View Full Version : How do you measure to the edge of a tapered feature?

12-15-2003, 09:21 AM
In making the blank for the dodecahedron, I had to measure from the end to the point where the tapered section meets the cylindrical section. I also had to measure the diameter of the flat area on the end, and the length of the cylindrical section.


All of these measurements involve features which won't "catch" the calipers, and using a scale seems iffy.

What's the proper way to make a measurement like this accurately? Can it be done without taking the part off the lathe?


12-15-2003, 09:29 AM
Well, a couple of ways to do it would be a toolmaker's microscope or an optical comparator, but most of us don't have those....

12-15-2003, 10:11 AM
If your trig is good put dowel pins over the angle with the flat & pins on a flat surface then mic over the pins.


12-15-2003, 11:12 AM
Don't dismiss the use of a scale too quickly, by the way. Use a 5X or 10X magnifying glass, and measuring to quite good precision suddenly becomes a lot easier. Given that the project at hand is of the "if it looks right, it is right" variety, that may be good enough.

Of course, there's nothing wrong with wanting to be fanaticial about it to half a thou, either....

12-15-2003, 11:55 AM

First you need to define "accurately. Do you need it to .010, .001, .0001?

You can set it on end on a surface plate and measure the heights with a height gauge and a magnifying glass. You may want to put some color on the part and then make a slight mark with the scribe on the height gauge to help determine the break point at the tapers.

Once you have the heights, you can use a precision protractor and determine the angles. By measuring the center diameter and using the angle along with some trig. you can determine the diameter of the ends.

You can also use a rule as has been suggested and lay it across the end and view it under a magnifying glass unless you need measurement to .0001.

Also as has been suggested, an optical comparator will do great job, but -- most of us don't have one or a CMM.

That's a really great project. One that would warm up the old brain cells to figure out how to do it.

Thanks for the post.


12-15-2003, 12:30 PM
The CAD program gave me dimensions to two decimal places. I did the trig and got them to three places. I'd like to get them to a few thousandths, but in some places I was introducing the reading error at each end. For instance, holding the scale on the round end after cutting the taper.

Marking the surface before cutting is a good idea. Since I don't have the other stuff, I'll try that and a magnifying glass.

My wife, who rarely ventures into the shop, was suitably impressed when she saw the lathe in action with that fixture. "Wow, no wonder you enjoy tools so much."


[This message has been edited by winchman (edited 12-15-2003).]

Paul Alciatore
12-17-2003, 10:59 PM

1. Any CAD program should do the math internally to at least 4.5 places (16 bit = 1 part in 65K). The really good ones work to twice that precison (32 bit math = 1 part in about 4,000,000,000) or over 9 decimal places. You should check the settings in your CAD program and find out how to get better numbers out of it. They are definitely in there.

2. Once a part like that is machinned, optical magnification is definitely the best way to measure it. A good scale can read directly in hundreths and with a 10X or 20X magnifier you can easily read to a fourth or fifth of that. That's within 2-3 thousanths. The magnifier would also be indispensible if using a height gauge or any other device.

3. A technique I learned a longtime ago might also help. A measurement can be made to about 10X better resolution than the scale used if the scale is randomly placed on the object many times and the readings are simply taken to the best accuracy of your ability to the edges of interest with NO ATTEMPT TO ZERO THE SCALE ON EITHER EDGE. So readings are taken for both the start point and the end point of a measurement. The differences between two numbers of a single readings are taken and recorded. After you take 10, 20, or more such readings, they are averaged. The average will be much better than any of the individual measurements.

This technique depends on the random placement of the scale and therefore the random averaging of your errors. It also can eliminate any bias you may have in your observations of the location of the edge or in interpolating between the lines of the scale. But use the magnifier here also.

Paul A.

12-18-2003, 03:13 AM

Seems like uncommon "Common Sense"!

Thanks for sharing.


12-18-2003, 03:35 AM

That is the Monte Carlo Method.