Direct Fiber Positioning System

© 2022-2024, Kevan Hashemi, Open Source Instruments Inc.




[30-MAY-24] Our Direct Fiber Positioning System (DFPS) uses the slight bending of a piezo-electric cylinder to displace an optical fiber at the end of a tube. Each fiber positioner consists of a cylindrical, piezo-electric actuator, a hollow tube that acts as a mast, and two or more optical fibers held in a ferrule at the tip of the mask. Accompanying the positioner are controller electronics that generate the actuator's ±250-V electrode voltages. One of the fibers is a guide fiber, which we use to determine the location of the mast tip. The remaining fibers are detector fibers, which we use to collect light from celestial objects. The DFPS provides one fiber positioner per 5 mm × 5 mm square area. The control electronics for each positioner fit beneath its 25 mm2 footprint. All positioners can be adjusted independently and simultaneously with no increase in power consumption. We can construct an array of eighty thousand positioners using the same fundamental design as we would use to construct an array of eighty fibers. Positioners share power and serial communication with their neighbors, so the number of electrical connections required by a large array of positioners remains small.

Figure: Sketch of Direct Fiber Positioner. By exaggerating the bending of the actuator, this sketch above shows how the mast and actuator move the fiber tip.

During construction of the DFPS, we measure the location of each detector fiber with respect to its guide fiber. A pair of fiber view cameras (FVCs) within the DFPS enclosure locate every guide fiber with respect to a set of fiducial fibers distributed around the perimeter of the positioner array. Star-guide image sensors adjacent to the fiducial fibers allow us to deduce the location of the fiducial fibers with respect to the stars. By combining these measurements, we obtain the location of each detector fiber with respect to the stars.

Figure: Model of Direct Fiber Positioning System 4A (DFPS-4A). Enclosure dimensions 90 cm × 34 cm × 20 cm.

Because the DFPS guide fibers are separate from the detector fibers, we can back-illuminate them for location measurement at any time. Because the guide fibers have a larger numerical aperture than the detector fibers, we can view them from a location that is invisible to the detector fibers. The result is a self-contained fiber positioner that we can adjust continuously during obervation without disturbing the observation itself. The DFPS is self-contained, in that it requires no modification to the telescope beyond bolding the DFPS enclosure to the telescope's viewing aperture. It is therefore easy to install and to remove.

Figure: Sketch of Self-Contained DFPS-80A Front-End. Each mast provides a guide and detector fiber. Two fiber view cameras measure the location of the guide fibers. Guide sensors and fiducial fibers sit in the focal plane next to the detector fibers. Yellow: guide fiber emission cone. Orange: detector fiber viewing cone. Blue: incoming celestial light cone.

The DFPS is the front end of a multi-object spectrograph. The back end is a system of diffraction gratings, mirrors, lenses, and image sensors that creates and records the spectra of the light from each detector fiber. We are collaborating with the Astronomical Instrumentation Laboratory at Texas A&M University (TAMU) to contruct a prototype, fully-functional multi-object spectrometer for deployment on the two-meter telescope at the Macdonald Observatory. We will provide the fiber-positioning front end. They will provide the spectrograph back-end.

Our work on the DFPS between October, 2021 and April, 2023 was supported by a Phase I Small Business Initiative Research (SBIR) Grant, number 2111936, from the National Science Foundation.


[31-MAR-23] For convenience, we present the following glossary of terms.

PositionerThe combination of an actuator, mast, and controller that together move the tip of a fiber.
ActuatorThe piezo-electric cylinder that bends when we apply voltage to its electrodes.
MastThe long tube that acts as a lever arm to turn the bending of the actuator into translation of the fiber tip.
FerruleThe cylinder with a precision center hole that presents the polished fiber tip.
ControllerThe logic, converters, and amplifiers that generate a single actuator's four electrode voltages.
Base BoardThe printed circuit board that supports all the positioners of a single cell.
Service BoardThe printed circuit board that holds the fiber controllers for all the positioners of a single cell.
Detector CellA base board, its fibers, its positioners, its service board, and its controllers.
Detector FiberA fiber used to capture and transport the light from a celestial object.
Guide FiberA fiber used to determine the location of mast.
Dead ReckoningMoving a fiber to a desired position and keeping it there with no use of guide fibers.
Guide SensorAn image sensor at the edge of the positioner array that records the position of guide stars.
Fiducial FiberA fiber used to locate the positioner array with respect to celestial guide sensors.
Fiber View CameraA camera looking down on the fiber tips.
Front EndThe multi-object detector: fibers, positioners, and the fiber view camera.
Back EndThe spectrograph itself: we plug the fibers into it and it records spectra.
Table Glossary of DFPS Terminology.

The detector and guide fibers may be the same fiber, or separate fibers. The fiducial fibers are stationary fibers whose location we know with respect to the guide sensors.


[30-MAY-24] When we add up the masses of the galaxies that we observe through our telescopes and apply the principles of general relativity to the distribution of these masses, we predict that the universe should be expanding more slowly now than it did ten billion years ago. We measure how the universe has expanded in the past by examining the spectra of distant galaxies. According to these spectra, however, the universe ten billion years ago was expanding more slowly than it is now. This disagreement between our expectations and observations is one of the outstanding mysteries of modern cosmology.

One possible resolution of this disagreement is that our existing measurement of the expansion of the universe is inaccurate. Another possible resolution is that our understanding of the universe is flawed. So far, astronomers have examined the spectra of ten million galaxies. If they were to examine one billion galaxies, their measurement of expansion would be ten times more accurate. If this more accurate measurement agreed with our expectations, our mystery would be solved. If it disagreed, this more accurate measurement would be our best hope for obtaining a clue to solve the mystery.

In order to examine one billion galaxies, astronomers need a large telescope equipped with an instrument that can examine fifty thousand galaxies simultaneously with a single hour-long exposure. This instrument would need to perform eight exposures a night for two hundred and fifty nights a year for ten years to gather enough data. No such instrument exists, but its construction is one of the long-term objectives of the astronomical community. Astronomers call it the "Stage Five Spectrograph".

In order to examine the light from a distant galaxy, we place the tip of an optical fiber upon the image of the galaxy in the focal plane of our telescope. To build the Stage Five Spectrograph we must pack fifty thousand optical fibers into the focal plane, along with all the mechanics and electronics required to maneuver each fiber into position. With exceptional effort and great expense, we can build a telescope with a 120-cm diameter focal plane. Even with such a large focal plane, each optical fiber must fit inside a 5-mm square. The smallest pitch of any existing, operational positioner is the 7.2-mm pitch of the 4MOST positioner, which was deployed on the Subaru Telescope in 2008 and provides four hundred fibers. The smallest pitch of any prototype positioner, prior to our Phase I work, was the 6.8-mm pitch of the MiniHawk prototype, which was built by the Australian Astronomical Observatory in 2012. The largest existing positioner is DESI, deployed on the Mayall Telescope in 2020, which provides five thousand fibers on a 10-mm pitch.

These existing positioners are ingenious and effective, but they do not lend themselves to the construction of the Stage Five Spectrograph. None of them can provide fibers on a 5-mm pitch. All of them dissipate one watt per fiber during adjustment, which is tolerable for an instrument with five thousand positioners, but intolerable for an instrument with fifty thousand positioners. The MiniHawk positioner is the most compact of the existing designs, but each positioner requires four high-voltage signals to be delivered to it from outside the focal plane. If we put fifty thousand of them together, we need to bring two hundred thousand high-voltage signals across the perimeter of the focal plane, which is not possible with any existing printed circuit board technology.

Our Direct Fiber Positioning System (DFPS) is designed for large spectrographs. Each positioner fits in a 5-mm square and dissipates less than fifty milliwatts continuously, with no significant fluctuation in dissipation as the positioner moves. The positioners share power and control signals. An array of fifty thousand positioners needs only two hundred high-voltage signals, for which there is ample space around the perimeter of a 120-cm diameter focal plane. In our SBIR Phase I work, we built and tested an array of sixteen direct fiber positioners. Our four-by-four array occupied a 20-mm square footprint. Each positioner provided 10-μm precision and 10-μm stability in fiber position. Our positioners contain no motors, no magnets, and no sliding surfaces. Their only moving parts are the slight bending of a piezo-electric tube and the displacement of a carbon fiber mast. The positioner is mechanically simple and easy to produce in large quantities.


[30-MAY-24] From August 2022 to February 2023, we built and studied Test Stand Two (TS2). This test stand consisted of a sixteen-fiber array, which we call the DFPS-16A, mounted ion a gimbal with a fiber view camera and all services required to operate and monitor the DFPS-16A. The purpose of TS2 was to demonstrate that we can load sixteen positioners onto a 20-mm × 20-mm footprint, with all necessary control electronics and high voltage amplifiers, beneath the same 20-mm × 20-mm footprint. Each 4×4 cell of positioners must be constructed in such a way that another identical cell may be loaded on all sides, so as to continue the array with no gaps between the cells. Another purpose of TS2 was to demonstrate that our fiber controller logic and amplifiers are capable of delivering the electrical stability and precision required to place fibers with 10 μm precision, and do so while dissipating less than 50 mW per positioner. Our focus in TS2 was upon the assembly procedure, the mechanical design of the plates and circuit boards, and the design and assembly of the electronic circuits and the miniature connectors. In TS2 we used carbon fiber tubes for our masts for the first time.

Figure: Test Stand Two (TS2), Sixteen Positioners Loaded. For video, see TS2.mp4. Each positioner consists of a 40-mm actuator, 300-mm carbon fiber mast, two 1.25-mm diameter zirconia ferrules and an optical fiber running down the middle. The fiber view camera is above the fiber tips.

Loading the sixteen positioners onto their base board turned out to be particularly challenging. Our hope was to solder the actuator electrodes to pads on the base board with the help of solder paste and a surface mount reflow oven. But the joints we made this way were unreliable. We ended up having to hand-solder the positioners onto the base board, which was hindered by the fact that every positioner was fully-assembled with mast, and a polished 1.25-mm diameter ferrule at either end. We succeeded in loading all sixteen, but the alignment of the fiber tips was no better than ±1 mm, which meant that the fibers interfered with one another when we eventually set them to move across their full dynamic range.

Figure: Test Stand Two (TS2) Horizontal. We have rotated the gimbal to put the fibers in the horizontal orientation.

At the base of each positioner is a ferrule to which we connect another optical fiber, which in turn runs to a light injector. We can flash all sixteen fiber tips independently. The TS2 masts are carbon fiber tubes. When we rotate the masts into the horizontal position, we observe them to sag by ±350 μm with a repeatability of 10 μm rms. For pointing angles up to 45° from the zenith, we can predict the deflection of the fiber tips to better than 5 μm rms. Predicting the sag to 5 μm rms is useful as a diagnostic check of each fiber, but is not necessary to maintain the fibers on target. Our concern with the sag is that if one mast sags by 1 mm while another sags by 2 mm at the same orientation, we will be unable to compensate for sag by rotating the telescope. It appears, however, that the sag will be within ±50 μm for all masts.

Figure: Test Stand Two Electronics, Annotated. Showing (1) ±250V power supply, (2) fiducial fibers, (3) positioner fibers and wires, (4) camera cable, (5) backplane, (6) actuators, (7) fiber controllers, (8) LWDAQ root cable, (9) command transmitter, (10) LWDAQ multiplexer, and (11) thirty-six way contact injector.

Packing sixteen fiber controllers beneath our array of sixteen positioners requires a connector capable of maintaining isolation between ±250 V signals, and holding each fiber controller in place with its own insertion force, all within a 25-mm2 area. Only one company makes connectors that meet these requirements: Omnetics Connector Corporation in Minnesota. In the photograph below we can see pairs of these connectors mating at the top of each fiber controller circuit.

Figure: Fifteen Fiber Controllers Loaded on Service Board Beneath Backplane. The sixteenth controller is missing because we broke its plug off the service board. Each controller provides four ±250-V control voltages for one actuator. Kapton tape insulates the controllers from one another and acts as a label for their identifiers.

In the photograph we see a three-cell backplane board equipped with one service board. The fiber controllers are plugged into the service board. A flex cable runs from the service board up to the base board. The actuators are soldered to the base board. The flex cable can be bent at a sharp 90° angle where it emerges from both the base and service boards, thus allowing us to the 19-mm square circuit boards on a 20-mm grid.

Figure: Base and Service Board. On the left are sixty-four pins for sixteen actuators. On the right are sixteen connectors for fiber controllers.

The flex cable carries all sixty four high-voltage signals required by the sixteen actuators. The fiber controllers receive high-voltage power from the service board, as well as serial logic control, both of which they share with every other fiber controller on the same backplane. Each fiber controller is addressed with its own four-digit hexadecimal identifier. The base and service board layouts were challenging. We were able to make all the required connections using 125-μm track spacing and a rigid-flex circuit, but we do not believe we could route the traces for a 5×5 cell, and certainly not a 10×10 cell. The number of traces required increases linearly with the number of positioners on the cell, but the width of the flex cable increases only as the square root of the number of positioners. We believe that 4×4 cell to be ideal for the construction of a large array.

We made several mistakes during the construction of TS2. We tore off one of the connectors on the service board. After epoxying the remaining fifteen in place, we discovered that five had been deprived of their ground connection. We aligned the fiber tips poorly by hand. We broke half the electrode connections within the base board when we applied pressure to the masts. Nevertheless, TS2 demonstrates that we can pack all necessary electronics in the footprint available, and control all actuators individually.

Figure: Perimeter Tracing by Three Correctly-Connected Fiber Positioners in TS2. Each perimeter marks the outline of a fibers dynamic range. For trace of spiral reset, see here. For a video of seven fibers moving, three tracing out their square perimeters, see here.

According to our fiber view camera, the range of motion of our TS2 fibers is 3.2 mm × 3.2 mm, which is significantly smaller than the theoretical maximum of ±3.8 mm. In TS2 we ended up soldering the actuators 6 mm above their bases, and we fastened the carbon fiber masts 5-10 mm from the tops. Our original assembly tooling was designed to reduced the amount by which they bend at their top end, but our original assembly plan turned out to be impractical, and we did not want to wait for another set of assembly fixtures before proceeding. The exact size of the dynamic range is not a source of great concern for us. Even ±3.2 mm covers 40% of the area occupied by the positioner. Our intended use of the DFPS is to permit each fiber to observe one object in each exposure, not to observe a particular object. According to our simulation, 40% coverage is perfectly adequate to permit a large-scale survey of the sky.

Figure: Fraction of Available Objects Observed Versus Fiber Range of Motion. We start with 2000 objects in a 50 mm × 100 mm focal plane. We move the telescope between each exposure and try to observe one object with every fiber. Obtained with our Observing.tcl simulation.

We are now constructing two self-contained fiber positioner front-ends. One is destined for installation on the McDonald Observatory's 2-m telescope in August 2024. The other will remain here at OSI for testing and trouble-shooting. This prototype, which we are calling the DFPS-4A, will be equipped with four positioners, each presenting two detector fibers and one guide fiber. The first DFPS-4A is on schedule for completion in mid-July 2024. Our collaborators at TAMU are assembling an eight-channel, fiber-fed spectrograph for use with the DFPS-4A. We plan to attach this spectrograph directly to the outside of the DFPS-4A deployed on the telesceope.

The DFPS-4A contains its own fiber view cameras, light injectors, high-voltage power supply, and data acquisition electronics. It communicates with the outside world through a Ethernet port. It is powered by a 24-V power jack. Light from the detector fibers is available on eight fiber-optic feedthroughs on the enclosure wall. The DFPS-4A can be expanded from four to eighty positioners without any significant modifications. If we are able to obtain reliable celestial spectra with the DFPS-4A and the TAMU, and demonstrate reliable positioning and tracking, we will seek funding for the construction of the eight-positioner version of the same instrument, the DFPS-80A.

Development Log: Development of the DFPS at OSI starting January 2022.

SBIR Phase I Application: "A Novel Dense Fiber Array for Astronomical Spectroscopy", application to National Science Foundation (NSF) Small Business Innovation Research (SBIR) agency by Open Source Instruments Inc. Submitted 04-DEC-20, awarded 01-JAN-22, award number 2111936.

Phase I Final Report: Final report on work done during Phase I, submitted to NSF in March 2023.

SBIR Phase II Application: "A Novel Dense Fiber Array for Astronomical Spectroscopy", application to National Science Foundation (NSF) Small Business Innovation Research (SBIR) agency by Open Source Instruments Inc. Submitted 24-JUN-23, proposal 2334185.

SBIR Phase II Reviews: Reviews of our Phase II proposal, 20-DEC-23, proposal 2334185.

Focal Radio Degradation in Multi-Modal Optical Fibers: White paper, 2023.

Direct Positioning of 50,000 Optical Fibers: Poster presented at Snowmass, 2022.

Base and Service Board (A3043): Combined base and service board for mounting fibers and controllers.

Auxiliary Circuits (A3044): Backplane, cell supports, and fiducial plate circuits.

Fiber Controller (A3045): Logic and amplifiers that generate control signals for actuators.

Fiber Positioner Circuits (A2089): Prototype circuits developed at Brandeis University for the DFPS.

Properties of Piezoelectric Tube Actuators: Study of the movement due to creep in piezo-electric tubes. Guadalupe Duran, Brandeis University, May 2020.

Direct Fiber Positioner System: A method to guide fifty thousand optical fibers. Kimika Arai, Brandeis University, May 2022.

Other Reports: Separate reports on aspects of the DFPS development.