Waterblock Surfacing

Overclockers is supported by our readers. When you click a link to make a purchase, we may earn a commission. Learn More.

Interesting observations on flatness – Ian Anderson

In lieu of the $25 he normally pays for an article, Joe agreed to send me the Cooltechnika White Water waterblock reviewed here HERE. From the surface markings it appears to be the same one pictured. I would be quite happy with the performance as tested, but just for fun I placed it in my optical flat testing rig. (I should note in this article imperial units are measured in inches (or fraction of) while metric units are measured in microns – 1/1 000 000 meters).

I used my assembly for testing diagonal flats for telescopes. It consists of a spherical reference mirror from Jarrel-Ash monochromator. I used a standard Rochi test with a 100 line/inch grating. The test was arranged with the surface of the waterblock placed at 45 degrees between the test mirror (placed at 90 degrees from the grating) and the grating to create a double pass test.

The test was arranged in the following order: lightsource, diffuser, grating, waterblock, reference mirror, waterblock grating, camera (or in this case, eye). Because the light reflects off the waterblock surface, twice the error appears twice as strong. This is not a quantitative test but it is quick gives a good idea of whether or not more accurate testing is needed.

While the waterblock was sufficiently polished to allow the surface to be tested, it was not flat enough to give an intelligible result. Further mechanical testing against a precision ground flat revealed a concave cylindrical figure. Deviation parallel with the fins was well within 1/1000″ peak to valley. Deviation perpendicular to the fins is slightly over 4/1000″ peak to valley. I should also note the mounting pressure from the die simulator (left over after Joe’s test) deformed the base. This area was too small to test with my equipment but I would estimate it to be around 1/1000″.

HERE you will find a handy little application to allow you to determine the deviation at the contact area from the deviation of the whole surface (assuming it is a smooth curve). To use this app, enter the deviation as the sagitta. Due to the fact that the deviation is not perfectly cylindrical and appears to be flatter toward the edges, it would be reasonable to assume 2/1000″ deviation over the area of the IHS. The deviation over the area of an uncovered die would be negligible. I would expect much of this stems from the fact that the base was machined to 0.5 mm (as measured) thickness, making it difficult to machine lap flat.

Given that the thickness of the base over the core is 0.5 mm, lapping could cause more harm than good. Rather than use sandpaper which will likely create a convex surface, I ground the block against a porcelain tile, first using loose 220 grit silicon carbide followed by 500 grit emery. This brought the gap to within 1/1000″ over the IHS. I did not want to go further and potentially damage the base. I then surfaced the block with a piece of 1500 grit silicon carbide paper held by hand. To prevent oxidation, I used a drop of mineral oil instead of water to lubricate the paper for this final step.

Should you choose to lap your block, there is one more consideration you should make:

When thermal transfer is through a TIM (thermal interface material), surface finish also plays a factor in performance.

The size of the particles in the TIM should be comparable to the size of the pits in the surface. If the grit (in the TIM) is too large, it will hold the two surfaces too far apart. If it is too small, there will be too much fluid between the particles, resulting in a less than ideal transfer rate.

The size of the pits in the surface will be ½ to ¼ the size of the grit used to lap it. The size of the TIM particles should be ¼ to 1/8 the size of the pits in the surface. 1200 grit paper is typical of the surface of most heat sinks and IHSs – this translates into roughly 9 micron grit. The pits in the surface will be about 2.25 microns, meaning the ideal particle size will be about 0.375 microns. As you will notice, this is almost exactly the published size of the particles in Arctic Silver thermal compounds.

Ian Anderson

Custom Optical Systems


Leave a Reply