In another thread someone asked about how to set up parts that are truly square. As in not just assuming that their machinist's square is actually square. But rather to try to dig down more and use the measuring tools we have as home machinists to make and prove a shop made master square without spending the money on a granite square or similar.

I suggested in that thread that a flat plate could have the four edges milled on their milling machine and with a sensitive dial gauge or dial test indicator with a suitable stand could measure for square using basic principles. To that end I posted this sketch...

The idea is that a thin but not too thin to be easily tested and used plate is bolted down so all four edges can be machined with one setup using the X and Y travels. And that the initial results would be a reading of how square the two axes are to each other. And how, if needed, the square could be put back on the mill and with any error known that the one side could be set up with a slight out of true amount which compensates for the measured out of square to true up the square by skimming the error away.

So the first step was done tonight. I cut out some 1/8 plate and drilled two holes. Some 1/4 pieces I have in my assorted clamping goodies drawer served as risers to lift the plate up so I could mill the edges. I chose to relieve two edges which would be lapped if needed and left the other two plain. The roughing cleanup and relieving was done in a few minutes and then a fine .002'ish climb cut to give a nice finish was done on all four edges. Here it is finished and just before I removed the hold down nuts.

So after deburring and a couple of light passes on a stone to ensure there were no burrs sticking out of the edges the plate sat easily vertical with no help on my surface plate. Apparently I leveled the benches fairly well....

First step in the testing was to check the opposite sides for parallel.

With the two relieved edges down on the plate the full edges were tested and I was a bit pleasantly shocked to see that they were dead straight AND parallel to the plate with only a slight waviness that was never more than +/- .0001 to .00015. My DTI's are both .0005 per division so a tenth is what I call a needle's width of travel where no white shows between the set 0 and the needle. A sliver of white I call a tenth and a half.

Anyway. with the parallelism for both pairs of opposite sides proven it was time to flip the plate 90 degrees and test the two vertical edges for leaning.

Things aren't as good in this department. What I found was a difference of .0030 +/- .0001 from one side to the other. This side to side test doubles the true error. The reading which is the true square value is half way between the low and high values as seen from each side. Distance between the bumper on the base to the finger of the DTI is 3.2". So it looks like my X and Y axes are out of square by .0015 over that 3.2". Not a big deal for a lot of things but if I'm doing anything more fussy I'll have to keep that in mind.

Next up is to put the plate back in the mill with the .0015 fudge factor indicated on the one axis or the other and skim the other two edges to see if I can get it either dead square or at least a lot closer to where I can lap the difference in or just use the middle reading for the true square value.

More as it happens. Might be tomorrow evening since I've got to go out for the day.

I made the square this small because it was a handy bit of scrap I have. And it's also a size I can use to check lesser squares such as my cheaper combination squares and the wood working squares. So it'll have a use other than proving it can be done and to tell me that my mill's table ways are not quite square to each other.

The whole process so far has taken about an hour and a quarter. If nothing else it's worth that just to know what my mill is slightly out of square so I don't go chasing fairy dust at some future time when I try to rely on the X and Y squareness and I find the error.

I suggested in that thread that a flat plate could have the four edges milled on their milling machine and with a sensitive dial gauge or dial test indicator with a suitable stand could measure for square using basic principles. To that end I posted this sketch...

The idea is that a thin but not too thin to be easily tested and used plate is bolted down so all four edges can be machined with one setup using the X and Y travels. And that the initial results would be a reading of how square the two axes are to each other. And how, if needed, the square could be put back on the mill and with any error known that the one side could be set up with a slight out of true amount which compensates for the measured out of square to true up the square by skimming the error away.

So the first step was done tonight. I cut out some 1/8 plate and drilled two holes. Some 1/4 pieces I have in my assorted clamping goodies drawer served as risers to lift the plate up so I could mill the edges. I chose to relieve two edges which would be lapped if needed and left the other two plain. The roughing cleanup and relieving was done in a few minutes and then a fine .002'ish climb cut to give a nice finish was done on all four edges. Here it is finished and just before I removed the hold down nuts.

So after deburring and a couple of light passes on a stone to ensure there were no burrs sticking out of the edges the plate sat easily vertical with no help on my surface plate. Apparently I leveled the benches fairly well....

First step in the testing was to check the opposite sides for parallel.

With the two relieved edges down on the plate the full edges were tested and I was a bit pleasantly shocked to see that they were dead straight AND parallel to the plate with only a slight waviness that was never more than +/- .0001 to .00015. My DTI's are both .0005 per division so a tenth is what I call a needle's width of travel where no white shows between the set 0 and the needle. A sliver of white I call a tenth and a half.

Anyway. with the parallelism for both pairs of opposite sides proven it was time to flip the plate 90 degrees and test the two vertical edges for leaning.

Things aren't as good in this department. What I found was a difference of .0030 +/- .0001 from one side to the other. This side to side test doubles the true error. The reading which is the true square value is half way between the low and high values as seen from each side. Distance between the bumper on the base to the finger of the DTI is 3.2". So it looks like my X and Y axes are out of square by .0015 over that 3.2". Not a big deal for a lot of things but if I'm doing anything more fussy I'll have to keep that in mind.

Next up is to put the plate back in the mill with the .0015 fudge factor indicated on the one axis or the other and skim the other two edges to see if I can get it either dead square or at least a lot closer to where I can lap the difference in or just use the middle reading for the true square value.

More as it happens. Might be tomorrow evening since I've got to go out for the day.

I made the square this small because it was a handy bit of scrap I have. And it's also a size I can use to check lesser squares such as my cheaper combination squares and the wood working squares. So it'll have a use other than proving it can be done and to tell me that my mill's table ways are not quite square to each other.

The whole process so far has taken about an hour and a quarter. If nothing else it's worth that just to know what my mill is slightly out of square so I don't go chasing fairy dust at some future time when I try to rely on the X and Y squareness and I find the error.

## Comment