Fiber Harness Test Results
Sloan Digital Sky Survey Telescope Technical Note
19970403
Matthew Buffaloe, Russell Owen and Walter
Siegmund
University of Washington
Contents
Introduction
The Sloan Digital Sky Survey operates in both imaging and
spectrographic modes. In the spectrographic mode, 640 optical fibers
transport light from a plug-plate mounted at the telescope surface of
best focus to a pair of spectrographs. The optical fibers and the
plug-plate are part of an assembly called a spectrograph cartridge.
Nine spectrograph cartridges are being constructed. This will allow
plug-plates to be plugged and unplugged during the day and rapid
exchange of cartridges at night.
A total of 5760 optical fibers are required to populate the nine
cartridges. To facilitate assembly, testing, record keeping and
maintenance, the optical fibers are organized into fiber harnesses,
each containing twenty individual fibers. On one end, fibers are
terminated in individual 2 mm diameter plugs. On the other end, all
twenty fibers terminate in a single V-groove block.
C Technologies, Inc. of Cedar Knolls, New Jersey, manufactured 322
fiber harnesses. This is enough to populate nine cartridges and to
provide spare harnesses. A fiber tester was developed that
illuminates the plug end of each fiber with an f/5 beam (the f-ratio
of the 2.5-m telescope) and measures the amount of light exiting each
fiber in an f/4 beam (the f-ratio of the spectrographs). It is
calibrated by mating the light source module to the detector module
with no intervening optical fiber (Figure 1). Fibers are tested by
plugging them one at a time into the source module. A motorized stage
properly positions the V-groove block with respect to the detector
module for each fiber (Figure 2). A microscope built into the light
source module allows the plug end of each fiber to be inspected for
contamination and flaws. Also, the position of the fiber relative to
the illuminating spot can be checked (Figure 3).
Two fiber testers were assembled. This allowed each harness to be
tested by C Technologies as manufactured and by the University of
Washington upon delivery.
Figure 1: Fiber tester calibration.
The light source and the detector modules are mated at the left
end of the optical rail. In this configuration, the light is coupled
directly to the detector with no intervening optical fiber. This
provides a reference value for subsequent fiber measurements. Other
features of interest are the linear stage (right end of optical
rail), the stabilized lamp (behind and coupled to the light source
module via an optical fiber), the pico ammeter and the linear stage
controller (on shelf above lamp).
Figure 2: Testing fibers. One of the twenty
fibers in a harness is plugged into the light source module. The
V-groove block, at the other end of the harness, is mounted on the
linear stage. The linear stage moves the V-groove block in order to
align the transmitting fiber with the detector. Upon completion of
this measurement, the next fiber is plugged into the source module
and the process is repeated.
Figure 3: Visual inspection. Russell Owen
uses the microscope built into the light source module to inspect the
end of a fiber for contamination or damage. This is done for each
fiber before it is tested. Even a small amount of contamination
affects the throughput of a fiber.
Results
The throughput distribution of fibers measured at the University
of Washington is tight (Figure 4). Fused silica has a refractive
index of 1.46 at 550 nm. Consequently, fiber throughput cannot be
higher than 93% because of the dielectric reflection losses at the
two ends of the uncoated fiber. This accounts for the abrupt cutoff
on the right (Figure 4). The standard deviation of the distribution
is such that if the distribution were Gaussian, no measurements would
fall below about 90.5%. Instead, a long tail on the left is apparent.
This indicates a non-Gaussian distribution exists.
Figure 4: Measured throughput of the fibers
in 318 harnesses that satisfied testing by the University of
Washington. The sharp cutoff at 93% is due to dielectric reflection
losses at the two air-silica interfaces. A long non-Gaussian tail on
the left side of the distribution is apparent.
Figure 5: Measured throughput of the fibers
in 321 harnesses that satisfied testing by C Technologies. The
smaller standard deviation of the University measurements may be due
to a difference in the cleaning technique used.
Figure 6
: Throughput
scatter plot. Each point represents one of 6220 fiber measurements.
The throughput measured by
C
Technologies
is plotted versus that measured at the
University of Washington. This plot is consistent with nearly all
fiber throughputs being intrinsically identical with scatter
associated with measurement error. A few fibers have lower throughput
in measurements made by
C Technologies or the
University but not both
. We attribute this to
contamination or stress that was present for only one measurement or
mismatching of fibers between the two measurements. The last group of
fibers was measured to have significantly lower throughput by both
groups. These are presumably intrinsically low-throughput fibers.
Less than 1% of the fibers fall into this category.
In that absence of measurement error and variable contamination or
stress, measurements performed by C Technologies and the University
would be identical and would fall along the diagonal of Figure 6. The
scatter seen in the Figure is consistent with most fiber throughputs
being identical and the scatter being due to measurement error. About
1% of the measurements fall on the diagonal trend line below and left
of the main cluster. Presumably, these are fibers with intrinsically
lower throughput. Also exceptional are a few fibers with
significantly lower throughput measured by C Technologies or the
University but not both. These may be attributed to contamination on
the ends of the fibers or stress that affected only one measurement
or mismatching of fibers between the two measurements.
The fiber harness contract specifies that the mean throughput of
the fibers of each harness be greater than 90% and that the
throughput of the worst fiber be better than 87%. The best fibers in
each harness approach the 93% maximum possible transmission discussed
above (Figure 7). The histogram of throughputs for the worst fibers
in each harness (Figure 8) shows that only a few harnesses have
fibers that are measured to have less than 90% throughput. (The data
plotted in Figures 7 though 10 were obtained at the University of
Washington. Only those harnesses that satisfied the throughput
criteria based on UW measurements are included.)
Figure 7: Maximum throughput. The
throughput for the best fiber in each harness is plotted. The
sharp cutoff at 93% is due to dielectric reflection losses at the
two air-silica interfaces. The data plotted in Figures 7 though 10
were obtained at the University of Washington.
Figure 8: Minimum throughput. The
throughput for the worst fiber in each harness is plotted.
Harnesses are rejected if this value falls below 87%.
Figure 9: Average throughput. This
graph shows the mean fiber throughput for each harness. Harnesses
are rejected if this value falls below 90%.
The mean throughput for each harness is very consistent (Figure
9). The standard deviation of the means is 0.22% and the range for
318 harnesses is only 2.1% peak-valley. Almost all of the harnesses
measured have mean throughputs 1.0% or higher than the specification
of 90%. The histogram of standard deviations for the harnesses
(Figure 10) shows that the harnesses have fairly uniform throughput
with few outliers.
Figure 10: The standard deviation of
the fiber throughputs for each harness is plotted. The throughputs
for most of the harnesses are very uniform, i.e., the standard
deviation is less than 0.5%.
Conclusions
- 318 fiber harnesses containing 6360 fibers were found to
satisfy contractual throughput specifications when measured at the
University of Washington. The mean throughput of the fibers of
each harness is greater than 90% and the throughput of the worst
fiber is better than 87%.
- Nearly all harnesses that were measured to satisfy
specifications by C Technologies were found to be satisfactory
when remeasured at the University of Washington.
- The great majority of harnesses that satisfied specifications
exceeded those specifications by a significant margin. No evidence
exists of significant skewing of the mean throughput distribution
by selection or rework of failed parts.
- The results for most harnesses are consistent with fiber
throughputs being identical to a few tenths of a percent. Scatter
is attributed to measurement error.
- A few harnesses have larger standard deviations than nominal.
It appears that only one outlier is responsible in most
cases.
Acknowledgments
We are grateful to John Phelps of C Technologies, Inc. for
supplying measurements of the fiber harnesses made by personnel at C
Technologies.
Date created: 4/16/97
Last modified: 12/30/98
Copyright © 1997, 1998 Walter A. Siegmund
Walter A. Siegmund
siegmund@astro.washington.edu