Report on Testing of the Belleville Spring Stacks in the Instrument Latch Life Test Fixture to Determine Preload

Sloan Digital Sky Survey Telescope Technical Note 19991214

Larry Bartoszek, PE Bartoszek Engineering


A report was written previously by Larry Bartoszek about the calculated force from a stack of Belleville washers (disk springs). This report documents the experimental results obtained when these stacks of disk springs were tested in the Instron machine in Jay Hoffman's lab at Fermilab.

The point of these calculations and tests was to determine how far the spring stack had to be compressed initially such that when the instrument latches latched on the life test fixture at Fermilab they would exert more than enough load equivalent to supporting the weight of the SDSS camera against various accelerations. The design goal was to have at least 700 lbf in the stack at the full deflection of the instrument latches. Calculation showed that an initial compression of .158" with the latch in the latched position produced a force of 800 lbs. The springs that were tested were the same size and manufacturer as used in the test fixture at ICB during the life testing of the latches at Fermilab. The actual springs used on the camera were a different size and were modified at Princeton. Appendix 1 gives the data for these other springs tested in the Instron. Princeton used measurements made by Jim Gunn and Mike Carr at the telescope to set the preload for the actual camera.

Testing showed that the number to achieve the 800 lbs was closer to .150" of pre-compression, but some aspects of experimental error would lead one to conclude that the number could be slightly higher

Discussion of Data:

Many experiments were run on several sets of springs. Some tests were run lubricated, some unlubricated. The following graph is a summary of many runs with springs unlubricated. The graph shows the force measured on the test machine versus travel of the machine head. One can see that the very first test, series A, and one other were anomalous, with the rest of the data falling within a small band.

The chart also shows that the data looks relatively flat at the beginning of the test as the backlash in the system comes out. For this reason, the data was corrected by subtracting out the flat part of the curve at the beginning to create a set of "delta" curves. The delta curves were used to predict the force of the springs as a function of crush.

Figure 1: Raw test data for the same set of springs, no lubrication

When the data is corrected for backlash and the anomalous tests are thrown out, the chart looks like the following figure:

Figure 2: Test data for the same set of springs, no lubrication, corrected for backlash in the test system

The springs were then lubricated and subjected to the same tests over again. The raw data is presented in the following graph. Notice that the first run after the test setup was taken apart, lubricated, and reassembled, showed another anomalous curve when compared to the remaining curves. It is hypothesized that there was some break-in of a stack at initial assembly, but how this is correlated with behaviour at the real telescope is not known.

Figure 3: Test data for the same set of springs, lubricated with molybdenum disulfide, uncorrected for backlash

When the data is corrected for backlash and the anomalous tests are thrown out, the chart looks like the following figure:

Figure 4: Test data for the same set of springs, lubricated, corrected for backlash in the test system


Many important things were discovered during the latch and spring testing. Perhaps the most important thing learned was the influence of the hardness of the bolt that the belleville washers were riding on. Many latch tests were confounded by the fact that these springs would dig grooves into any bolt that was less hard than a grade 8 bolt. Once a groove was worn into the bolt, the springs would catch in the groove and not unlatch fully. Grade 8 bolts in the test fixture solved this problem.

One thing to note about the delta curves is that at zero deflection, the force is not zero, but about 25 lbs. This is an example of some of the experimental error encountered in the testing, but which is mirrored in the setup of the latch test fixture. It was hard to get zero initial load on the Instron, and hard to know exactly when the stack is at zero compression during the tightening of the bolt in the test fixture. It was felt that it was better to have slightly higher load, than slightly less than desired load. On the Instron, there was always about 10 pounds of preload on the stack when the load was zeroed out prior to making the compression test. On the test fixture, the bolt was always a little snug producing slight preload before the real tightening happened.

Another point to note about the delta curves is that lubrication made slight difference to the scatter in the data, but the basic curves are not that different for lubricated and unlubricated springs. The test fixture was always assembled with lubricated springs. The graph of Figure 4 was what was used to set the preload in the life test fixture.

Appendix 1: Data from the actual springs used in the camera

We can see from this that the force/deflection curve for the actual springs is somewhat different from that of the springs used in the test fixture. These springs are stiffer and there is significantly more scatter in the data, especially at the upper end of force. They achieve a preload of 800 lbs at about .120 inches of deflection (unfortunately right where the scatter starts to get pronounced.)