The BBC News article Telescope tracks 35 million galaxies in Dark Energy hunt says:

The aim of the five-year programme is to shed light on Dark Energy - the mysterious force thought to drive an accelerated expansion of the Universe.

The instrument effectively contains 5,000 mini-telescopes. Each one can image a galaxy every 20 minutes.

In just one year scientists will have surveyed more galaxies than all the other telescopes in the world combined.

I wonder if instead that it would be better to call them 5,000 mini-spectrometers or 5,000 mini-slits, but I'm not yet sure how this whole thing works.

Inside DESI are 5,000 optical fibres, each acting as a mini-telescope. This enables the instrument to capture light from 5,000 different galaxies simultaneously, precisely to map their distance from Earth, and gauge how much the Universe expanded as this light travelled to Earth.


  1. How exactly will DESI simultaneously capture individual spectra from 5,000 galaxies using optical fibers? Are the fibers used to route the light to a very, very long single slit which is carefully imaged on a 2D detector, or does it do something more complicated?
  2. Considering that the exposure cadence is 20 minutes (per the article), how are all 5,000 fibers quickly repositioned between exposures? 20 minutes is only 1,200 seconds!

update: Wikipedia's Dark Energy Spectroscopic Instrument says:

The Dark Energy Spectroscopic Instrument (DESI) is a new instrument for conducting a spectrographic survey of distant galaxies. Its main components are a focal plane containing 5000 fiber-positioning robots, and a bank of spectrographs which are fed by the fibers.

Wikipedia does not always serve as an authoritative and accurate source, especially for details. Five thousand robots sounds like the name of a Kraftwerk album, and "bank of spectrographs" is less than fully quantitative.

enter image description here

DESI will scan more galaxies in a single year than all the telescopes in the world combined, UCLA/ Lawrence Berkeley National Laboratory

This answer to What is the dimensions of the smallest object detectable by an optical fiber from a specific distance? shows several implementations of multiple optical fibers picking objects from a focal plane and bringing them to a spectrometer. However I don't think any of them come anywhere near to 5,000 separate fibers!

enter image description here

above: GIF from the video (with exciting music) A 2dF night at the Anglo-Australian Telescope

  • 1
    $\begingroup$ @PeterErwin thanks for simultaneously removing the simultaneous duplications of the word simultaneous in two places simultaneously!! $\endgroup$
    – uhoh
    Oct 30, 2019 at 8:30
  • $\begingroup$ Ah yes, Dark Energy, the Mysterious Force. $\endgroup$ Oct 30, 2019 at 19:41

2 Answers 2


You can probably get most if not all of your questions answered by perusing the main DESI web site, which I encourage you to check out.

There is, for example, a nice video describing the assembly of the main focal plane elements (the fibers and the associated robot positioners) here.

But in simple terms: the circular focal plane is divided into ten wedges (or "petals"). Each wedge holds an array of 500 fibers and their associated robotic positioners. Each set of 500 fibers is collected at the back end and passed outside the telescope to a spectrograph, where the fibers are arrayed into a line to pass their light onto a slit. That is, there are ten separate spectrographs (that's your "bank of spectrographs"), each of which has a single slit with the light from 500 fibers passing into it. (Each spectrograph then has dichroics splitting the light from the slit into three different wavelength regimes and three different cameras, each with a different grating -- 360-555 nm ["blue"], 555-656 nm ["red"], and 656-908 nm ["infrared", though this is still handled by a CCD and it not really what modern astronomers usually call "infrared"]).

The fibers are pretty densely packed into the individual wedges, and so the robotic positioning is done by tweaking/wiggling each fiber a small amount around its default position within a circular hole (you can see glimpses of this in the video), rather than the wholesale movement of individual fibers across the whole focal plane that the 2dF instrument in your video link does.

enter image description here One of the ten focal-plane wedges of DESI, with about 60 (of its eventual 500) fiber-plus-robot-positioner elements inserted.

  • $\begingroup$ Okay I've got it. It's probably likely that for any given exposure not all 5,000 fibers will have a suitable galaxy within their individual ranges of motion, but that's par for the course with massively parallel architectures. I'll do some further reading now. Thanks! $\endgroup$
    – uhoh
    Oct 30, 2019 at 8:41

Supplemental to @PeterErwin's answer, some more details on the five thousand "robots".

Each fiber has a circular "patrol area" with a diameter of 12 millimeters, and these are located on a hexagonal array with a pitch (nearest neighbor distance) of 10.3 millimeters.

Motion is implemented with eccentric axis (Θ–Φ) kinematics. Instead of x-y or r-Θ which use two and one degree of linear translation, two rotations are used, in a "shoulder and elbow" configuration.

That the fibers' range of motion overlap improves the probability of finding a target for each fiber, and reduces the dead space (area with no fiber coverage possible) to zero. The robotics systems are aware of the positions of adjacent fibers so collisions are avoided.

See also Dark Energy Spectroscopic Instrument (DESI) Fiber Positioner Production Leitner et al. 2018 in Researchgate and arXiv

Here's a crop of a part of Figure 1 of the arXiv paper linked above, for a detailed view of the robot's (Θ–Φ) kinematics.

enter image description here

From SPIE's Robotic fiber positioners for dark energy instrument

enter image description here

Figure 1. Left: Eccentric axis (Θ–Φ) kinematics. Whenever the Φ arm (R2) is retracted within the dashed circle E, the positioner is guaranteed free rotation about Θ without obstruction from its neighbors. Right: The patrol region (the area within which every point can be reached by the robot) coverage extends beyond the pitch. This assures complete coverage but requires implementation of collision avoidance algorithms in the move control software. R1, R2: Kinematic lengths of the Θ and Φ arms. xc, yc: Center of robot. x, y: Position of fiber. Θo, Φo: Home positions. 1×, 2×: Coverage by a single positioner and two positioners, respectively.

enter image description here

Figure 2. DESI (Dark Energy Spectroscopic Instrument) 10.4mm pitch Θ−Φ fiber positioner. The centers of the two axes are indicated on the insert (top right).


You must log in to answer this question.

Not the answer you're looking for? Browse other questions tagged .