employing simulation, the engineers made a quick calculation to determine if the unit could be cooled with free convection using a formula to determine temperature rise in a package. In this case, the rise was predicted to be 86 C, much higher than the design specification of 15 C. Forced air was needed to cool the head, but what type of forced-air cooling would provide the best results? The fan-exchanger configuration normally used positions the fan on top of the head, blowing air downward onto the heat sink and exiting around the sides of the unit. This approach requires a large package height, and the connectors need to be positioned toward the edge of the unit, which was not compatible with the front-end connector placement of instrument.
The engineers then looked at the alternative of a blower/cross-flow heat-exchanger design in which the airflow is perpendicular to the face of the heat sink. In this application, the inner walls of the case form a curved channel that directs the air around the top of the heat sink. This approach offered the advantage of reduced height requirements and enabled the connectors to be centered on one side of the calibration head. Since the wrapped flow configuration was previously unproven, access to simulation was critical in optimizing the design.
Simulating airflow and heat transfer
The primary concern of the cross-flow design was that the flow rate of the blower would be reduced due to the pressure drop associated with redirecting the airflow over the heat sink. The blower manufacturer provided a fan curve showing the flow generated at any particular pressure drop, but the pressure drop through the calibration head was initially unknown. The team created a proposed design and used CFD to predict the pressure drop at various flow rates. They reasoned that the pressure drop could be reduced by increasing the heat-exchanger channel area, so they created a second simulation model. Simulation results showed that the second configuration did indeed create less pressure drop, resulting in a higher flow rate sufficient to cool the calibration head.
To ensure that this cooling airflow would be sufficient, engineers then built a heat-transfer model with 1.5 million tetrahedral elements. The integrated circuits were configured as heat generators, and the conductivity of the housing was based on aluminum. The airflow predicted by the first simulation was used as a mass flow input. The heat-transfer simulation took about 15 min to solve on a personal computer; it showed that the case temperature rise was well within the 15 C design specification.
Simulation results match experiment
Agilent Technologies engineers went one step further by constructing a cross-flow prototype using on-hand parts. Included were a similar heat sink from another project, a blower and a plastic duct. The prototype was constructed to match the latest design configuration as closely as possible. Physical measurements showed that the heat sink rose in temperature by 9.5 C. The engineers modeled the prototype in CFX and ran a simulation to validate the simulation method. The simulation predicted a temperature rise of 9.5 C, a perfect match with the physical measurements. Minor differences in pressure drop and flow between the simulation and physical measurements were within the margin of error of the measurements. These results helped to build confidence in the simulation methodology.