The 0.0867" diameter bits were held by an custom-made aluminum collet. To minimize runout of the tool, the hole in the collet was machined while the collet was mounted in the NC mill. Scratches were made on the collet and machine so that the collet could be re-inserted in the same orientation. The collet was split using EDM to minimize burring from this operation. Runout of the collet was measured at 0.000,05".
All holes were drilled using a spindle speed between 3500 and 5600 RPM (exact speeds given later), and a feed rate of 5.0"/min. Reamers, when used, were fed at 10.0"/min. The runout of at least one of each type of bit was measured at the collet and was always found to be 0.000,05" (as expected).
During drilling, the plate was held in a custom jig flat against a backing plate. The backing plate had an over-sized pit under each hole in the test plate, so the bits never contacted the backing plate. After drilling the plates were run through an automatic dishwasher; this removed most or all metal shavings from the holes.
Non-circularity is defined as the difference in radius between the point closest to the hole's center and the point farthest from the hole's center. Note that this definition is based on the extreme measurements of the eight measurements at the given depth; hence it will tend to be dominated by noise in the measuring machine and residual dirt in the hole. The measuring machine introduces an average of approximately 5 µm of noise into non-circularity according to Robert Riley (one of the people made the measurements). Non-circularity is useful for comparing different drilling methods, but is not easily used to predict hole morphology.
The x-y position data for each plate was corrected for overall errors in offset and rotation, meaningless artifacts of the way the plate was mounted in the measuring machine. The correction was applied by fitting only the mid-level data because that was less prone to errors from hole damage. The ends of the holes on the sky-facing surface will actually determine the position of the optical fiber, but it has not yet been determined whether plates will be drilled from the sky side or fiber side. Note that scale errors were not removed (unlike in the 1/92 drilling tests).
The x-y position errors of each hole at top and bottom were divided by the z distance between these measurements to generate the tilt of that hole.
Method Pos. Err. Diameter Error Non-Circ. Tilt RMS mean std. dev. RMS RMS RMS (µm) (µm) (µm) (µm) (µm) (mrad) Twist Drill 7.5 1.3 5.6 5.7 10.1 3.2 Spade Drill 9.1 5.0 3.9 6.3 7.0 2.0 End Mill 5.6 74.7 55.9 93.2 7.8 1.6 Drill & Ream 6.9 4.7 7.3 8.7 8.4 3.0 Mill & Ream 9.7 5.0 11.1 12.2 10.2 6.1
Method Entry Pos. Err. Diameter Error Non-Circ. Tilt Angle RMS mean std. dev. RMS RMS RMS (deg) (µm) (µm) (µm) (µm) (µm) (mrad) Twist 0.00 8.7 0.9 8.3 8.4 13.3 5.2 Drill 0.25 8.8 0.4 8.2 8.2 15.4 4.4 0.50 6.0 1.5 2.4 2.8 5.7 1.2 0.75 6.8 1.9 2.4 3.0 5.5 1.2 1.00 7.0 1.7 2.4 2.9 5.6 1.2 Spade 0.00 8.2 4.7 4.0 6.2 6.2 1.6 Drill 0.25 9.4 4.9 3.7 6.1 4.6 1.6 0.50 8.9 4.9 3.6 6.0 8.2 2.0 0.75 9.2 5.0 3.9 6.4 5.2 2.1 1.00 9.9 5.7 4.2 7.0 9.5 2.4 End 0.00 5.4 71.2 54.2 89.4 7.6 1.5 Mill 0.25 5.9 72.5 56.4 91.7 8.6 1.4 0.50 5.0 74.0 55.8 92.5 7.8 1.7 0.75 5.8 77.3 56.5 95.7 7.6 1.6 1.00 6.0 78.3 57.0 96.7 7.6 1.8 Drill 0.00 5.1 5.2 4.0 6.5 5.4 1.6 and 0.25 5.3 5.5 4.0 6.7 6.1 1.4 Ream 0.50 5.0 5.6 3.9 6.8 4.8 1.5 0.75 10.4 4.3 10.2 11.0 13.9 5.3 1.00 7.3 2.9 10.7 11.0 8.4 3.2 Mill 0.00 2.6 6.3 4.4 7.7 4.5 1.2 and 0.25 3.2 5.5 3.7 6.7 4.5 1.3 Ream 0.50 3.1 5.5 4.0 6.8 5.3 1.5 0.75 19.9 3.8 22.7 22.9 14.8 12.8 1.00 7.0 4.0 7.4 8.4 15.2 4.4
Method Speed Pos. Err. Diameter Error Non-Circ. Tilt RMS mean std. dev. RMS RMS RMS (RPM) (µm) (µm) (µm) (µm) (µm) (mrad) Twist 3500 3.0 1.1 1.6 2.0 5.7 1.1 Drill 3500 3.7 -0.1 1.6 1.6 5.4 1.4 4500 7.5 0.3 7.9 7.9 13.1 5.2 4500 12.3 4.2 3.4 5.4 7.8 1.0 5600 7.3 0.7 8.0 8.0 14.6 4.4 Spade 3500 11.1 6.2 3.4 7.1 7.1 2.1 Drill 3500 5.2 3.1 2.7 4.2 4.6 1.7 4500 13.1 2.9 1.7 3.4 5.0 2.3 5600 3.5 7.9 4.5 9.1 10.0 1.7 End 3500 5.1 146.1 57.3 156.8 8.0 2.1 Mill 4500 5.7 73.8 45.1 86.4 8.3 1.7 4500 7.9 53.7 36.3 64.8 7.9 1.3 4500 4.7 64.4 33.1 72.4 7.4 1.4 5600 3.9 35.2 27.2 44.4 7.6 1.4 Drill 4500 7.6 2.0 9.5 9.7 8.8 3.1 and 4500 6.2 2.0 1.6 2.6 6.1 1.6 Ream 5600 4.7 5.2 1.7 5.4 3.8 1.1 5600 8.6 9.5 9.0 13.1 12.4 4.8 Mill 4500 17.7 8.4 20.9 22.4 11.6 11.5 and 4500 4.0 3.3 2.4 4.1 7.8 1.5 Ream 5600 5.9 3.7 6.0 7.0 13.3 3.8 5600 3.6 4.7 2.8 5.5 6.5 1.7
The results by method ( table 1 ) show that spade drills and twist drills both work very well. By comparison, end mills work very poorly. Reaming after drilling was slightly worse than drilling alone. This may be because the undersized holes were drilled conventionally (standard holder, standard quality bits). However, the 1/92 tests showed that drilling followed by reaming was inferior to simply drilling when all operations were performed with standard tools and techniques, so reamers may simply not work as well as plain drilling for such small holes. Reaming did help following end milling, but the results were still inferior to plain drilling with a spade drill or twist drill.
The results by entry angle ( table 2 ) show that spade drills suffer some increase in diameter error at 3/4 degree, and some increase in tilt error at 1/2 degree, but no noticeable loss of position accuracy at higher angles. By contrast, twist drills work poorly at low entry angles (a potential problem should we need to spot face the holes) compared to entry angles of 1/2 to 1 degree.
The results by run ( table 3 ) show too much variation from plate to plate to allow any firm conclusions as to the effects of spindle speed on hole quality. The variations are probably due to variation in bit quality particularly accuracy of centering of the point. Twist drills may work significantly better 3500 RPM than at higher speeds but further tests would be necessary to verify this.
The spade drills we used were 3/4" long from tip to collet, whereas all other bits were 1/2" long. The spade drills used include 5/16" of undersized shaft just above the blade; if the bit manufacturer is willing to supply bits without this region of undersized shaft we could hold the bits 1/2" back from their tip. Also, the spade drill blade itself could probably be shortened without ill effect, giving an even shorter bit. Shorter bits will probably make better holes and be less sensitive to entry angle.
Transverse Position Error µm RMS Astrometry 17 Transformation to focal plane 1 Scale, rotation and guiding 2 Hole location 9 (10) Temperature gradients 5 Plate deformation 5 Plug/fiber concentricity 8 Plug/hole concentricity 8 (23) Total transverse error 24 (32) Axial Position Error µm RMS Focus monitor 15 Registering surface 8 Temperature gradients 10 Plate deformation 25 Plug/fiber location 10 Plug/hole registration 12 Total axial error 35 Principal Ray Misalignment mrad RMS Hole drilling 2 (4) Plate deformation 10 Plug/fiber alignment 5 Plug/hole alignment 5 (16) Total alignment error 12 (20)
The plug/hole concentricity error was computed as follows. The hole diameter standard deviation is 3.9 µm (a factor of 3 improvement over the 1/92 drill tests). The plugs we are planning to buy have a diameter tolerance of +/-2.5 µm, which roughly corresponds to a standard deviation of 0.8 µm assuming the tolerance applies to 99.9% of the plugs and the errors are normally distributed. The standard deviation of clearance on the diameter is 4.0 µm (the plugs contribute very little). Assume normally distributed errors in plug and hole diameter and that we wish less than 1 hole in 1000 to require reaming; then the required average clearance on the diameter is 3.1 times the standard deviation, or 12.3 µm (ignoring non-circularity). In addition, we must include an allowance for hole non-circularity. The measure of non-circularity is hard to interpret (as explained in section 3), but as a crude estimate I propose including half the RMS non-circularity to the clearance (half because non-circularity is the difference between the closest and farthest points). The total non-circularity is 7.0 µm RMS, intrinsic machine noise accounts for approximately 5 µm RMS, so the non-circularity of the holes is approximately 5 µm RMS (using quadrature subtraction). The average clearance on the diameter is 14.8 µm (including non-circularity). The RMS variation in clearance on the diameter is the quadrature sum of the average and standard deviation, or 15.3 µm.
Transverse concentricity between plug and hole is half the clearance in diameter, or 7.7 µm RMS. Tilt is clearance in diameter divided by the length of the ferrule tip or hole, whichever is shorter. The proposed ferrule has a 2.87 mm long tip (excluding the chamfer), so tilt is 5.3 mrad RMS.
These methods are different than those used for the 1/92 report. The error budget below includes errors based on the 1/92 data but recomputed according to the methods described here. Transverse concentricity error was little changed, but the recomputed tilt error was twice that listed in the 1/92 report.
Based on the 11/19/92 fiber optic test results [R. Owen, unpublished], we would expect the principal ray misalignment of the plug to cause approximately 1% light loss. Light loss due to the transverse error will depend on the shape of the image, but the error is small compared to the core diameter of the optical fiber (180 µm), so the loss should be tolerable. The axial error is negligible because the light is incident on the plug at f/5.
However, we should review this issue when the optical design is finished. We will then know the range of entry angles and whether or not we will have to spot face each hole to set the plug depth. Spot facing would reduce the entry angle to zero (good for spade drills, very bad for twist drills) and offer the possibility of drilling a small pit to start the bit (which might improve either method). If the final drilling conditions are significantly different than those already tested we may wish to run one more set of drill tests. If we do run another set of tests, I suggest using shorter spade drill bits, if available.
