At these velocities the flow through the cooling plenums is turbulent and may lead to flow-induced vibration of the MCM cards in their Rohacell mounts. There is concern about (1) possible vibration-induced noise in the electronics and (2) possible gradual loosening of the MCM cards as mounting groves in the Rohacell deteriorate (see Sec. 3).
Another potential problem not discussed in Sec. 2, and not yet evaluated using COSMOS Flowplus, is thermal stratification of the flow in the cooling plenum because of buoyancy effects. That is, as the air traverses the cooling plenum and is warmed 10 C, the less dense, hotter air will move to the top of the plenum contributing to the temperature gradient across the MCM's.
It was found that by adding a 0.2-mm-thick aluminum layer on the base of the MCM's to spread the waste heat, the temperature non-uniformity across the MCM's could be reduced from 23 C to 6 C (see Sec. 2).
While increasing the mass flow rate of air between MCM's by increasing the spacing between boards was evaluated, it was found that the maximum MCM temperature decreases only slowly as the spacing between boards is increased.
We observe that an obvious way to lower the air temperatures along the plenum, and thus reduce the velocity of air required to cool the MCM's, is to lower the plenum inlet air temperature from 20 C to a lower value. This could be accomplished using a water chiller coupled to the air loop through a water-air heat exchanger. This approach has not been studied or designed.
In the event that the difficulties associated with an air cooling system outweigh the problems associated with a water cooling system, a water cooling system could be designed.
Gravitational displacement. It has been determined that using the current Rohacell C-module design will lead to acceptably small deflections and stresses because of gravitation forces (see Sec. 4).
Hydroscopic expansion. From finite element modeling, it was found that the expansion of Rohacell foam in the C-modules because of moisture absorption will cause distortions in the support framework, and high stresses in the area of the foam-silicon glue surfaces (see Sec. 5). It is noteworthy that even small changes in humidity may result in values of stress that exceed the tensile stress of the foam. Without additional design considerations, this result poses serious concern.
A full-scale prototype air cooling system should be set up in the lab using resistance strip heater to simulate the MCM heat loads, and that measurements of air and exit MCM surface temperature be made to insure that the cooling is adequate. These measurements can be made with or without a chiller to lower the plenum inlet temperature. Preferably, the test will include actual MCM's with aluminum plate heat spreaders. These tests will confirm the required plenum air velocity, flow patterns, and confirm or refute thermal stratification.
At the required plenum air velocity, it is recommended that tests be performed on the prototype to evaluate whether vibrations and resonance are expected to be a problem. While accelerometer measurement are useful, these date depend on the laboratory set-up and plenum mounting configuration. The level of g-loads may or may not reflect the actual PHENIX installation, depending on how much effort is made to duplicate the real configuration. For this reason, it is suggested that the internal flow itself also be characterized by pressure and velocity (turbulence) measurements.
Measurements of pressure and velocity fluctuations using pressure transducers and hot-film anemometer equipment are advised. Pressure power spectra will be used as input to the COSMOS plenum structural model to predict displacements. Both pressure and velocity rms (turbulence intensity) levels and power spectra will be used to evaluate the effects of design improvements.
It is suggested that care be exercised in the design of the cooling loop, for mistakes in this design can produce high turbulence levels; on the other hand, the employment of fluid mechanics design tricks can significantly lower turbulence levels. In addition, attention should be paid to the MCM configuration down the plenum to produce a uniform heat load across (transverse in ) the plenum.
A computational fluid dynamics model (COSMOS Flowplus) can be used to evaluate thermal stratification in the plenum flow (see Sec. 6).
In the event that the difficulties associated with an air cooling system outweigh the problem associated with a water cooling system, and alternative water cooling system should be designed and tested.
Hydroscopic expansion. Since it was found that the expansion of Rohacell foam in the C-modules (because of absorption of small amounts of moisture) will cause distortions in the support framework, and high stresses in the area of the foam-silicon glue surfaces, several tests and design ideas are suggested. These tests and ideas are discussed in detail in Sec. 5 and summarized here.
A desirable test is to bond a Rohacell frame around a single silicon strip and expose it to cycles of humidity change to check for cracking, separation, or tearing of the foam.
An accurate stress-strain curve for Rohacell should be determined to obtain Young's modulus and the elastic limit for the foam.
A more careful determination of the moisture coefficient of expansion of Rohacell should be made. In particular, the expansion for fully stabilized (equilibrium) conditions should be measured as well as the time-dependent behavior. This determination should be made with and without coatings.
Silicon-to-foam bonding and attachment methods that allow for a flexible bond or for relative motion between the silicon and foam should be investigated with a view toward mitigating the potential moisture-expansion-induced distortion and stresses.
Alternate structural material. in the event that the difficulties associated with implementing a Rohacell foam structure are deemed to be too risky, a redesign of the silicon structural supports utilizing more conventional, non-hydroscopic materials such as low-density, graphite-epoxy laminates or thin, sheet aluminum should be initiated as soon as possible.