Report on the SDSS Latch Life Tests and Operating Characteristics
Sloan Digital Sky Survey Telescope Technical Note 19991228
Larry Bartoszek, PE Bartoszek Engineering
The instrument latches on the Sloan Digital Sky Survey Telescope hold up the CCD camera and are required to be 100% reliable over the lifetime of the Survey, a design life of 10,000 latch cycles. To achieve this reliability the latches were tested in the laboratory using a programmable controller, a fixture to load the latches, and an air supply to simulate conditions at the telescope. The test latch was tested to 8000 cycles at room temperature and 2000 cycles in a freezer to simulate winter on the mountain. This ratio of cold to warm testing was thought to match the likely ratio on the mountain given fewer observable nights in winter due to weather.
This report summarizes the results of that testing and what was learned during the testing process. Ultimately, six latches were delivered to APO that had all the final design changes incorporated, were tested and found reliable. The original log book of the testing is available through SDSS management.
AutoCAD drawings of the latch and test fixture are referenced in this report. They are in DWF format, which requires the Whip! plug-in for Netscape and Internet Explorer, available at Autodesk. Clicking on any of the jpeg images of AutoCAD drawings will start the plug-in and allow the reader to zoom in on any feature of the CAD file that may be inaccessible or indistinct in the jpeg image. Using the "back" button on your browser returns the reader to the report.
Overview of Test Plan:
Testing of the instrument latches happened in three major phases. The first phase was room temperature (referred to as "warm") testing on the camera simulator. The second major phase was cold testing of the latch on the camera simulator in a freezer. The third phase was actually not a life test of the latch as much as it was a life test of the stirrup mechanism that supports the plug cartridges. The latches are used to support two completely different pieces of equipment alternately, the CCD camera and the Plug Cartridges. The camera weighs more than the plug cartridges so it loads the latch more heavily and is the main reason for the reliability requirement. The reason for testing the plug cartridge stirrup was to guarantee its reliability so that the survey would not be impacted by problems in this component during cartridge changes.
The assembly drawing of the latch is in Figure 1. The fixture that loads the latch by the same mechanism that the camera uses is shown in Figure 2. The fixture used to life test the Plug Cartridge Stirrup is shown in Figure 3.
Figure 1: Assembly of the instrument latch.
Figure 2: Assembly of the latch life tester. The numbers shown in this drawing for the spring compression are not the ones used during the life test. Proper spring compression was determined by another test. A report is available on the Belleville Spring Test at: need link to other report here.
Figure 3: Assembly of the Plug Cartridge Stirrup Tester.
Discussion of Problems Encountered During Testing:
A series of problems was encountered and solved all during the latch load testing to end up with the final reliable product. The latch design evolved from the original concept developed at Princeton, and what was first tested was Bartoszek Engineering's redesign of the latch that attempted to recycle as many of the original components as possible. The testing determined that many of the original components were not adequate to achieve the reliability, so they were gradually replaced over the course of testing. What follows is a brief description of the changes that happened due to problems discovered during testing.
Proper Latch Pre-load--Setting and Maintaining
The load that the latch hook sees upon latching was entirely determined by the amount of preload a stack of belleville washers was put under by a 1/2-20 bolt in the test fixture. It was originally thought that the load could be set with the latch in the unlatched position, but this was later found to be incorrect because the stirrup in the camera simulator could be positioned by two small screws in the bottom which took away the relationship between the loads in the latched and unlatched configuration. The most important thing to realize about the latches is that they are fixed stroke devices. They must exert enough force to change from the latched to the unlatched state, but the stroke that the hook goes through from one state to the other is always fixed. Because of this, the force on the latch has to be set by adjusting the preload in the springs with the latch in the latched position.
The preload was intended to be between 700 and 800 pounds to make the latches hold three times the weight of the camera. The set of the springs was determined by the Belleville spring test mentioned above. Once the correct spring compression had been determined by the Instron test all the latches were tested with a compression of .130 inches, corresponding to a load of 700 lbs. 800 pounds was doable, but required more air pressure than 95 psi and was not thought necessary.
A perrenial problem that occurred throughout the test was that the 1/2-20 bolt would periodically loosen from the vibrations, or the belleville springs would wear grooves in the bolt and then get stuck, changing the preload that the latch hook saw. This problem was solved by adequately tightening a lock nut on the 1/2-20 bolt, and by replacing the bolt with a grade 8 one that was hard enough to resist the grooving of the springs wearing on the bolt shank.
Component Material Problems
The original guide rollers inside the latch were oil-lite bronze. The oil-lite bronze was discovered to be unable to take the forces of latching and would develop grooves or deformations that eventually caused the latch to jam and misfire. The rollers were all replaced with stainless steel. To reduce friction between the stainless rollers and axles, Glacier DU bushings were introduced between them. Latch forces are high enough to cause plastic deformation of the DU bushings which tends to elongate them, but this was discovered in testing and the bushings were shortened so that they couldn't get long enough to jam in the mechanism.
The original latch housings were aluminum. These were replaced with cold drawn steel coated with Wear-Cote Plus CFx to make them corrosion and wear resistant. The latch hooks were originally also coated steel, but the plating flaked off in contact with the pusher roller and the flakes were considered hazardous to the reliability. The hooks were remanufactured from stainless steel.
Component Geometry Problems
The latch hook profile was kept the same as the Princeton original during the steel hook testing, but after it was discovered that the profile had sharp edges that were destroying the pusher rollers, they were modified to the current profile to lower forces on the latching mechanism.
Some of the early latch failures could be traced back to insufficient flow of air through the solenoid valves in the controller. These were replaced with larger models that solved this problem. The larger models never needed replacing for the remainder of the testing.
Moisture Problems and Other Air Supply Problems
Moisture in the air supply lines became a big problem during testing in the freezer where it would condense and freeze the lines shut. We were eventually forced to switch from using shop air from a compressor to dry nitrogen in bottles ganged together. Another problem with shop air was that the pressure required by the latch was close enough to the limits of the compressor on/off dead band setting that the pressure would drop too low for reliable latching before the compressor would cycle on. This made the latches look unreliable when it was really an air supply problem. We were eventually forced to log the air supply pressure on a chart trace to track this problem down. The gas bottles also solved this problem.
After many trials, the conclusion was reached that the latch would never latch reliably significantly below 95 psi. An original design constraint was to operate at 80 psi, but this could not be met. The air system at APO was upgraded and the latches have worked reliably there (at lower pressures than they would operate at in the lab).
Another condition that did not lead to problems in the lab, but was thought to be a problem for the camera was that the latching and unlatching were not gentle. A distinct banging was happening at the latch. Eventually, flow restrictors were put on the air cylinder ports to throttle the flow rate down and make the latches move with force but lower speed.
Switch and Mag Sensor Problems
Periodically throughout the testing the number of unlatch signals and commands would exceed the number of latch commands. This was thought to be caused by switch bounce in the microswitch that indicated latch hook retraction. This was a problem that came and went and was never completely understood, but did finally go away with the addition of flow restrictors to the air cylinders. It seems to have been mainly associated with the "intermittent" mode of operation of the controller. In "I" mode, the controller turns off the supply of air to the cylinder as soon as it gets a signal from the microswitch that the latch is in the desired state. Before the flow restrictors were put it the speed may have been high enough that the latch hook bounced in and out of contact with the microswitch causing the controller to send more unlatch signals. As long as the number of unlatch commands matched the number of unlatch signals, and same for latch commands and signals, these tests were allowed to be final.
A problem associated with the magnetic proximity sensor that indicates the latched state is that it is adjusted and locked into a dovetail groove on the outside of the air cylinder with a small pan head screw. This screw has no means of being locked and did loosen on a couple of occasions. The sensor has been glued in place with RTV on the final latches.
Latch Spring Problems
The first latch retraction spring problem that we had was with a spring that had gone many thousands of cycles on the test latch and suddenly the loop on one end cracked off. Metallurgical testing determined that the spring had fatigue failed and the failure could have been caused by a tiny imperfection on the drawn wire that created a stress riser. This was written off as a fluke. It was deemed too expensive to examine every spring with a microscope, and there was too much area hidden inside the coils of the spring to make it worthwhile. A report is available about this failure from Bill Boroski.
In October of 1998, Corrector Latch #1 was returned to Fermilab from APO because it would not signal that it was unlatched and the hook stuck out slightly. It was determined that the latch retraction spring had been over stretched. It was kinked and about one and a half coils longer than a new spring. This was a failure that had never happened in the lab, but was traced back to the fact that the retraction springs could move back and forth on the roll pins that hold them in place. It was possible for a spring to move far enough off center that it could be pinched between the latch hook and housing during retraction. The solution was to design some bronze sleeves that would keep the spring centered. A web report is also available about this failure at NEED LINK HERE.
Abbreviated Summary of Test Results:
On 2/18/98, the test latch was cycled at room temperature to 8247 cycles with no lost counts. The air pressure was set to 95 psig
On 2/20/98, the latch test fixture was installed in the freezer. 2/21/98 was when the broken retraction spring was discovered. The test latch was cycled in and out of the freezer after reassembly several times to check for reliability. Pressure definitely needed to be turned up while the test was in the freezer (between 103 and 110 psig). There were several tests with failed latching that were probably due to water in the air lines.
By 3/17/98, all changes had been made to run dry nitrogen through the tester in the freezer. Successfully ran 2398 cycles at -18.2 degrees C.
The final version of the Plug Cartridge stirrups were tested on 10/11/99. They successfully latched and unlatched at least 400 cycles at room temperature. We did not test these cold.
The test latch was sent out to APO as a spare, but its housing needs the chamfer cut that all the rest of the latches have.
Summary of Latch Operating Requirements:
The latch needs dry air at 95 psig for room temperature.
The reliable operating pressure in the lab was about 105 psig when the temperature was at -18C.
The magnetic proximity sensor indicates the latched state.
The microswitch on the back of the latch indicates the unlatched state when it is closed.