Chemically lapped cpu and heatsink

[Modelworks]

I was lapping a cpu & heatsink today and it turned out well.
It originally had some very high sides with a dent like shape in the middle.
As I was finishing up I just wasn't happy with the surface.
It felt smooth but in the light I could see scratches.
Some say that it will cool better because the compound will go into those micro scratches.
I decided to try something I hadn't thought of before and haven't seen mentioned elsewhere, so if it has please post me a link.

Anyway, I develop my own pc boards sometimes, etching them in a solution that removes copper.

WARNING !
WARNING !

DO NOT DO THIS INDOORS OR CARELESSLY. YOU CAN INJURE YOURSELF AND OTHERS AND DESTROY THE CPU.

You need two things.
Muriatic acid, available from a hardware store, about 5.00 a gallon.
Hydrogen peroxide , standard 3% stuff from the drug store, about 1.00
A plastic container
Some q-tips or cotton swabs
some gloves, latex or rubber work fine
Paper towels.

Outside or in a well ventilated area, this stuff causes fumes that are really bad to breathe.
Pour a small amount of the muriatic acid in the container, I use about 1/4 cup.
Add about half that amount of peroxide.


You won't see much happen.
If you want to know if you got it mixed right, take a q-tip and wet it with the stuff, then touch a penny or anything copper, it will instantly turn green.

You really need to get a feel for the stuff so try using it on a penny before using it much on a cpu.

Carefully take the qtip and place a small drop on the top of the cpu where you used the sand paper. Rub it around a bit then wipe it off with paper towels.

Don't just put the stuff on there and leave it.
It takes it awhile to dissolve copper, but it can dissolve it pretty fast depending on the solution strength.

In the end you can get a heatsink and cpu with a mirror like finish with absolutely no scratches or dirt, etc.

After a bit the solution will stop working, usually you can tell by the color turning all green.

If anyone else tries this let me know.
And please be careful with this stuff.

 

[MadScientist]

If you try this definitely do it outdoors. The reaction of HCl (muriatic, hydrochloric acid) and hydrogen peroxide generates chlorine. And remember from Chem 101, always add the acid to the peroxide.
2HCl + H2O2 -->2Cl + 2H20
2Cl + Cu -->CuCl2
or neat:
2HCl + Cu + H2O2 -->CuCl2 + 2H2O
The etching reaction:
Cu + CuCl2 -->2CuCl

I prepare my etching solution by adding 1 part Muratic acid (the common kind that's sold in hardware stores, ~20 baume = 31.45%) to 2 parts 3% Hydrogen Peroxide (the common kind that's sold in Walmart & drugstores).
Be very careful when opening the cap of 31.45% HCl. It will fume.

 

[aigomorla]

im kinda lost in this.

So basically your going after that super mirror finish? however it involves acid disolving the copper. How is it flat if the acid is disolving the copper? Wouldnt that leave micro pits on the copper surface which would then destory the whole point in lapping?

When you say rubbing, that doesnt appear very flat and even for the acid to disolve, so i dont understand the point in this over a lapping up to 2000grit.

You get pretty good results at that grade.

 

[Rubycon]

Sounds dangerous and "flying blind" in the name of precision. Also if it were possible to maintain truth (in surfaces) some residual highly corrosive hydrogen chloride residue can drift into tiny nooks and crannies of the CPU doing lots of damage that may not surface right away. On the heatsink side LOTS of post prep is required to make sure ALL HCl is completely neutralized.

 

[EarthwormJim]

Why would you do this? Shiny does not equal flat.

The point of lapping isn't to remove all micro pits but to make the heatsink and ihs flat.

 

[Modelworks]

The point was to remove the fine scratches that are left by any sand paper .

The acid is dissolving the fine scratches that remain.

I think there is some misunderstanding as to how strong the acid is.
Its not a process where, you drop it on there and instantly the copper is gone.
The process takes time, which is why I said to rub the area with the qtip.
Agitating it makes the process move along quicker, otherwise you might be there for awhile and eventually have a slight (we're talking hairline, depression), where the acid was left.

The only thing dangerous about it is improper handling.
I and thousands of others use it everyday to make pc boards and its not going to cause a chip to short or anything else unless you dunk the chip in it.

Cleaning it off afterward is also very easy. You can use the same method that they suggest in preping a chip thats in the arctic silver instructions. Or you can wipe it off with soap and water. Its not that strong a mix of hcl to begin with.

The biggest thing I was concerned with was someone doing this indoors as the fumes are bad.

Mirror image does not mean flat, but if you can see any scratches at all then that surface is not truly flat. This is a very easy way to remove those final scratches for those interested .

 

[Ika]

I believe that people who have tested CPUs and heatsinks that have been lapped found that the mirror finish actually deteriorates the cooling ability of the heatsink. I don't recall the exact numbers, but the results showed that the grit that gave the best temperature was 1200, and a mirror finished lap job was 1-2C higher.

 

[Modelworks]

Originally posted by: Aflac
I believe that people who have tested CPUs and heatsinks that have been lapped found that the mirror finish actually deteriorates the cooling ability of the heatsink. I don't recall the exact numbers, but the results showed that the grit that gave the best temperature was 1200, and a mirror finished lap job was 1-2C higher.

That was one of my thoughts as well.
I'm not sure how the compound interacts with the rougher surface.
I guess it would also depend on the particle size of the compound.


Another method I thought of thats a lot harder to do correctly would be to lower a heatsink or cpu just far enough into the etchant to contact the surface. The etchant would have to be perfectly still, which really doesn't make the etchant work well, but the way liquids always seek their own level it would give a truly flat surface with very little work.

 

[Rubycon]

Originally posted by: Modelworks

Another method I thought of thats a lot harder to do correctly would be to lower a heatsink or cpu just far enough into the etchant to contact the surface. The etchant would have to be perfectly still, which really doesn't make the etchant work well, but the way liquids always seek their own level it would give a truly flat surface with very little work.

Or export CMM plot into a printer that can spray resist on the lowest parts.

 

[DerwenArtos12]

If we're going to go through all this trouble why not use a fixed spectrum lazer attached to a milling head to do the hard work. It would take some time and probably a custom ground lens but, that'd be the way to get it perfectly flat and smooth. I only worked on high powered yag lazers but, some of the other projects in my lab could etch business cards onto aluminum or etch 3D images into solid blocks of lexan so I know it can be done. Or, you could use a fixed sample decking mill to get it perfectly flat, I've seen those that guarantee les than 1/1000th of an inch across a foot. Why don't we? cause it's really just not worth it in the end.

 

[MarcVenice]

Just a silly question I came up with. Why don't we use 'curved' heatsinks, with the heatsink and the IHS making a perfect match ? Wouldn't the contact area be bigger? Could be done, right ?

 

[DerwenArtos12]

Originally posted by: MarcVenice
Just a silly question I came up with. Why don't we use 'curved' heatsinks, with the heatsink and the IHS making a perfect match ? Wouldn't the contact area be bigger? Could be done, right ?

It would be substantially more difficult to match and convex and concave surface perfectly than two flat surfaces but, theoretically it could be done and it would increase surface area.

 

[MadScientist]

Originally posted by: Rubycon
Sounds dangerous and "flying blind" in the name of precision.

This is dangerous in the wrong hands. I've never tried this on a CPU or HS and never will, but have used it to etch copper.
You are "flying blind" because it's impossible to know whether you are etching one part of the IHS or HS more than another.

Originally posted by: Modelworks
Mirror image does not mean flat, but if you can see any scratches at all then that surface is not truly flat.

The question where to stop when lapping a heatsink or CPU has been discussed many times here and on other forums. On one side you have the "It's got to have a mirror finish, so lap it up to 2000 grit sandpaper" and on the other the "Get it flat and stop at 600 - 800 grit". I came across this article awhile back at Overclockers. Link The article is old but the theory, in my opinion, is still valid for today's heatsinks and CPUs.
It makes sense to me. Of course opinions are like you know what, everyone has one. Personally I'm in the "lap til it's flat" group and stop at 800 grit.

 

[Modelworks]

Originally posted by: DerwenArtos12
If we're going to go through all this trouble why not use a fixed spectrum lazer attached to a milling head to do the hard work. It would take some time and probably a custom ground lens but, that'd be the way to get it perfectly flat and smooth. I only worked on high powered yag lazers but, some of the other projects in my lab could etch business cards onto aluminum or etch 3D images into solid blocks of lexan so I know it can be done. Or, you could use a fixed sample decking mill to get it perfectly flat, I've seen those that guarantee les than 1/1000th of an inch across a foot. Why don't we? cause it's really just not worth it in the end.

No, we wouldn't use that method because its seriously complicated and expensive.
I'm using something that is under 10.00 cost and easy to implement.

 

[Modelworks]

Originally posted by: Rubycon

Originally posted by: Modelworks

Another method I thought of thats a lot harder to do correctly would be to lower a heatsink or cpu just far enough into the etchant to contact the surface. The etchant would have to be perfectly still, which really doesn't make the etchant work well, but the way liquids always seek their own level it would give a truly flat surface with very little work.

Or export CMM plot into a printer that can spray resist on the lowest parts.

Thats an idea I had not thought about.
I'm wondering if I couldn't apply a coat of resist then scrape across the heatsink with a straight edge removing just the high areas of the resist.
That would expose only the high areas to the etchant.
It would then be possible to etch the parts without ever using sand paper.

Apply etchant, wait, check flatness, apply etchant, repeat until its done.

 

[Zepper]

Things are flat because the top of the silicon wafer is flat. To work well, everything from the inside out would need to be curved.

.bh.