No announcement yet.

Qualitative flatness testing, cheap. (pics)

  • Filter
  • Time
  • Show
Clear All
new posts

  • Qualitative flatness testing, cheap. (pics)

    This will be posted in three sections to accommodate the pics so if you get here before I finish please don't post until I finish the three posts.

    I had a lot of questions about how I managed to get the mill table that I just anodized as flat as it is. I won't go into all the details but it's a matter of a lot of careful work and comparing to a known accurate reference such as a surface plate the old fashioned way with layout dye. I use acrylic artist paint. I spent quite a few hours on it since it is one of the single most important parts of a mill. If the table is off then so is everything you do on the mill.

    I thought I would show how to do an easy and inexpensive qualitative test of surface flatness. This is an optical non-contact test and gives an overall indication of the degree of flatness. It does not give a truly quantitative result although that can be inferred.

    It also does not give zone information although with modification that is also possible.

    This method relies on a cheap laser level as a light source. Although I use a surface plate qualified to 0.0001" as a reference, in this test it isn't necessary once you know what to look for.

    In pic 1 is the test setup. The laser level is on the right, the test surface just to the left of it (surface plate) and on the far left of the bench is the white screen.

    The laser level is set to project a horizontal line. It is adjusted so that the line just grazes the surface of the test object with the angle of incidence nearly zero. The angle and height are adjusted so that it is possible to split the thickness of the line with the far edge of the test object. A portion of the beam illuminates the screen directly while the rest of the thickness of the beam is reflected from the surface at grazing incidence.

    Note that at grazing incidence it is not necessary for the surface to be a reflective polished surface or even metallic. It merely needs to be fairly smooth.

    There are three main considerations in this setup. You must be able to adjust the angle of incidence, the height of the laser and the parallelism of the beam to the test surface.

    If the beam is not parallel to the surface it will result in this condition shown in image 2.

    When properly adjusted in all axes it will look like this:

    In most of these images a six inch scale is included.

    Image three is the signature produced by my surface plate which is certified to .0001" flatness.

    The bottom line is projected directly on the screen and represents a reference as it is a portion of the top edge of the projected light from the laser that overshoots the test object. The top line on the screen is the portion that is reflected from the test surface. Because of the extremely low angle of incidence the entire test surface is illuminated and the entire surface contributes to the reflected image that composes the top line. This is obtained by careful adjustment of the height and angle of incidence of the laser.

    The angle of incidence is adjusted so that the beam sweeps the entire surface of the test object. When it is correct the separation between the top and the bottom lines on the screen will be about equal to the thickness of the original laser line multiplied by the ratio of the distance of the laser to the center of the object and the distance from there to the screen. In this setup it is about 5 to 1.

    By sprinkling a bit of salt on the surface plate it can be seen in image 4 that the entire surface is illuminated by the beam.

    Because of this a multiplication of any errors of slope is obtained. In this setup it amounts to about 240 times. This is arrived at from the ratio of the width of the laser line, about .1 inch, compared to the length of the area it illuminates, about 12 inches. This gives a factor of about 120 but is doubled because the angle of reflection is opposite and equal to the angle of incidence. That gives 240. The amount of multiplication will vary depending on the angle of incidence. If the angle of incidence is kept the same for other test objects the multiplication will remain the same. The lower the angle of incidence the greater the multiplication. The angle of incidence must of course be greater than zero.

    The amount of error in slope on a surface depends on what distance that error covers. A concavity of .001" over ten inches produces much less error in slope (about 1/10) than the same .001" over a distance of one inch.

    Part II next post.
    Last edited by Evan; 02-16-2010, 09:29 AM.
    Free software for calculating bolt circles and similar: Click Here

  • #2
    Now that the setup is adjusted for use we can proceed to test some surfaces. The first is a section of 1/4" thick aluminum mill plate. It hasn't been polished or otherwise altered, only the edges have been machined.

    It is apparent with the rule that there is a slight gap in the center. It measures with brass shim stock to be about .002" bowed downward.

    In the laser signature it is obvious that this piece isn't flat as this error shows as a much thickened top line compared to the bottom reference line. Not all parts of the laser light are reflected to the same part of the screen because of the slight curve in the part.

    The thickness of the top line is about .5" which corresponds to the .002" error magnified around 240 times.

    Part III next.
    Free software for calculating bolt circles and similar: Click Here


    • #3
      The next part to be tested is a piece of 3/4" tooling plate. This is pretty flat stuff and an eyeball check with the rule confirms this.

      But, it isn't all that flat. A check against the surface plate revealed that it is bowed slightly, about .0004" in the middle. The laser signature confirms this as it produces a double line.

      Last is my mill table. There are some gaps and spurious reflections produced by the edges of the four rails that make up the table. Also, the very ends of the center rails (last 1/4") are very slightly turned down by a few tenths from the finishing process. In all though it compares very favorably with the surface plate and approximates it in flatness. Keep in mind that any out-of-flat condition in side to side or end to end of each rail relative to the others would result in a horizontal or vertical misalignment of the line segments.

      Free software for calculating bolt circles and similar: Click Here


      • #4
        Evan: You should really send this to Canadian tire to show what can be done with one of their laser levels. Heck, I'll bet that you get a gift certificate or better you can be the new Canadian Tire pitchman.



        • #5
          This is the most interesting thread I have read since becoming a member on this forum.I know I need to reread it to make sure I understand what is going on.I have never seen or heard of this before.



          • #6
            Hello Evan,
            Thats impressive ingenuity.


            • #7
              Evan -

              Thanks for posting this. Yet another good reason to visit the BBS every day!

              Frank Ford


              • #8

                That is very interesting, I think I understand what you wrote and the pictures. Is this something you have used (mirror grinding?) elsewhere or did you "head scratch" the "problem" to get a method from "applied physics".

                If you get a chance, maybe you could make a "simple" CAD drawing, in profile, "showing" the laser. I am unclear as to location of the laser, is it more or less bisecting the "y" axis at 90*, or is located toward a corner, "shooting the diagonal" ?

                Do you "need" to rotate the table 90* and check again?

                Sorry I am not asking this well tonight.
                Today I will gladly share my experience and advice, for there no sweeter words than "I told you so."


                • #9
                  Damn Evan, that's sweet...I've been waiting for a post like this since


                  Been playing around w/ a Horrible Fright laser level when time permits and I've bought a couple optical reference books and an article on collimation. It would be real nice to have a series of articles on the history of optical precision and how the now ubiquitous laser diode opens up new turf. I'd renew my lapsed subscription asap!!!

                  That post as is looks like it would be a great Machinist Workshop little time so much to learn. Thanks!!! Great photos too, as always.


                  • #10
                    This technique isn't something I have seen done before although I would be highly suprised if it isn't used somewhere. I have played around with optics since I was a small child and lasers since shortly after they were invented. My father had many interesting goodies in his science lab and I encountered many more helping him at the Lawrence Berkeley National Lab.

                    This I thought up from my knowledge of first principles of optics. It seems to work pretty well and is easy to apply. Here is a schematic of how it works. It depends entirely on several things.

                    1: The laser beam is composed of a bundle of nearly parallel rays.

                    2: The beam has a finite non-zero diameter.

                    3: It is essentially monochromatic.

                    In the diagram the top part of the beam (red) overshoots the surface and produces the bottom reference line on the screen. This is accomplished by carefully adjusting the height of the laser.

                    The middle of the beam (green) is shown striking a sloped surface near the end (highly exaggerated). As the angle of reflection is always equal and opposite to the angle of incidence the beam is reflected upward. The bottom of the beam (blue) strikes a different part of the surface since the beam has a non-zero diameter. Because the surface isn't flat the angle of incidence and therefor the angle of reflection is different. This produces a spread on the screen that is directly proportional to the degree the surface isn't flat. If the surface is flat the angle of incidence is the same along it's length. There is a small cosine error involved due to laser divergence but it is negligible.

                    Last edited by Evan; 04-15-2006, 09:50 PM.
                    Free software for calculating bolt circles and similar: Click Here


                    • #11
                      I think thats really great. I should probably buy the issue of the magazine if you publish this, but, since I already subscribe.........
                      mark costello-Low speed steel


                      • #12
                        Interesting Evan, I am always impressed with your work.

                        So the table is all nice and flat, but how is the parallelism between the table and the way surfaces?
                        Last edited by mochinist; 04-15-2006, 11:04 PM.


                        • #13
                          Evan, thanks for the great post. Randy, good idea, but based on my experience at crappy snappy, there is no one there that would understand one word of it
                          in Toronto Ontario - where are you?


                          • #14
                            Hi Evan... Looks good cept... It looks like you are checking the "flatness" of the top surface optically. Optically your checks can zero in to what appears flat but the ratio seems abit low for optical measurement systems.

                            Not really my question though.

                            It seems you are able to quantify the top surface but how is it put in relation to the bottom surface as far as parallelism is concerned? Where is the "squareness" derived?

                            Looks good and I enjoy seeing "out of the box" thinking. JRouche


                            • #15
                              It seems you are able to quantify the top surface but how is it put in relation to the bottom surface as far as parallelism is concerned? Where is the "squareness" derived?
                              The thickness of the table is an entirely different matter and this doesn't measure it. My mic does and it is within a thou or so. That's what shims are for where it mounts to the linear bearings. BTW, aluminized mylar sold as "space blankets" or emergency blankets make excellent shims .0005" thick. It is very dimensionally stable, conformal and mylar has less than 1% water absorption.


                              As for the sensitivity of the system, it seems to be sensitive to around .0001". That is good enough for most of my work.
                              Last edited by Evan; 04-16-2006, 12:50 AM.
                              Free software for calculating bolt circles and similar: Click Here