Saturday, September 8, 2012

Squaring techinques on the milling machine

A year or so ago I started corresponding with one of my readers about a problem he encountered. Mitch posed a question that I thought was of general interest to machinists and metalworkers and I have never seen it described in any book. Mitch has a business making gun leather products. And in his spare time also has his own machine and welding shop were he does a little unwinding. His problem was related to cast iron right angle plates. Similar to this one,
The problem he was encountering was the plate as it came from the manufacturer was not to specification for squareness and parallelism or at least not as good as he wanted within his ability to check. We communicated back and forth via email to discuss how he could measure it with the equipment that he had available in his shop. His first clue that the plate had a problem was just checking it with a high quality combination square and seeing some light between the blade. The part that made the problem intriguing was Mitch didn't have any squareness standard to compare the angle plate to for a quantitative measurement. Its all pretty easy to evaluate if you have a nice Taft Pierce cylindrical square or a squareness master like Hermann Schmidt combined with a squareness gage. Incidentally none of these is worth a hoot for inspection to a known limit without some calibration and traceability to an accepted reference standard.

For fun I checked one of my twenty something year old Starrett combination squares with a setup exactly like the one pictured above when I had access to one of these setups. I was pleasantly surprised that in ten inches which was the maximum I could measure to with the squareness gage the combination square showed less than .002 over the ten inches. Pretty good for a square that has been dropped in the weld shop more times that I care to admit.

All Mitch had to do his analysis were some basic machinist measuring tools and a decent milling machine. This got me thinking about how you would go about measuring and repairing a damaged, out of spec angle plate with only basic measuring tools and the machine that you use the plate on. It hardly helps to have angle plates and standards ten times more accurate than the machine you use them on. The basic accuracy of the machine cannot be improved greatly by adding super precision accessories. In another article we will talk about accuracy, precision, repeatability and resolution. If you don't use sound machining practice and check your output you will never produce superior work. Believe it or not I don't think I have ever made an angle plate before so it turned out to be fun and enlightening for me which is just what I like in a good problem.
So to start I found a chunk of fairly heavy steel angle as the demonstration victim. The ends were just saw cut and were actually fairly square to the other faces of the angle. Right out of the gate you can see there are quite a number of interconnected perpendicular faces. A proper angle plate has all these faces at accurate right angles and parallelism to one another. Now this angle is smaller than the cast plates that Mitch was having a problem with but I figure its really the underlying principles were looking at not so much a step by step solution to his problem. The principles shown apply to many jobs found in the typical milling machine.
The next step was to verify that my machine was trammed as accurately as possible. For this job I intended to use the vise for all the work so I trammed the machine related to the bottom of the vise. Normally I like to use the machine bed as my reference for tramming. I ran the indicator around all over the surfaces I would be using. Everything looked pretty good with only a sprinkling of tenths showing anywhere.
So here is where I started. I chose this orientation for a couple of reasons. Three point contact is almost always good, and I wanted to use only the bottom cutting edge of the cutter. Side cutting introduces errors caused from tool deflection and the surfaces would have to be done in two setups. For small parts using the side of the cutter is fine and produces parts as square as the machine ways perpendicularity if done carefully. 

In the picture above I have the back saw cut edge supported on two gage pins on the more square side of the two ends. The first order of business is to establish two edges that are as parallel as I can make them. We wouldn't want to stray too far from whatever the average is compared to the other surfaces but you have to start somewhere. This is an important point to establish qualified reference surfaces that you can use to build on in the subsequent setups. Against the movable jaw of the vise you see a round rod clamping the angle. This is positioned to make the two vertical edges of the part against the fixed jaw sit in full contact with the jaw face and not have their alignment be influenced by the rough saw cut on the bottom.Incidentally this rod is the handle for a Starrett tap handle that I have been using for this purpose for years. I always feel a little pang of guilt using it this way but in thirty years the handle is no worse for the job. Now Murphy heard me and will cause me to bump an end mill into it and cement the guilt fully.
The cutter I'm using is a 1.5 diameter Sandvik inserted shell mill mounted on an R-8 arbor. I just skimmed enough material to clean up the edge surface fully. After de-burring the edges I lapped it a few strokes on the surface plate with 400 grit paper to remove any high spots or bumps.
This is probably the hardest setup to get right in the whole operation. The just machined surface is in the bottom of the vise sitting on three points. At the fixed jaw the angle is sitting on two gage pins and at the moving jaw it sits on a stack of gage blocks equal in diameter to the gage pins. I didn't have three gage pins the exact same diameter so I had to use the stack of blocks to make the third bearing point. Clamping the whole thing is a thin sheet of soft copper between the movable jaw and the workpiece. The goal here is to position the copper to make the angle sit down in the bottom of the vise and clamp the three gage points so they are all tight. By moving the copper up and down and by gentle seating the angle down in the vise the three gage points were equally clamped. This took a few tries to get it equal.
Work is only accurate if you verify it. After cutting the second edge I lapped the second edge a little on the surface plate on some paper so the indicator runs smoothly along the surface. Sweeping the surface showed less than .0005 no matter which way side was up. This was pretty good for the first two surfaces. In some cases on a job like this you might have to to a series of passes over all the faces improving all the surfaces with each round of machining. For vise work on a R-8 milling machine this was acceptable. If this was done on a high quality Jig bore or grinder I would expect something like half again better for the first cut.
In this picture the setup shows the preparation for surfacing one of the large faces of the angle plate. Remember those first two parallel edges? One of the edges is indicated vertically as close as possible. At this point it doesn't matter about the part of the angle that's in the vise. The vise jaws will just average whats there. What I care about it the reference surface related to the machine axis. 
After roughing the surface I added a little support under the under the overhanging part. The inserted cutter made a little noise in this area so I stabilized it. Hidden between the adjustable parallel and the angle plate is a strip of post-it note paper. This is so I don't have to use a lot of force pushing upward on the overhang to get the necessary damping to get a smooth cut with the shell mill. I could have also switched to a cutter with a more positive cutting geometry to eliminate the problem.
Here is the setup for the second large face. The first cut face is against the fixed jaw of the vise and the tap handle rod is assuring that the face is in full contact with the jaw face. The vertical edge is indicated again and aligned with the machine axis. 
In this picture you can see the adjustable parallel and paper used to eliminate chatter. 
After the large faces were machined the two remaining small edges were cleaned up.
Time for some verification. First check was across the faces that were cut last. This test shows the parallelism of the small edge to the face sitting on the surface plate. It read .0002 over three inches. I always advocate using as crude a measuring tool as possible in the early stages and progressively increasing the resolution as the interrogation gets finer and finer. A tenths indicator will drive you nuts if your probing for features in the .010 range and just adds unnecessary detail when its not needed.
Here is an example of getting the indicator to settle by tapping on the stem gently. Hysteresis in the measuring tool can be a source of inaccurate measurements and operator frustration. You will want to interrogate the part enough times to make sure the reading is repeatable. Measuring features this small is not easy. Just because you have the tools and trappings doesn't mean you can automatically reliably resolve down to such small limits. At this level everything affects everything else. Ask any grinder that has chased tenths around a part for a couple of days. Temperature, technique, and tools all play a part in taking accurate measurements. Don't fool yourself into precision. After all the measurements on this angle plate I would only realistically say that all the features have a high probability of being a little less than a grand (.001) related to one another in parallelism and perpendicularity to one another. If some other inspector says I did better its just a bonus. If the features had to be closer that this then I would have needed access to some different equipment and measuring tools.
Since I don't have a squareness gage anymore I checked the large faces the only way I could using a granite angle plate and an indicator. So what am I actually measuring? All I'm doing is comparing the plate I made to another plate that some other person made. I still don't know the current status of my granite angle plate or my surface plate for that matter. Fortunately these don't wear very fast but they do wear. So you can see the finer and finer the resolution you seek the greater the need for multiple comparative references and the need for calibration. In the end the two large faces showed about .0003 as compared to the granite plate. 

This is one way of skinning the cat. I'm sure there are machinists out there that use some other equally valid methods to achieve similar results. Keep in mind just because I happen to write it down does not necessarily make this the say all do all way of handling this problem. For each shop, person and toolkit there will be just as many ways to get the job done. This just happens to be the way I did it today. 

There is an excellent reference for further reading on the efforts and difficulty of achieving truly accurate geometry and relationships. Its an expensive book but required reading for folks trying to achieve very accurate work. Foundations of Mechanical accuracy by Wayne Moore is the classic text in the metrology and precision machine design.

I hope you found this article useful. I had fun doing it. And thanks To Mitch for providing a good problem.


  1. Tom, Interesting article and an even more interesting problem.
    I need to make a light duty "box" parallel, to calibrate a squareness comparator I plan to make.
    I will machine it then hand scrape it, but I'm wondering if there is really a difference between making an accurate angle plate or box parallel. Would it have to be set up differently?

  2. Hi Andre,

    I would go for the box section for your calibration standard if you're going to go through all that work. Its just a more stable shape. More surface area to scrap but heck its only elbow grease and free.



  3. Although I had planned to make a box parallel, I was rummaging through my friends scrap box (actually a to the scrapper box) and found some real heavy angle iron. about 3" long and the perfect piece for an angle plate, I've done the sides and one face so far, tapping it for 10 minutes or so with a plastic mallet to relieve stress. My mill is worn, so I will have to do a good bit of scraping after.

    Would a good condition Starrett V block standing on end with a Taft Pierce parallel bar clamped to it make a half decent angle reference? That's all I have right now.

    Really appreciate the article, and hope to see you do some scraping on your new machinist level you showed on your youtube page.

    Happy machining.

    1. Hey Andre,

      The vee block and parallel sound like they would work pretty well. It won't tell you quantitatively how much out of square it is but it would be a decent comparative square. Put the part in your kitchen oven at 400-500 F for an hour or two to help relieve stress before final machining.

      All the best and thanks for the comment.