Some answers on a common practice – USASMA
Lapping is a very complex subject that I haven’t seen treated in detail – so I wrote this based on my experiences shared with me by metalworking mentors (a metallurgist and 2 gunsmiths) over the years. This paper will be split into three parts:
- Single Piece Lapping
- Is It Flat Enough For Me?
- Do You Even Care If It’s Flat?
- Flat vs Mirror Shine
- Two Piece Lapping
- Is It Flat Enough For Me?
Single piece lapping is the attempt to flatten an item by rubbing it against a flat surface using an abrasive to remove the “high” spots.
The Flat Surface
Commonly we use glass as our “flat” surface. If you visit an older house, you can look at the windows and see the imperfections in the glass. Has glass come any further since the olden days? Yes it has, but it’s still not flat.
How Flat Is It?
A Google search turned up many different results, some of which have tolerances measured in millimeters
Placing the glass on a flat surface (when we lap) helps to minimize the flex. But, depending on that surface’s flatness, flexing of the glass can still affect your lapping efforts
Abrasives are particles that are harder than the surface being lapped. They scrape away the material that they come into contact with
Different size particles remove material at different rates. The composition of the abrasive also affects the rate of material removal. Size is also affected by use. The more that you use the abrasive (without replenishing it) the more the abrasive particles are worn down (new sandpaper cuts better than old sandpaper).
Fine quality abrasives have a more consistent particle size than lower cost abrasives. In short, you gets what you pays for! The different sized particles in the cheaper compounds cut unevenly. It’s not easy to see or to measure the difference, but it is a source of error that should be considered.
The Piece Being Lapped
Presumably you’ll be lapping a chip of some sort and a heatsink of some sort.
Friction and Drag
Friction in lapping is caused by the contact of two surfaces, whether it be the compound and the piece being lapped and/or the compound and the flat surface underneath it. This, along with the difference in abrasive particle size, can cause uneven lapping.
Also, as you push along while lapping, the uneven particle size causes the lead edge of the piece being lapped to drag more than the rest of the piece – causing more metal to be removed from the lead edge than is removed from the other edges.
Back and forth lapping removes more metal from the lead and trail edges than from the rest of the piece. Circular lapping removes more from all the edges than the center (and even more from the corners). The result of all this is that hand lapping actually causes the piece being lapped to be slightly convex in shape.
Pressure and Flex
Now, in addition to the convex shape, add the effects of pressure on the piece being lapped and the effect of flexion on the piece being lapped and the flexion of the flat surface.
Typically, uneven pressure results in more metal being removed from one side or the other, depending on the lapping technique being used. And, all of this varies some more with the varying degrees of hardness for the flat surface, the abrasive, and the piece being lapped. Phew! That’s a lot of error creeping in here – what’s an avid overclocker to do?
Is it flat enough for me? How much time, effort, and money do you want to invest? How important is it that your CPU mate exactly with your heat sink? Arctic Silver is a wonderful compound! Presumably you want the surfaces to mate well enough to enable effective heat transfer.
Do You Even Care If It’s Flat?
So, does it even matter if the surfaces are flat, as long as they mate together very closely? Heat is transferred from one object to another (for our purposes) largely by conduction. That is, if 2 pieces are touching each other heat will be conducted from the hotter material to the cooler material.
So, as long as the pieces maximize the touching of surfaces (in other words: The entire surface of the CPU completely contacts the heatsink), then the maximum amount of heat will be transferred.
It is the amount of contact between the 2 surfaces that matters – not the flatness. This is because, if the surfaces don’t mate exactly, there is air in the gaps and air is a very poor conductor of heat when compared to the metals we are working with. BTW – it also varies based on the materials that are being used – each has a different rate that it conducts or gives up heat – but that’s beyond our control (with the exception of adding thermal paste or tape).
Flat vs Mirror Shine
There has been some debate over whether a mirror finish is better than a surface that approaches a mirror finish. As long as the surfaces mate exactly (impossible for us to achieve) and the 2 objects conduct/release heat faster than any intervening substance (like thermal paste or tape) then a mirror finish should prove more effective.
Unfortunately, experimentation doesn’t prove this out. My guess is that the mirror finish surfaces don’t mate exactly – thereby diminishing the heat transfer – while the “not quite” mirror finishes have more surface area for the thermal paste or tape and thereby transfer the heat more rapidly.
So, once again, what’s a poor overclocker to do? It seems that there are lots and lots of problems with lapping a CPU or a heat sink. This brings to mind a trick that an old gunsmith taught me for mating parts. That is:
Since it’s only the mating of the surface that matters, not the flatness, why not lap the two pieces (ie CPU and heatsink) together in order to achieve the mating. Assuming that you use a high quality lapping compound, the results should be more encouraging.
The theory behind this is that, given a uniform particle sized abrasive, two surfaces that are lapped against each other will abrade the high areas on each that are preventing the precise mating of the surfaces.
Further, if you lap outside of the 2 mating surfaces, you will introduce some error. The most common method of two piece lapping is to spread the compound on one part and lap the other with it using circular motions. This tends to minimize the flex and pressure issues, but it brings up another issue.
In making circular motions, you will be lapping outside of the areas that will make contact when you mount the parts together. It seems to me that this error is less of a concern than the errors that we normally introduce in single piece lapping. I use this method when I lap my parts and haven’t had any problem with them – I don’t have any data to back this up (Although I’ve got a brand new P4 – hmm…I’ll try it both ways and report the results later!!!).
It seems to me that the ideal way to mate these surfaces is by lapping them together using a circular motion. This will maximize the lapping of the mating surfaces and will minimize the lapping of the surfaces that are beyond the mating surface (because the CPU surface is a square and not a circle).
This could be accomplished by mounting the CPU on some sort of drill press. There’s lots of problems with this method, but it seems to hold the highest promise of a nearly perfect mate between the chip and the heat sink.
Well, enough of this! I don’t know if this will be useful to anyone, but it was a subject that needed discussing.