Building a Lapping Machine

[Neomoses]

In an effort to produce waterblocks that give consistent results from system to system, I'm going to build a lapping machine. I cannot afford a real surface grinder, but here's my thoughts:

First, make the WB bases flat on a mill with a facing cutter.

Make the surface out of a 9"x12"x3" granite surface plate, guaranteed flat to within 0.0001." Lay a piece of sandpaper face up on the granite block and secure it if necessary (not exactly sure how I'm going to secure it yet, ideas are welcome).

Use a 1/3 HP electric motor to create linear motion, giving roughly 6 inches of displacement. Motion of the Waterblocks should be no faster than 10 inches per second. (any input on WB speed would be appreciated) Attach WBs to the crank assembly on the electric motor and have them moving on top of the sandpaper/granite block. Sand with 400, 600, then 1000 grit sandpaper. Use a flood system to continually wash the sandpaper with soapy water to carry away the grit/copper particles.

I think I can build this machine for between $400-$500, which is much cheaper than a surface grinder. Do you think it will produce sufficiently flat and smooth surfaces for WB bases? Also, do you think the motion should be linear or elliptical? Linear would be easiest, but it has a potential to produce "grooves" or "waves" in the bases of the blocks if the sandpaper is uneven. This would be a great thread for people like BillA to jump in on, as an anal-retentive nature and attention to detail would be perfect for this (no harm intended).

[ChrioN]

I don't have any tips for you but it sound REALLY interresting. Want one myself.

good luck!

[bigben2k]

An intriguing project!

On the "linear vs elliptical" issue: it's been recommended to keep it linear, but only because of the hand process (i believe). The important question is: how are you going to keep it flat?

You've got that granite slab as a start, but how on earth are you going to maintain even flatness? To me, the elliptical solution sounds better, but I'm not sure, nor convinced, that you'll be able to design a block clamp that'll remain parallel to the granite's surface.

As for holding the sandpaper down, you should definitely avoid using some kind of adhesive, between the sandpaper and the block, because that would make the surface uneven. You probably want to secure the sandpaper at its edges. The problem, again, is that we don't know how flat the sandpaper itself can be...

You've got another problem to resolve: how are you going to measure your results?

Don't get me wrong, I like your idea. I think you could actually market this as a service, if you can get right. It's going to depend on you doing one thing: identifying the sources of error/deviation, and solving them.

[gone_fishin]

If I were going to make a rudimentary time saver such as this, I would start out with a contraption such as the sketch I drew up. Then through trial and error refine it. The weight of the granite slab itself would make for a good hold down of the sandpaper edges if it were folded over the edges and under the slab. This method would still require a hand touch as it is but you would only be required to hold the wb still in one place. This would give an oscillating motion to the slab. The ball rollers can be easilly bought in woodworkers magazines as they are frequently used in shop jigs. The size of the pulleys on the shafts would control the speed.
Does this help at all?

[BillA]

looked at this for a loooong time, but don't have any volume to justify more than the granite surface plates

sandpaper is crappy, got to use PSA film to keep it from 'riding up' in front of the piece
ANY rotary motion will rock the piece

far and away the cheapest is the flat lap end of a faceting machine (for gemstones)
BUT you will need at least a 8" diamond lap, probably in several 'grits'
and 10" will be FAR better, and a LOT more expensive

you need to lurk on eBay, I have seen 20" Lapmasters (the gold standard) go for $800 or so
(you do need to investigate how it is done commercially)

but you also have the cart in front of the horse (ahem, as usual NeoMoses)
-> how can you distinguish between flat - and non-flat ??

THIS is flat, to 5 millionths ! (from a yet to be released new wb)



if such flatness is a concern, suggest you get an optical flat and a monochromatic lamp to first get some 'flatness lessons'
- then worry about how to do it

[Since87]

One potentially useful place to look for lapping information is in the area of amateur telescope making.

Most of the information is likely to be on grinding parabolic mirrors, but if you search for info on making "diagonal mirrors" you may find info useful for creating flat surfaces.

[N8]

Heh, I have Lapmaster 15's here at work, and getting a 20 or two in the near future, when we move into our new building (no room for any more equipment until then.) The 15 and 20 refer to the diameter of the platen, otherwise they are essentially the same. Awesome machines, but you have to 'understand' how they get things flat to keep them maintained to keep things flat in the first place. You can use different grit slurries on them, but I would recommend only using ONE grit size and type, otherwise you will have to tear the machine and pump apart every time to avoid grit cross contamination. I have separate machines for different grit sizes and types. The coarser ones use SiC (silicon carbide) 600 grit, and finish laps use 1400 grit Al2O3 (aluminum oxide - sapphire). After this the parts can be polished if need be.

You can measure flatness mechanically with 'GOOD' equipment, but optically is your best best. BUT, this means you have to polish the surface to get it reflective. Polishing then can get the piece un-flat again (usually convex) when you round the edges off. This isn't entirely bad, as usually (for our purposes), the center region where the heatsink touches the processor is still the flattest region and the rounded edges don't touch anything.

Here is a KCl (potassium chloride) optic crystal bonded to an OFHC copper substrate buffer. KCl is a very soft salt that is soluble in water, breaks at 90 degree angles with any pressure, and scratches very very easily (Q-tips will gouge the surface.) I lapped both sides of both pieces on the Lapmaster (with lapping jigs to protect the edges - yes, you will need to machine jigs for heatsinks, etc., as the parts need to rotate (therefore need to be round.)) I then polished both sides of the copper and KCl, bonded them together, then lapped the KCl down to 315 microns thick, then hand polished it down to 300 microns thick with 5 different micron diamond polishing oil based solutions from 6 microns down to 1/10 micron. At this point, with this soft of a sample, there is no mechanical way to measure flatness anymore as it would damage the surface polish, and optics are the only way to double check your work.

As BillA stated, you can lap anything with the best equipment, but you need to be able to determine if you are truly flat or not. If you notice in this picture, I have 'almost' a perfectly flat surface with straight lines (sorry for the difficulty of seeing the lines, this is not a metallic surface, but an optically transparent surface, very hard to get interference lines.) This is with a 1/20 waveplate. I can measure flatness mechanically with a digital micrometer down to 1 micron of deviation, but with optics you can measure down to about 1/100 of a micron.

Let me stress again, that even if you have a Lapmaster, it has to be maintained/adjusted properly and often to keep the platen flat, and you can NOT assume that your parts are flat unless you have some sort of accurate measurement techniques. If there is a Lapmaster on Ebay or whatever, it is probably because the platen is worn down unevenly to the point where they couldn't properly adjust it back to flatness, the pump went bad, or bearings are wearing out, etc. These are beasty machines and can last 40+ years with proper maintenance and care, so be cautious of buying. Research before just jumping into this.



I have a lot of pix on the main page of my site right now that are just related to lapping and polishing, with examples of what some stuff look like right off the lapping machine and what they look like at various stages of polishing. There is even a picture of a Lapmaster 15 that uses the 1400 grit.

http://www.wsu.edu:8080/~i6735189/cp...oledpelts.html

[BillA]

lol
I do like your posts N8; and people think I'm 'far out' ?
lol

[N8]

Originally posted by unregistered
lol
I do like your posts N8; and people think I'm 'far out' ?
lol

lol, yea, I don't post very often, so I have to make up for it... right?

Jam as much content in each post as I can. Sorry, I work with rocket scientists for 9 hours a day, I need to vent once in a while.

[Neomoses]

This is great guys, exactly what I was looking for. Here's a few questions for y'all. First to BillA: In an article at overclockers.com entitled "Waterblock Bench Testing" you note that you almost always saw an improvement in temperatures when you lapped blocks. What is your method for lapping? Is it just the standard sandpaper on a flat surface, or do you use a lapping machine? Since you now have an optical flat and a monochromatic light, I'm wondering how flat is "flat enough?" Over the die area, is it 1/1000"? 1/10,000"? 1/100,000"? 1/1,000,000"? Also, if you are lapping these blocks by hand, how flat can you reliably get them? Are your results repeatable within 15% each time, or is there a lot of variation?

Now comes a question for anybody. Where can I get an optical flat/monochromatic light for a good price? Keep in mind, I would like to accomplish this entire project for less than $500. I realize that the measuring equipment that I currently have is not even in the ballpark for measuring these tiny changes in elevation.

N8, you have a lot of good information. I'd love to hear more.

Basically, this is my situation. Since I live in the northern hemisphere, and there's now snow on the ground, my money trees have lost all their leaves. (AKA I don't have a lot of money) Buying a machine such as a lapmaster is out of the question.

I currently face mill my blocks to make them "flat" and then hand lap them on a granite surface plate to get what I deem a "good" surface finish. I would like to automate this task and (if possible) improve the flatness and surface finish on the blocks. In the process, I'm sure I'll invest in more inspection/measurement tools to ensure that my bases are actually improving. Thanks for all the input so far, and keep it coming!

[bigben2k]

There's still a long way to go...

I've thought (a long time ago, before I knew better) that I could use an orbital sander for lapping

But now that I think about it, if you could hook up a block to it, that might get you somewhere. The problem of course is that the sander is not likely to stay flat, and I mean "not at all"!

Like Bill said, circular motions will result in the block tilting.

So the linear motions are definitely the way to go.

But your first step is getting to be able to measure it.

[gone_fishin]

Makes me wonder, what ever happened to the "600 grit is fine enough" montra? Now the talk is in microns

[BillA]

flatness and finish are not the same, at all

[gone_fishin]

Micron sized polishing material has been mentioned.

Here is a lap machine, would it save on the elbow grease?
click me

[bigben2k]

Originally posted by gone_fishin
Micron sized polishing material has been mentioned.

Here is a lap machine, would it save on the elbow grease?
click me

That's the closest thing I've seen to a potentially easy solution! Nice going!

[BillA]

you guys need to dig into the info already provided
there are reasons for the recommendations made, and for things not recommended

'flat' is defined quite differently by users with different purposes in mind

BTW, I have probably 20 books or so on lapidary techniques and equipment, a past hobby

no free lunch, but there are some very good buys on eBay,
IF you know exactly what you are after

[Neomoses]

So, we just need to set some goals that are reasonable. How does flatness to 0.0003", repeatable between all blocks sound? I may be way off on that number, but it's a number to shoot for. Does anyone know if that is a high or low value for typical waterblocks of today?

BillA, do you happen to know any typical values for the flatness of Swiftec, DangerDen, or any other waterblock manufacturers? I'm just looking for a ballpark figure here. Any help would be appreciated.

[BillA]

Neomoses
you don't need numbers on other mfgrs' products
(which I would not want you quoting in any case as I do not trust you)
my articles and various threads have a bunch of this (general) info

-> but you, as a wb mfgr, need the ability to inspect/measure for yourself

the numbers vary greatly, but with the present attention most mfgrs seem to be improving - some hugely
note that 'very' non-flat, specifically concave, can = a 4°C CPU temp 'hit' (I have measured such)

let us not return to the recurring difficulty of unsubstantiated claims
- what on earth can a dimensional limit represent without the ability to measure such ?
(and pay attention to the accuracy/uncertainty, because I will)

[RoboTech]

Very intersting thread, with lots of good info...

I, like many, hand lap my waterblock and heatsink bases with the hope of improving heat transfer thru better surface area contact. I use plate glass, wet/dry paper and linear motions and try to keep things as clean and consistent as possible. I realized some time ago that Flatness was my goal and not a pretty surface finish. I usually stop at 1200 grit and rely of the "vissibility" of high and low spots to judge flatness. When I can no longer see high/low spots then I have gone as far as my limited tech will take me. Thru comparative testing, I can regularly see a repeatable improvement in temps - from pre-lap (not flat) to post-lap (as flat as I can get it) so I know I am doing some good. I hold the impression that the effort involved with lapping produces diminishing returns - the closer you get to Flat, the less change in thermal transfer that will occur.

I don't have an optical flat and can't justify the cost in buying one to support my hobby and curriosity... (Maybe someday)

So what is the average enthusiast to do??? For those of you who do have the equipment and experience to know FLAT and be able to measure it, are we basement lappers just kidding ourselves (in my case I think not - because I can measure (with some degree of repeatability) an improvement in temps)? Can someone give us some rough guidelines for what typical flatness would be for a fine machined finish, hand lapped surface, or professionally flattened surface???

And I would love to know how FLAT the top of the CPU core is... I assume they are pretty flat, but what does a typical XP core or P4 heatspreader look like under an optical flat??? Anyone have pics?

[BillA]

http://forums.procooling.com/vbb/sho...&threadid=5463

[N8]

Originally posted by gone_fishin
Makes me wonder, what ever happened to the "600 grit is fine enough" montra? Now the talk is in microns

600 grit 'IS' fine enough, since 99% of you do not have the means to measure flatness, therefore there is not a need to do a fine polish job.

I was stating that to use optical flatness measurement techniques with optic flats, the surface needs to be sufficiently reflective (polished.)

You can talk about lapping/polishing papers/pastes in grit or micron size.

Surface height deviation from true flatness can be talked about in (percentage of) wavelength of light, or in SAE or metric terms such as millionths of an inch or in microns, etc.

Be careful not to confuse the issues and measurements.

Heatsinks are still large enough that you can measure surface height deviation (flatness) mechanically with a good dial indicator (dial or digital) and a pair of accurate granite or steel block flats, down to 1 micron, but you have to know what you are doing.

CPU surfaces are NOT large enough to do this, and their flatness can really only be measured by optical means or with a profilometer, etc.

[N8]

To add more about surface finish:

Surface finish is an absolute measurement. The means to get there (sandpaper, grit sizes, polishing compounds) are only relative - as in there is a variation.

Lets take 600 grit for an example. 600 grit is just an average size assigned to the material. The deviation from average could be small or large. The sanding paper may be labeled 600 grit, but it may have anywhere from 400-800 grit particles on it. Higher quality papers will have a smaller deviation from average.

The material on the sanding paper may be different. Silicon carbide grit will remove material differently than aluminum oxide grit or even diamond grit.

The SHAPE of the particles on the sanding paper will be different. You could have two different pieces of 600 grit sandpaper with silicon carbide on both, but 600 grit is just a size range, it has nothing to do with shape. What if there are silicon carbide pieces on one paper that are more pointy and the other has more rounded grit?

Another issue is the density of the grit on the paper. You could have two pieces of 600 grit sandpaper, but the amount of grit on each piece could be different. The piece with a higher density per square inch will produce a finer, more consistant finish, while the piece of sandpaper with less grit on it will produce a more variable scratch pattern.

The length of time you use the sandpaper will affect the shape of the grit on the paper. It will clog the higher density papers faster than lower density sanding paper. The longer you use the sandpaper, the more rounded the tips of the grit become, altering the shape of the grit.

NOW, consider that all of the above variables are combined. Just think of the different variations in conditions you could come up with for 600 grit sandpaper. You can end up with a surface finish that is TOTALLY different depending on these variables, even if two different people finish up with 600 grit sandpaper.

Now think about operator differences. Even if two people have the exact same 'movement' techniques, exact same sanding paper, one of the two might push very lightly on the paper, the other will push very hard. This pressure difference with produce very different results for surface finish.

Whether you use water or dry sand will also change the surface finish.

There are many many variables that affect what we are doing.

The only measurement that is absolute is the surface finish on the material itself (heatsink, etc.)

For a good read on grit sizes and issues, conversion tables, etc., go here:
http://www.newportglass.com/grit.htm

One word of caution, grit size conversion tables are NOT standardized, so there are some variations to be found from different tables.

[Roscal]

A more complete table here with several normalization in world:



FEPA (european norm) go to 4000 (5µm) but not standard

[RoboTech]

Thanks guys...

(Sorry I missed your earlier thread Bill - I should have searched first before blurting out questions in my enthusiasm.)

N8: I understand that not all wet/dry papers and films are created equal but don't know how to properly choose one brand/type over another. I generally use NORTON Tufbak Durite T421 600-A and 3M Imperial Wetordry 1200. Are these OK or would you recommend something different???

[N8]

Originally posted by RoboTech
Thanks guys...

(Sorry I missed your earlier thread Bill - I should have searched first before blurting out questions in my enthusiasm.)

N8: I understand that not all wet/dry papers and films are created equal but don't know how to properly choose one brand/type over another. I generally use NORTON Tufbak Durite T421 600-A and 3M Imperial Wetordry 1200. Are these OK or would you recommend something different???

In all honesty, I don't have any specs on either one, but I have had good luck with stuff that is approved for automotive paint refinishing. Check with your local auto-body shops to see what they use. They can't afford to use sub-standard stuff, otherwise they would be ruining their own paint jobs.