machine#construction: foundations: scraping


Hand scraping, old style

Someone may wish to write a rant here about how machinists in "civilized" countries are still trained in the practice of hand scraping.

Power scraping and tricks

Les Watts wrote this short article on using an anglegrinder and machinist's level to aid in hand scraping.


When one is contemplating a design for a shop built cnc router or mill often
long precision flat surfaces are required for mounting linear motion rails.
The rails are flexible enough to easily take the shape of whatever they are bolted to
thus the requirement for the accurate mounting surface.

Traditionally this would be accomplished with a large Blanchard or Mattison type Grinder. The logistics and cost of getting a machine frame to one of these monsters Is often prohibitive.

I will briefly describe a way to do this surfacing in the small shop with only hand power Tools. It is a good bit of hard work-but the results can be as good as a machine ground Surface.


Such a machine might be made of aluminum or steel. I suggest using HOT ROLLED
Mild steel such as ASTM A-36 or 1018. Other materials such as cold rolled and drawn Steels and extruded aluminum can have significant internal stress. Removing material From one side will cause very significant warping.


It is well known that scraping techniques can generate straight edges and flat surfaces in sets of three without having a reference standard. We were looking for something that would work on one piece. We further wanted to flatten a rough piece that might be 1/16" or more off at the start. With conventional scraping one thousandth of an inch is a lot to take off.

The method we have developed involves using an offhand angle grinder, small surface plate, and machinist's level. (See level and whitworth if you live in antarctica -fenn.) An Eastern European made level can be bought for less than $100 with 0.0005"/10" slope measurement per bubble division. This is what we use and recommend. The small surface plate can similarly be an import B grade 9"x12" size or a little larger . These sell for some tens of dollars. ($14.99 at enco)

The technique is to integrate level readings along the work to obtain straightness and twist information. We then analyze the data to lessen the effects of accumulated error.

Let's say the machine has a four foot 1"x3" solid steel hot rolled way mounting surfaces. We will try to make one edge (the 1" wide surface) precisely straight and flat.
First we will prep the material a little. It will be lumpy, bumpy, twisted, and bowed. Get the thick Mill scale off with dilute acid or a rough snagging with the angle grinder. Then take a reasonably flat piece of metal used as a sanding block to go down the surface with 80 or 120 grit paper. We are not going to hand sand this flat... we are just using it as a marking tool. Take a few swipes and note the sanding scratches on the high spots. Get the angle grinder and blast away at these. Really bear down ... this is just rough prep.
Repeat this cycle a few times until you have ground off the really nasty bumps. Perhaps a third of the surface will have been ground.

Now it is time to level the surface. It is typically already on the machine frame so we will shim the base. Put the level on the surface. It will rock and the bubble will be pegged almost every where you try. Keep shimming until there are several spots where you can get a reading that is not pegged. Then firmly attach the base to the floor. It should not rock or bend when you clamp it. I usually temporarily mortar some blocks with embedded studs to a concrete floor to put the machine on. It does take a while for the fresh mortar to stabilize. Drilling holes for studs in old concrete is quicker. Patch them later.

At this point the level will often rock in the narrow direction from twist and rock in the long direction in places from bow. We must control this by making it rest on the work kinematically. This just means it is like a three legged stool- it can't rock. Most of the levels have four feet near the ends so if you always bias the rock in the narrow direction the same way it will rest on three of the four feet. If your level has a flat base you may need to affix it to a metal sub base with 3 contact points . This does not have to be accurate because it is kinematic. I have clamped the level to a 10" piece of ground stock and then epoxied three little buttons (I used just some scrap) to the bottom. If the top of the block is not ground just put a little epoxy putty on the top then aluminum foil. Gently press the level into the foil and clamp with padded u-bolts or similar. Note that the level does not have to be precisely calibrated. You can adjust the vial if you want.

Now we will take readings along the surface every level length. My level is 10 inches long so that is my increment. The positions should overlap just enough so one end rests on the spot the other end rested on with the previous measurement.

Start at one end and write these readings down. Probably many will be pegged and you won't get a reading. (Pegged means the measurement is outside the range of the level.) But, write down the position and bubble reading for the ones that do. Now look at the pegged readings on either side of the good ones. Is the surface going up or down? If the pegged reading indicates low Grind the nearby area with the good reading. If it reads high grind the pegged reading area. What if you have one good reading area with it pegged low on one side and pegged high on the other? Grind both the good area and the pegged high area.

Repeating this a few times will expand the areas where you get a reading. They won't read zero- but they won't be pegged. As a check you can take the small surface plate and slide it upside down along the surface. High areas will allow the plate to easily rotate when moved a little. It is easy to feel.
Don't drop the plate! I tie it with a safety rope or stack boxes or wood along the surface so it can't fall off. Sometimes I grind an area that feels high with the plate.

A note about grinding technique... use short heavy strokes. We are taking off huge amounts of metal. We may end up taking off a pound or so! This will tend to make the surface barrel shaped. We usually tend to take off more near the edges. This may reduce our kinematic support, so make sure the center area is a tiny bit lower than the edges. Often I just take the grinder with a light pass down the center every few cycles to make this so. I can check it with a small ground scrap to see if it rotates if nudged as before. The center should just be a thousandth or two lower all the time but it is not critical as long it IS lower.

At some point all the readings will be good (not pegged) . You have been getting an idea how much you are taking off with varying degrees of pressure on the angle grinder. If readings get worse lighten up some. You are getting calibrated. At this point you may begin to realize that you can take off one thousandth of an inch controllably with an angle grinder!

Now get some graph paper or a computer plotting program. Mark off the reading position horizontally and the bubble reading vertically. Start at 0 on the left. Let's say you have an "up" bubble of 2.5 divisions. (You can easily figure which side of the vial is up or down) Draw a line from zero to +2.5 at the first reading increment (10inches for me). Now the next reading... let's say a down bubble of 5 divisions... and plot it from the LAST vertical position. So in our case we have a line from 0,0 to 10" , +2.5 to 20", +2.5 -5 or -2.5. so at 20" we are actually lower than the start. We are INTEGRATING or adding the slopes to get the profile.

When you are finished you will have a graph that probably looks like the Himalayas. But let's take a close look. It may mostly just go up along the length, or it may trend down. Well that is just how we shimmed the base.
At this point you can reshim it if it is really bad. If it were 50 divisions high at one end then that end is 25 thousandths high. But you don't have to perfectly level it because we are going to go for a straight graph line-not neccesarily a flat one. Try to do a best fit straight line through the curve with a pencil and straight edge. If you are plotting on a computer do a linear regression line.
What is the worst DIFFERENCE? That is what we will be trying to correct.

Now we will start the regular grinding cycles. Each time we will plot the curve comparing it to a best fit straight line and hand grinding the high spots. But at this point we will have a second plot to keep track of-TWIST.

For this we put the level crosswise on the way. It may have to rest on a scrap block. Again we will have mostly pegged Bubbles. Grind off the high side as before. When we get areading everywhere we can plot in a similar fashion.

Note that when we work on twist we will have messed up the length measurements some. But the stock is narrow. That reduces the interaction a bit.

I generally alternate between four or five length measurement cycles for every twist cycle. I mark spots to be ground early on with a magic marker. I use hatch for light grind pressure, cross hatch for heavier pressure, and solid for really leaning on it pressure.

An important note- if you are working hard enough to get the way warm cool
it between cycles with some air or just time. It will temporarily bend a WHOLE lot from having the top warmer than the bottom.

It takes a while to get calibrated. If it gets worse after a cycle use less pressure. If it is going too slow use more.

This method has one problem... it is prone to accumulating error. Just one bad reading will make the whole rest of the part appear low or high. And there will be bad readings. How to we minimize this? The first way is just scrutinise that line. If the data says we ADDED metal or removed it where we didn't grind then we have an error. Repeat the measurement a couple times. How consistent is it?
Another way to spread out the error is to measure starting from the other end alternately. The graph will be displaced vertically but it should be the same shape.

At the point where we are getting twist and length measurements improving we will add the surface plate to measure every two or three cycles. This time it is coated with a thick layer of prussian blue marking paste. Slide it down the way.The high spots will be clearly marked.
Grind only those areas on this cycle. This step also prevents sampling error from measuring at discrete (10") intervals. We want the areas in between to be flat as well!

So- that is the basic technique. It is a combination of scrutiny and sweaty work with a grinder and lugging a chunk of granite. It seems complicated at first- and the first results are often a surface that gets worse.

But once you get calibrated it works pretty well. If I have been doing it a bit I can get an eight foot way measuring straight and flat to a thousandth or so in one very hard days work.
The final surface will still feel can easily see and feel a one thousandth high spot. But even with a few bearing points per square inch you have a good surface for mounting linear slides.

The trick is that an angle grinder with a 36 grit wheel can snag off huge amounts of metal and also remove well under 0.001" with the right technique.

When it's done right ANYWHERE you put the level you will get the same bubble reading over and over. And remember... those divisions are 0.0005" in 10!

Les Watts


Below are several pdf files on hand scraping. They were copied from some of the Yahoo groups. When I find which groups I will post acknowledgements.

Attach file: file2648-Scraping.pdf 43 download [Information] file2633-Scraping.pdf 50 download [Information] file05-Duplex_Scrapea.pdf 57 download [Information] file2652-Scraping.pdf 55 download [Information]

Front page   Diff Backup Reload   List of pages Search Recent changes   Help   RSS of recent changes
Last-modified: Thu, 14 Dec 2006 19:07:52 GMT (1489d)
The Gingery Machines wiki is published under the
Creative Commons Attribution-Share Alike 2.5 License Creative Commons License