Final Design

There were two designs that were developed for this project. There is an ideal design which is co-fired but has a lead time of around 8-12 weeks. As a result, an alternative design was created for intermediate testing concurrently with the ideal design. The first solution uses a thick-film co-fired metalized aluminum nitride substrate with gold plated tungsten internal metallization.

The second solution involved fabricating the same material with the exception of the circuitry layer (made from glass fiber printed circuit board to speed up production) and a support bracket that is made from Teflon. Teflon is an extremely soft plastic that is non-metallic but is not a very good thermal conductor. For the sake of the alternative design this was not the ideal material to use but could still be used for intermediate testing. This secondary solution does not use the co-fired process that the “ideal solution” did, which significantly reduced lead-time to about 3 weeks (1 week for delivery and 2 weeks for fabrication). This solution was tested at SPAWAR and was found to work but not go all the way down to temperature (went to 9K). The reasons for this bracket not going all the way down to the needed temperature are described in Chapter 4. Both of these designs are discussed below in more detail.

Ideal Design

The final design of the Microchip stand incorporates the best qualities of each design solution, with an emphasis on reliability and performance. As shown in the figure below, the basic static design is that of a 90° L-bracket, the bottom of which is composed of 2 types of layers. The Circuitry Layer is layered with electrical traces which carry the microchip signal to output pins. The chip makes electrical contact on the Circuitry Layer by the buildup of gold that is directly on the exposed ends of the circuit traces. The microchip sits within a square depression in the Restraining Layer, and is held in place by a removable bracket (Fastening Plate) to be screwed in above it. The Restraining Layers function is make sure that the chips pads and the bumps are lined up correctly without creating a short circuit. The Restricting Layer was designed to have the chip just be able to protrude out of the top when it was placed in so that the Fastening Plate would make good thermal contact. By placing thermal grease on the surface of the Fastening Plate that makes contact with the chip and the bracket, thermal contact is maintained.

Prototype Design

Instead of all the layers and the tungsten being co-fired into one bracket with internal circuitry, the concept design is held together by nylon screws and nuts. This cut down the turnaround time to two weeks maximum to get all the materials that were needed to manufacture the part. This also allowed for strict control of tolerances that are necessary for this project.

This design also got rid of the buildup of metal on the ends of the exposed traces. Because of the way that the circuitry was made, the chip could be placed directly on to the traces and maintain good electrical connection. This design is shown in the firgure below. However, if the chip fails to make a strong connection to the PCB, an alternative solution would be to a precision buildup of metallization onto either the ends of the exposed traces of the PCB or the Chip to create the necessary connection.

Co-Fired Design

The Section Cut below shows how the chip sits directly on the circuitry. The Fastening plate is what makes thermal contact with the chip by pressing down onto it. Even though the chip and the Fastening plate are really flat, they don't match up directly so the use of thermal grease is implemented on the bottom side of the Fastening Plate that makes contact with the chip. This thermal grease makes up for any small gaps that might occur.

Prototype Design

Two of the big drawbacks of this design was the use of Teflon for the Joining Bracket and the Restraining and the use of PCB. Both of these materials are poor thermal conductors when compared to the aluminum nitride (AlN = 140-180 W/m*°K, Teflon = 0.23 W/m*°K, PCB has a better thermal conduction coefficient than Teflon but it is still verry low). This creates a longer time for the bracket to cool down. The added mass coupled with the change in materials and the fact that the bracket wasn’t co-fired increased the risk that the chip would never reach the desired 4K (7.2°R).

Another thing that was different about the two designs is the layout of the circuitry. The company that made the PCB required that all the traces be at 90 or 45 degree angles. For ease of modeling and manufacturing, the traces were made with 90 degree angles. Unlike the Ideal design the traces could be a 90 degrees without causing a possible spike in resistance. This was assumed to have no effect on the predicted performance of the Prototype. A figure of the PCB is shown below with the Chip outlined in red and the circuitry outlined in black.

Section Cut

The below figure shows how the circuitry is laid out. To makes sure that the traces will have the maximum thickness that was called out in the drawing, the traces were made not to have any 90 degree angles. This is because of the tooling that is used to print the circuitry on to the AlN is made of wire mesh. When there is a 90 degree angle in the circuitry, the circuitry might get cut off or become very thing causing a high resistance in the trace.

Circuitry of Prototype

Circuitry layout

Preformance of Prototype

After the prototype was completed, it was taken to SPAWAR for initial testing to see if the chip would cool to 4K (7.2°R). The prototype was connected using nylon screws and thermal grease between the horizontal aluminum nitride and the vertical aluminum nitride back plate. Also, thermal grease (Apiezon) was added to the empty cavity on the bottom where the chip was exposed to attach a temperature gauge and keep it thermally connected to the prototype. Finally, once the prototype was pieced together, it would then be attached to the silicate cold tip using nylon 4-40 tapped screws as well as more thermal grease. This cold tip would then lead into the remainder of the system leading up to a vacuum and refrigerator pump.

For the testing stage of the project the team was particularly interested as to whether the chip would cool down to 4K (7.2°R). The assumption was that if this prototype could cool the chip to 4K (7.2°R), the ideal solution would have little or zero problems duplicating the results.

Results

The data was given in Temperature with respect to Time. The figure below shows the plot of temperature of the Prototype as a function of time. It also shows temperature of the cooling tower as a function of time.

Temperature of the Prototype and cooling tower as a function of time

Summary of Results

After running the cooling pump in a vacuum sealed dome, the results showed that the prototype only cooled down to 9K (16.2°R) instead of the desired 4K (7.2°R).

However, there were multiple factors that could have caused this discrepancy in the experimental value versus the ideal value.

1) The position of the temperature gauge was larger than the open cavity in horizontal layers and as the result, the temperature gauge was not reading the chips temperature, but the temperature of the PCB circuitry layer. Due to the fact that the PCB is not a great thermal conductor, this could have played a large role as to why the readings on the temperature gauge did not go down to 4K (7.2°R).

2) The two pieces of aluminum nitride were connected via thermal grease and the Teflon “L-bracket”. The L-bracket had a larger area of contact between the two pieces of aluminum nitride than the thermal grease, so it’s safe to say that a lot of the heat transfer to the horizontal aluminum nitride was being made through the Teflon rather than straight from the back plate. This alone would cause a decrease in thermal conductivity as the aluminum nitride was our main thermal conductor and would be the reason the chip would even cool to 4K (7.2°R). Since Teflon is not a great thermal conductor it slowed down the transfer process to the chip and could be the main culprit as to why the chip failed to cool to 4K (7.2°R). Ideally the aluminum nitride would be a single piece, therefore eliminating the issues caused by the Teflon as the two single pieces of aluminum nitride would not have a medium to transfer heat through and would instead provide an extremely large thermal conductivity coefficient and easily cooling the chip down to 4K (7.2°R).

Conclusion

Even though the results were not ideal, the team is still very confident that the co-fired ideal solution will provide more than enough thermal conductivity as well as contact to cool the chip down to 4K (7.2°R). Some actions that could have been taken to improve future testing would to use a different temperature gauge, preferably a small silicon diode temperature sensor that could be placed directly onto the chip to get a more exact reading rather than using the larger sensor which the team believes made a reading on the PCB circuitry rather than the actual chip itself.

Page updated

Report abuse