02-01-2018, 10:51 AM
Well we thought about it and decided that our next step should be to test the repeatability of the measurement post roll. The reproducible nature of the roll and repeatability of the measurement will determine just how many things we can test for. If the post roll measurement is +/- 20-30 we can likely test for a lot of different factors. If the post roll measurement is +/- 50 we will likely be able to test for fewer factors.
We conducted our tests using our new radius surface impactor at several different locations on the same edge. The results were something on the order of +/- 60. Not good on a scale of 300-350 (the extent of roll we intend to induce) . What is good is our edge tester's ability to point to and describe the extent of these very small edge inconsistencies. We want to do better than +/- 60. This kind of resolution/repeatability would only be useful in determining the grossest differences in edge structure. Of course, averaging of the readings would improve the error range from test to test. Statistically speaking, a group of four readings would reduce the error by 1/2. We would like to minimize the use of and reliance on averaging in these tests. From our electronic days we learn that the way to reduce the need for measurement averaging is to improve the signal to noise ratio. In this case, an uneven roll represents "noise". If we homogenize the roll, our "noise" will be reduced and the ratio improved. It is obvious from these tests that the roll produced by the current impactor design is not homogenous across the impact area. In retrospect, one look at the photos we took previously confirm that. Here is one example:
The roll in the above picture appears not to be even. Depending on where the test media impinges on the edge, a wide range of readings may result. In view of the fact that the entire rolled section of the edge comprises a length along the edge of some .020" (20 thousandths) of an inch, predictable test media placement is impractical. We must, simply, improve on how we roll the edge. The better we get, the more information that can be derived from the instrument.
We have ideas and are already at work on them.
We conducted our tests using our new radius surface impactor at several different locations on the same edge. The results were something on the order of +/- 60. Not good on a scale of 300-350 (the extent of roll we intend to induce) . What is good is our edge tester's ability to point to and describe the extent of these very small edge inconsistencies. We want to do better than +/- 60. This kind of resolution/repeatability would only be useful in determining the grossest differences in edge structure. Of course, averaging of the readings would improve the error range from test to test. Statistically speaking, a group of four readings would reduce the error by 1/2. We would like to minimize the use of and reliance on averaging in these tests. From our electronic days we learn that the way to reduce the need for measurement averaging is to improve the signal to noise ratio. In this case, an uneven roll represents "noise". If we homogenize the roll, our "noise" will be reduced and the ratio improved. It is obvious from these tests that the roll produced by the current impactor design is not homogenous across the impact area. In retrospect, one look at the photos we took previously confirm that. Here is one example:
The roll in the above picture appears not to be even. Depending on where the test media impinges on the edge, a wide range of readings may result. In view of the fact that the entire rolled section of the edge comprises a length along the edge of some .020" (20 thousandths) of an inch, predictable test media placement is impractical. We must, simply, improve on how we roll the edge. The better we get, the more information that can be derived from the instrument.
We have ideas and are already at work on them.