Precision square tester

[rhb]

For ordinary workshop tooling the typical specs are:

5" Sine bar:

roll diameters  0.0001" x 2  =  8 arc seconds
roundness  0.0001" x 2 = 8 arc seconds
parallelism 0.0001" = 4 arc seconds

Gauge blocks

2 B grade blocks 2*0.00005 = 0.0001" = 4 arc seconds

B grade surface plate 0.0002" = 8 arc seconds

The total assuming all tooling is in calibration is 32 arc seconds  tolerance range. 

No.  One does not simply add individual uncertainties in a linear fashion to get the overall uncertainty.

Oh, master.  Enlighten me.  Perhaps you would be so kind as to explain the correct calculation.   Where did you come upon such great wisdom?  It is such a mystery to me.

So far all you have done is complain about a 30 second response to a question.  Do you even own a set of gauge blocks, sine bar and surface plate?    I've not seen your cal certificates yet.  In fact, all you have done is intrude to pronounce what I wrote is wrong, but without supplying the correct answer.  I think you  are a pretentious fraud.  You should go watch "Magnum Force" and pay attention to what Harry Callahan says near the end when the car blows up.

But I *do* think your choice of user name highly appropriate.  You really are a vegetable.  Be careful lest I slice and dice you for my salad.  The vast majority of my friends have PhDs from Stanford, Austin and Mines.  And most of the people whom I can call by name have PhDs from a major school.

You're new here.  You need to understand that there are a substantial number of PhDs both papered and common law who hang out here.

So do yourself a favor.  Go away before I decide to get unpleasant.

[tomato]

Oh, master.  Enlighten me.  Perhaps you would be so kind as to explain the correct calculation.  And cite the appropriate  references.  Where did you come upon such great wisdom?

The combined standard uncertainty is a quadratic sum of the uncertainties of individual input variables, each weighted by the (squared) partial derivative of the output function with respect to each input variable, evaluated at the nominal values of the input variables.

Here is one link:

https://www.nist.gov/pml/nist-tn-1297-appendix-law-propagation-uncertainty#EqA3

So far all you have done is complain about a 30 second response to a question.

I'm not complaining about anything, simply pointing out an error of a metrological nature.

Do you even own a set of gauge blocks, sine bar and surface plate?

Yes.

I've not seen your cal certificates yet.

I'm not sure what that would prove.

In fact, all you have done is intrude to pronounce what I wrote is wrong, but without supplying the correct answer.

See above.

I think you  are a pretentious fraud.  You should go watch "Magnum Force" and pay attention to what Harry Callahan says near the end when the car blows up.

Sorry, not a Clint Eastwood fan. 

But I *do* think your choice of user name highly appropriate.  You really are a vegetable.  Be careful lest I slice and dice you for my salad.  The vast majority of my friends have PhDs from Stanford, Austin and Mines.  And most of the people whom I can call by name have PhDs from a major school.

I'm very happy to hear that you have lots of friend with PhDs. I'm sure they are all very nice people. But, it doesn't negate the fact that you posted something that is incorrect.

You're new here.  You need to understand that there are a substantial number of PhDs both papered and common law who hang out here.

I'm sure they are all very nice people as well, and very knowledgable.  But it doesn't change the fact that you posted something that is incorrect.

So do yourself a favor.  Go away before I decide to get unpleasant.

I think you're already there.

[rhb]

In a conversation with Conrad I mentioned a technique for fabricating an arc second class reference square which he had not heard of.  It's not original.  I learned it from some 100 year old book.  All I've done is update the recipe for modern tooling and materials.  As it was unfamiliar to him I thought I'd share it.

Take six pieces of 1/8" x 1" ground tool steel only slightly  longer than the center height of your lathe.  Stack them and tack weld the ends.

Set up a blank plate on the faceplate of the lathe, faceoff and turn a shallow groove between the center and the outside.  Then check the flatness of the center relative to the outer ring with a straight edge.  Most lathes are set up to face off with a slight concavity.  Using the compound adjust the rings to match a  0.001`" straight edge.  Make sure to let the plate cool off before taking the final cut.

Set up on the plate to bore  a hole at each end of the stack of bars on the lathe.  Drill and ream to a light press fit for a hardened steel pin.

Set the pins pins at each end and saw off the ends so that the distance from the center of the pin to the end is less than the distance to the sides.

With the pins in place, harden the bars.  Then setup and grind one face of the stack with the pins resting on parallels.   Assemble with a top and bottom pair and the verticals in between and the ground faces outward.   Attach fine pitch turnbuckles on the diagonals.

Set up a tenth indicator on a stand with a hard stop near the base.  Place this against the verticals and adjust the turnbuckles to split the tenth from one side to the other.  Done carefully, that will give about 5-10 arc seconds of accuracy independent of the actual accuracy of the indicator.  The errors will be due to the pin and hole diameters and roundness and the accuracy of the parallels.

Once the square is adjusted, fit insulated handles so it can be picked up without touching the metal parts.  Then make a box to store it in.

If you don't have access to a grinder, leave the bars unhardened and scrape the stack to a surface plate.

This should also make clear why I'm interested in producing the same result using a crude square and electronics.  Properly the above belongs at Practical Machinist, but I thought it an interesting contrast to my electronic version.

As an exercise in gratuitously accurate bench work, I worked a 0.75" x 1.0" x 3.125" aluminum block to plane and parallel to 5-7 tenths on the 1" x 3.125" faces using files, scraper and lapping on wet or dry sandpaper on a surface plate. This will get cut up tomorrow to form the fixtures for gluing the sensors to 8d nails,  lapping the faces of the QRE1113s and a proper  mount for testing the sensors.

[Conrad Hoffman]

Got fascinated with the making of a decent square, so I fooled around in the shop and the result was yet another web page- http://www.conradhoffman.com/refsquare.htm
This is the first cut, so let me know if any gross errors stand out.

[rhb]

Nice job.  I wish I had a grinder.  I did notice one issue:

"I made my square out of common cold rolled bar. "

Did you anneal it before grinding?  If not, it is likely to warp over time. Generally hot roll is preferable for such uses.   A lot of writer's I've read recommend an initial machining pass to remove the scale, annealing and then finish machining.

I did most of the work on  the mounting frame for my inclinometer today. It's made from CRS tubing, so I'm trying to figure out the best way to anneal it.  I wish now I'd bought some SS tool wrap and wondering if a bunch of aluminum foil would reduce the oxidation.

No progress on the square tester.  I need to hunt down some clear acrylic sheet I've got and make the fixture for gluing the sensors on to the 6d finish nail heads.

Edit:  Low carbon steel which is what is generally is used for cold worked stock needs 875-925 C to fully anneal it.  So back to work on the square tester while I wait for some SS tool wrap to arrive.

[Conrad Hoffman]

I'd be surprised if it moves, but I do have a small kiln for heat treating. The stock was very old, so maybe its done whatever it was going to do. I'll keep an eye on it- the beauty of the simple block is I can check it any time curiosity gets the better of me. 

BTW, my grinder was near to free and is as ugly as they come, but with some TLC I brought it back to life. For some reason surface grinders are common and have little resale.

[rhb]

It will be interesting to see if it does move and if so how much.  A lot depends upon how much material you removed.  Make a record of the readings on all the faces.

In one of the "Machinist's Bedside Readers" Guy Lautard has a story about someone who wanted a long precision straightedge and came to the shop with a length of CRS.  He was told that was not suitable stock, but he persisted.  When they unclamped it from machining it warped like mad.

Not a lot of manufacturing where I am, so as a consequence, not a lot of used machine tools for sale.  It's very different in the Northeast.

I had not intended to revisit the sine bar kerfuffle, but I came across my copy of "Handbook of Dimensional Measurement" 3rd ed by Farago & Curtis.  So out of curiosity, I looked up what they had to say about sine bars.  I think this one page sums it up nicely and amply justifies my pessimism as to what a hobbyist can expect from cheap tooling.  Note how tight the laboratory grade sine bar specs are and the expected accuracy stated.  My sine bar was made in India.   I doubt my sine bar would meet the commercial spec given here. Almost all my tooling is cheap because that was what I felt I could afford for a hobby. And what was not originally cheap is old and of unknown accuracy.

I don't mind being told I'm wrong.  In fact I encourage it.  But I do expect an explanation with sound logic and references where appropriate. Not a blanket "That's not the way you do that."

[Conrad Hoffman]

Good pdf- thanks! I suspected that 45 degrees was about the max. My sine bars are certainly not lab grade, and may not even qualify for commercial. One is a Fowler, the other I've not a clue. On the good side, I can measure most of the things that matter, and have yet (in 40 years) to need angles known better than a few arcminutes. Since getting a set of angle blocks, the sine bar hasn't seen much (any) use.

It would surprise me if the block of cold rolled moves enough to detect. The aspect ratio is block-like, nothing like the aspect a straightedge would have.

I had another thought on the reference square. If one wanted a bigger baseline, it would be easy to start with a piece of iron pipe, maybe 4" diameter or so. There would be plenty of metal to grind a 1" wide flat down two sides and follow the same procedure as for the block type. Being a pipe, it wouldn't be crazy heavy like a solid cylindrical square. Hey, I'm all about cheap sources of material!

[rhb]

That's an interesting idea.   You've got me wondering if you could make a cylindrical square on a surface grinder.

I'm all about cheap materials and simple tooling.  Take a look at my tiltmeter construction post today.  And many thanks for pointing out the differential screw.  I'd have thought I should have heard of that, but I had not.  I think I may have seen one, but not realized what it did.

While I wanted to do that with minimal tooling, it was a lot of work.  But I felt the practice would be useful and I wanted to make clear that you do not have to have a machine shop for such projects.  I've now done enough flat filing that I can again feel when the object is not flat but has a curve to it.  It's very subtle.

I still like the adjustable box square concept I outlined.  A 10" version adjusted to split a tenth would be accurate to 1 arc second.

I'm quite amused by how many people in an electronics forum have fairly well equipped machine shops.

Off to work on the optical sensor jig.

[rhb]

Well, I constructed a better test setup, destroyed the sensor I had so carefully lapped and made another, but skipping the lapping part.

I'm seeing >10% drift over the course of a few hours.  Putting the bench PSU in current source mode helped a lot with noise, but I'm beginning to think that, while extremely sensitive, these sensors are too inaccurate.   If it is the tempco, then 1-2 F is causing 10% drift.

The room will cool down overnight, so I should find out if it's temperature related or not.  But it is increasingly looking like a dry hole.

Edit:  It's beginning to appear that this is an initial operation aging effect rather than a thermal effect.  There was no increase in current at starup this morning and it has continued to go down all day, but by much less than yesterday.  The rate is also reducing from 4 uA in 2:40 to 2 uA in 3:40.