PDA

View Full Version : Lapping??


XprincoX
12-30-01, 02:30 AM
Ok will someone pleeease explain the idea of lapping to me? TIA

xCarne_Asada
12-30-01, 03:45 AM
Lapping is when you sand down the bottom of a heatsink/waterblock so that it gives a mirror finish (you can see yourself in the bottom of the heatsink =). anyways it also gives you better contact between the heatsink and the cpu core and chipsets. I am still a newb with this, so anyone feel free to add.

Shadow рс
12-30-01, 04:13 AM
for the heatsink, start with wet/dry sandpaper (keeping it wet) at around grit 320.
at this point you'll be amazed just how "un-flat" your sink is. Make sure to keep the sink on a very flat surface (I recommend a piece of glass) and use figure 8 motions. Once you see the surface is evenly sanded, repeat in small increments. I recommend 400, then 600, then 800, then 1000 and finally 1500.

it won't be a mirror finish by any means, but will be a truely flat smooth surface.

The point? The flatter your surface between the heatsink and die (or cpu) is, the less microscopic air gaps there are between them. A truely flat surface would not need thermal compound, as this is only to help fill those air gaps. Air doesn't transfer heat as efficiently as metal to metal.

If you wanna get really technical, you can also lap your die. I'd recommend starting with 800 or more, and finishing with 1500. you only want to sand it down a tiny amount.....just enough to be sure it is flat as well. I don't recommend anyone who's unfamiliar with this "tweak" go this far tho...the heat sink alone will make a difference.

XprincoX
12-30-01, 01:01 PM
interesting idea... i see the pros, but what are some of the cons to such a techinique? Don't you run the risk of over sanding and risking the core all together?

Boilerhead
12-30-01, 11:07 PM
I worked in a safety valve shop for a number of years (part of the family business) and we used to do a lot of lapping. After machining the valve seat and disc on a lathe we would sand the valve seats using lapping plates with sandpaper on them and sand the discs on a surface plate which had sandpaper clamped to it. The figure 8 motion is absolutely correct and you should only remove as much material as is needed to get rid of the low spots.

Once we had a flat surface from sanding (usually with 120 grit) we would lap the seating surfaces. The discs were easy, we just tossed them into the lapping machine. It had a spinning plate with grooves in it and rings sat on top of it held in place from the center spindle to the outside by sealed bearings on posts. You put the discs inside a ring and walked away and 5 to 15 minutes later you had a perfectly flat seating surface on the disk. We also used that machine to make new lapping plates in various sizes. I wish I had that machine today for doing HS and water block work. To lap the seats inside the valve we used the lapping plates with Clover valve grinding compound which I have always called lapping compound. It came in various grades all the way from what almost felt like pebbles to a very fine toothpaste like goo. That stuff should work great for superfine core and HS finishes for the complete anal retentive lapper. No thermal compound would theoretically be needed between them, but I think it's a good idea to compensate for unequal thermal expansion alone. You wouldn't need much, just the tiniest little dab. Glass laps would work just fine as long as the glass is flat to start with - not all glass is flat so beware.

Surprisingly, I haven't lapped a CPU yet, but I wouldn't hesitate to do it if I felt the need for the closer tolerances. If you're going to try it I would recommend 1000 or 1200 grit paper and be absolutely certain you don't rock it at all, you shouldn't have to take much off at all. If you want to get it even flatter then consider making a small glass lap (2x to 3x core size) and getting your hands on some extra-fine valve grinding compound. Get a good magnifying glass or low power microscope and when you don't see any pits or scratches you are done. Clean it with alcohol and you are good to go.

I imagine that only the most extreme setups need anything more than 1200 grit sandpaper flatness, but I know some of you _need_ to have it purrrfect and this is about the only way I know of to do it that well. Sandpaper only gets you so far...

Henry

CrystalMethod
12-30-01, 11:56 PM
Originally posted by Xprincox
interesting idea... i see the pros, but what are some of the cons to such a techinique? Don't you run the risk of over sanding and risking the core all together?

Yep, that's why most of us just lap the heatsinks. I personally don't trust myself to lap a CPU die, I'd probably be so nervous about something going horibbly wrong, I'd screw it up anyways. Kind of funny, that I'll grind a $500 camshaft by hand, but I'm scared to do it to a $80 CPU.

XprincoX
12-31-01, 01:59 AM
What type of sandpaper who ya suggest for lapping hs? How do u know when its smooth enough?

Shadow рс
12-31-01, 07:33 AM
as I said...finish with 1500...some finish with 2000.

there's a debate over what's smooth enough, but in my mind, anything above 1500 is probably 1000 times better than the factory surface already.

ultimately, sand it to a mirror finish..sand the die the same way. I'm not willing to go that far, as I can see (usually) a 10 degree f difference with just doing that.

I actually did sand on waterblock to 2000...then polished it with Crest (yes toothpaste) to an almost mirror finish and posted pics of the pretty block here. But truth be known, I bet there wasn't a whole degree difference between testing after a finished 2000 grit and an almost mirror finish surface.

Sand with each progressive grit. Once you've finished with 320, start with 400....Look at the surface...you can see the difference.

progress to the finer sand papers.....each step look and see the progress. It's visible trust me. Even between 1000 and 15000.



As I said before....use wet/dry sand paper...it can be found in any auto body supply store. (3-M makes a great product)