Working on CIRCE at UF

CIRCE is the Canarias InfraRed Camera Experiment, led by Professor Stephen Eikenberry at the University of Florida. I am one of the graduate students who has been involved in this project over the course of about 15 years and I’d like to write about a bit of what I’ve done here on my blog.

CIRCE is a 750 kg (1650 lb) instrument built mostly out of aluminum. It’s about 1.75 meters (5 feet) long and over 1 meter (3 feet) in diameter. I think of it as a her mostly because the name is derived from the sorceress Circe from the Odyssey. On the outside she’s powder-coated black.


CIRCE being lifted to the Folded Cassegrain focal station at GTC.  The silver lines are for liquid nitrogen (LN2) transfers to her LN2 tanks.

I started working on CIRCE’s cryogenic mechanisms in 2012, testing its linear slide (which holds a focal plane mask and half wave plate rotator) and filter box (with 5 wheels). Both mechanisms are designed to work at both room temperature and pressure and in a vacuum at 77 Kelvin (-321 F), which is the operating temperature of the instrument.

This is a tricky thing for a few reasons:

  1.  Lubricant (for shafts and bearings) turns from nearly friction-free liquid to solid rock. We have to degrease EVERYTHING that is inside the vacuum jacket (the black cylinder) because of this. Additionally, we don’t want the lubricant to off-gas while we pump out all the air; any volatiles that boil off when we remove the air from the inside will settle on the optics and detector, which is extremely difficult to remove and therefore bad.
  2. Gear teeth must mesh at both room and operating temperatures. That’s tricky to design, especially if you have to use mixed metal gears (such as stainless steel and brass). CIRCE got around that by using custom-cut aluminum gears for the filter box, but it’s still not easy.
  3. Different materials shrink at different rates. Our shafts are precision-ground stainless steel, but the linear slider bearings are most definitely not stainless steel. As a result, some bearings have to be cut with a slot along their long axis to allow them to grip the shaft without binding.

What this means is we purposely degrease the bearings and shafts as much we can using acetone and isopropyl alcohol. There’s a fair amount of clever design work involved as well — the gearing required by our projects must be carefully chosen so that the gears can be cut and deliver the required step resolution. For example, CIRCE’s filter wheels take 1200 full steps per full revolution and have 4 filter positions and a wedge-shaped ‘Open’ position. We end up 1/8 stepping the motors for better control, which means that the filter positions are about 250 steps from each other. The gearing on the outer rim of the filter wheels is helical; the drive gears are split in half radially and connected by a spring to give minimal backlash so that no matter the direction they are turned they provide nearly instantaneous response to commanded motions.

We don’t use absolute encoders for our mechanisms because 1) they’re expensive to design for cryo conditions, 2) a source of radio frequency interference, and 3) we don’t need something that complex. We find that counting the number of pulses, or steps, the motor is commanded to move sufficient nearly all of the time.

As a result, however, we need to be able to ‘home’, or zero the step counts for all of the mechanisms so that we know where the mechanism is in its motion, especially if the moving part can run into and damage other components.

Testing the filter box and linear slide for repeatability and reliability meant putting it in our small test Dewar, wiring it up, verifying that it worked warm, then pumping out the air and cooling it with LN2 until it reached equilibrium some days after starting the test process. This allows time for the filter wheels to cool via conduction through the bearings, shaft and shaft bearings to the aluminum body. It’s not very efficient, so some of the cooling happens radiatively as well.


The wheels in the box go round and round….

We then tested the filter box by commanding something like 250 motions per wheel for 4 gravity vectors (over 5000 motions). We also tested how reliable homing was by triggering the homing switch and counting steps. It was not the most exciting two weeks of my life.

After that the CIRCE team began constructing the active and passive thermal shields. They are made out of aluminum sheets and are designed to minimize the heat load put onto the optical bench (as well as create a nice dark environment inside the instrument. We iterated a lot, working on welds, seams, joints, and aluminized Mylar jacketing to further improve their thermal characteristics. I spent about a month doing little more than trimming the Mylar and aluminized tape to shape the oddly-shaped components. I also helped out in the machine shop, cutting the sheet metal to shape the shields. It was a fantastic learning experience.


This was temporary thermal shielding to test how long CIRCE would stay cold without refilling the LN2 tanks once it was cold.

CIRCE’s detector electronics arrived in 2013 and we began tuning the detector and readout electronics to optimize the performance. I was only involved tangentially, but I did run the detector on occasion. I also did some design work for some clamps and wire guides both in and out of the Dewar.

The neat thing with the readout electronics is the flex cable system. These cables, about 15 inches long, carry the a data from the detector to the outside world (there are lots of details there, but it’s ancillary). These flex cables have Hirose connectors which are finicky beasts and tend to disconnect if the flex cables are under tension, which (spoiler alert) they always are. Our solution was to print a clamp in three parts, with the joined cables held precisely in place by the clamps (much like making a mold). This brings the chance that the cables will spring apart to 0. However, connecting the flex cables is a bit of a bear; it requires one person to hold the interface plate tight against the vacuum jacket while a second person reaches through the interface hole and connects the flex cables and screws the last clamp piece into place. It’s not my favorite job because any static discharge near the flex cable (which protrudes from the back of the interface plate by about 4 inches) will fry the detector or delicate readout electronics, and if the first person drops or lets the interface plate slip, the flex cables will be damaged beyond repair.


The black batting cage and clamps. The batting cage keeps the thermal break cables from thwacking into the flex cables while offering attachment points. The flex cable connector is the white-ish bit in the middle of the amber bit. Note the final shielding is more robust than the previous image.

Next was lab acceptance testing, where we ran CIRCE through her paces and demonstrated to the GTC that CIRCE would in fact work at GTC. That went off with only small hitches, such as GTC’s shock at the (admittedly) poor quality of the engineering-grade detector. In CIRCE’s defence, IR detectors are more flawed than optical detectors, and despite 2 of 32 amplifier channels being dead, the CIRCE detector’s flaws are averaged over when dithering. Once we proved to them via simulation that the detector’s flaws were acceptable, they accepted CIRCE as a GTC instrument. That meant we had to build a crate.


CIRCE being installed on its handling cart, which is bolted to a floating platform inside the shipping crate. Featured is Steve’s back and the mustachioed Nick Raines.

Over the course of about a month we built a shipping crate about 10 feet long by 5 feet high and 6 feet wide. It was quite the undertaking, but eventually we had CIRCE safely ensconced and shipped. As our mechanical engineer said, it was a good sight, seeing the truck’s taillights as it pulled out of the loading dock.

While that was going on, we began fabricating the cabling for use at GTC. The cables run from the electronics rack (which is hard-mounted to the telescope elevation ring) to the Folded Cassegrain rotator cable wrap (designed to allow cables to go from a fixed panel to a panel that rotates with the instrument) to the interface plate on the backside of CIRCE. We had constructed lab-testing cables that were about 2 meters (6 feet) long, but we needed about 10 meters (32 feet) of cable with two connectors in between (at the fixed and rotating panels). Building them was an exercise in patience, but I did improve my soldering skills considerably. I also helped build the electronics rack crate, which we had heat-treated in Jacksonville; we then packed and shipped it to the Canaries.

Once we got there….more to come!



Leave a Reply

Fill in your details below or click an icon to log in: Logo

You are commenting using your account. Log Out /  Change )

Google photo

You are commenting using your Google account. Log Out /  Change )

Twitter picture

You are commenting using your Twitter account. Log Out /  Change )

Facebook photo

You are commenting using your Facebook account. Log Out /  Change )

Connecting to %s