Automated Star Pointing

By Grantland Hall:

Over the past two years, I have been working diligently on the control system for BICEP Array. That encompasses all of the software that is responsible for moving the telescope, capturing data from the receivers and other sources, and then pulling all of that data together and storing it onto hard drives. Usually that work doesn’t leave too much to look at. Even our user interface, the Generic Control Program (GCP), looks superficially unchanged, even though the underlying code base has seen over a million lines of code changes!

While our telescopes largely run themselves during winter observations, there remains one major manual task that the winterover retains manual control over – optical star pointing. This is when we use a small optical camera, pointed in the same direction as the telescope, to find stars in the night sky, and then use those known star positions to calibrate the absolute pointing of the telescope. This task is typically performed every two weeks and takes about three hours diligently clicking on images of stars. Sometimes, due to the buildings slowly sinking into the snow, the telescope mounts need to be re-leveled, causing the pointing to change, and so a full 7 hour shift of star clicking is required.

Of course, this is not an insurmountable challenge for a modern computer. So a few weeks back, after becoming frustrated with one of the electronic components needed to continue to run Keck receivers in the new BICEP Array mount, I decided to spend the afternoon trying to automate the process of star pointing. Out winterover engineer Nathan had already purchased some cameras to be used for optical star pointing, so I stuck one to the top of the telescope with some foam tape and a fisheye lens and set to work writing the software necessary to automatically locate stars.

IMG_20190801_143428IMG_20190802_181900

Nathan made an offhand suggestion about dangling a headlamp for use as an artificial star, so I wrapped a headlamp around the overhead bridge crane in our highbay to allow me to dangle it over the telescope and have the telescope track the headlamp around the room. It turned out that I was restricted to a small section of the highbay due to the sun beaming in the highbay windows and drowning out the light from the headlamp.

MVIMG_20190802_162743Finally, after about a day of figuring out how to produce usable images from the camera, how to translate those images into motions, and how to communicate those motions back to GCP, I ran our very first automated star pointing! Because the fisheye lens severely distorts the image near the edges of the frame, the movements weren’t very precise, which should improve when we switch to our rectilinear lens (a 500mm cassegrain lens, for any photographers). But I was quite happy with the results!

Advertisements

An Operational Receiver

We’ve been very busy in the last few months, and although we’ve taken a lot of photos there hasn’t been much time to write an update. At the end of May we received the BA2 cryostat from the California Institute of Technology. This cryostat was the second to come off the production line and had been shipped to Caltech after being outfitted here at Minnesota. At Caltech, the sub-Kelvin insert and cryogenics were integrated, these allow our detectors to reach their final operating temperature of ~270mK (just 0.27 degrees above absolute zero!) They tested out its performance, then sent it back to us so that we could integrate it into the mount while they continue to work on developing, fabricating, and testing the batch of detector tiles we will be deploying this coming winter.

 

The BA2 cryostat shipped in a disassembled state, so the first task was to rebuild it. For most of us locally, this was the first time we’d been able to interact directly with this part of the receiver so we jumped at the chance and took plenty of photos. For this post I’m going to focus more on the receiver and less on the mount.

 

First a quick breakdown:

The detectors we use for our experiment are superconductors. This requires that they be maintained at an extremely cold temperature, just a fraction of a degree above absolute zero (around -459 degrees F). However, for something that cold anything at room temperature shines like the sun and provides a lot of heat. Therefore we need to provide a lot of shielding, we call this structure the cryostat (shown below).

cryostat_aug19

The largest (by volume) portion of the cryostat consists of three nested shells, color-coded red, green and blue from warmest to coldest in the image above. The red shell is our vacuum jacket, it sits at about room temperature and is air-tight which allows us to put everything inside it under vacuum which means there is no gas bouncing around to transfer heat to the cold components. The green shell sits at a nominal temperature of about -370F and serves mostly to intercept the heat radiated inwards by the outermost (red) vacuum shell. The blue shell sits at a nominal temperature of about -452F and serves both to block heat radiated inwards by the green shell as well as a place to attach our optics. The green and blue shells are both cooled by a pulse tube cryocooler which sits inside the smaller canister on the right hand side of the image. Although the shells need to be rigidly supported inside each other, heat conduction across any support system must be very small if we want to keep our internal components cold. Our support system is made out of a number of fiberglass and carbon fiber rods at the bottom end and very thin titanium strips at the top end.

This pulse tube can only cool us down to about 4 Kelvin however, which isn’t cold enough for our detectors. Therefore, inside our “4 Kelvin” stage (blue) we have an additional “sub-Kelvin” stage (below left) which is cooled by a 3-stage Helium Sorption Fridge (below right)

 

The sub-Kelvin stage is similar to the larger nested shells, in that its tiered structure provides a way to shield the coldest elements of the receiver from as much heat as possible. To that end it has three primary tiers, each cooled by one of the fridge stages. The first at about 2 Kelvin (-456 F) the second at about 0.340 Kelvin (-459.05 F) and the coldest at about 0.250 Kelvin (-459.2F). It is to this last stage that we attach our detector modules.

 

Each part of the receiver has to be installed carefully and precisely, and there are massive numbers of cables that carry data between different elements. And of course, with such a complex piece of hardware, plenty of tests happen between each individual component installation to ensure that nothing was accidentally bumped or disconnected. The whole process can take 1-2 weeks to build up a receiver from all its composite parts. Below is a collection of images from that process:

 

Unpacking the sub-Kelvin stage from its shipping container and installing the fridge

 

 

Installing the heat straps between the fridge and the three sub-Kelvin stages. Checking the correct installation of our temperature sensors.

 

Installing the pulse tube cryocooler and the heat straps that go the the 50 Kelvin and 4 Kelvin shells

 

A frequent sight: running electrical tests between each subcomponent’s installation to ensure nothing has become accidentally disconnected.

IMG_20190525_110840

 

Installing a single detector module (covered with a copper plate) and shorting boards in the rest of the slots. We only had one module available for us to use here at Minnesota for testing in the mount, the green shorting boards are just place-holders in the electrical system. After installation, everything is covered with reflective material which serves as both a shield to electronic interference and to reflect radiated heat from warmer components.

 

Left: The Niobium (superconducting) magnetic shield that helps to keep our detectors from picking up stray magnetic signals. Right: The Vacuum jacket, 50 Kelvin and 4 Kelvin shells waiting to be installed. The two cold shells are coated in reflective material to help them reflect radiated heat. Bottom: looking downwards from the top of the cryostat after the three shells have been installed.

 

For this run, we didn’t have any optics available which is why the focal plane can actually be seen from the top of the tube. For a science run there would be two filters and two lenses blocking the view. For our run we ran “dark” which means we blocked off the line of sight from the detectors with two aluminum plates.

 

Finally, the cryostat operating on the ground and in the mount.

 

After installation in the mount, we spent a significant amount of time getting all of the data acquisition and control software working so that we can read out data reliably.

Encoders and Hoses and Cryo, oh my!

We’ve been quite busy the last two weeks. After successfully testing out the winch system we were almost ready to run a true receiver cryogenic test in an operating mount, but we still needed to put a few more things on. First, we attached the encoder tapes and their read-heads. The encoder tape is a long adhesive backed tape with precision spaced magnetic North and South poles. As the mount turns, the read-head can count the number of poles it passes. We can then feed this information into our control software in order to track the orientation of the mount to extremely high accuracy! Below is a picture of the Azimuth encoder’s read-head (left) and a close-up of the encoder tape as viewed through a magnetic field viewer.

We have a second encoder tape and read head up on our theta axis for accurate control of what we refer to as our “Deck” angle. The elevation axis also has an encoder to measure its location though it is self-contained and we did not have to install it. With the encoder information in the system we can now define our “home” point and reference all our rotations as a number of degrees away from that point, which is a critical step in being able to scan the telescope.

 

Before we could run a cryostat up in the mount, we also had to install the Helium lines. Each cryostat uses a Pulse Tube cryocooler in order to get the two largest internal stages down to temperatures of 50K (-369 DegreesF) and 4K  (-452 Degrees F). These pulse tubes use compressed Helium to extract heat from the internal structures and each one has to be hooked up to a compressor that is detached from the rotating parts of the mount. As a result, we have to run a set of Helium lines from each cryostat, through the Theta axis rotary joint, then through the elevation cable carrier, then down into the Azimuth rotary joint, then down to the floor and into the compressor.

 

We have 8 sets of lines lines that run the four cryostats, 2 sets of guard channels that help to control leaks within the rotary union, and one nitrogen set. With so many hoses running up into the structure, it’s imperative that we take care in routing them so that they aren’t in our way. To do that, we decided to run all the hoses underneath the various beams of the mount. When they approach one of the rotary joints, we attach them to a set of custom low-profile rigid sections that bridge the connection between the fixed rotary union and the flexible hoses.

These are a really cool set of fixtures, and each one is unique. We have only manufactured enough for one set of lines so far, but the rest are in production. This area is going to look very neat once the full set is installed.

With the encoder and Helium lines installed, we were ready to unload the dummy receiver and load in a real Bicep Array receiver with a pulse tube. Although we’d installed the dummy receiver using the new winch system, we hadn’t actually installed the real thing so we weren’t sure if there were going to be additional quirks in the procedure that we needed to figure out.

 

In the end, it installed without a hitch. Everything went smoothly, and we were able to secure it to the mount using the harness system in about 2 hours. We then attached a bunch of extra hardware to the front of it which is necessary for the operation of the pulse tube. With everything plugged in, we turned on the pulse tube and were rewarded with it’s characteristic chirping noise. With the cryostat installed and cooling we took the time to shoot some video.

The chirping noise you can here is the pulse tube. High and low pressure Helium gas are alternately pushed through the pulse tube cold head in order to produce the cooling power required to achieve cryogenic temperatures.

 

We also took a moment and got just about everyone to sit on the mount and shoot a quick video of a partial scan.

 

 

Finally, once we had the telescope cooling and scanning…

IMG_20190503_111441

 

We threw an open house for the physics department. Quite a number of people attended and the donuts were quickly claimed. For an hour we chatted with other members of the department and anyone else who happened to walk by the high bay. Each of us got to talk about different aspects of the experiment we worked on and talk about the overall project while the telescope scanned in the background. It was great to see so many people interested!

 

 

Receiver Loading System

For the last few months we’ve been taking advantage of the overhead gantry crane in order to load the receivers into the mount because it’s quick and convenient. But we won’t have access to such a crane at the South Pole (at least not regularly). This means we had to design a hoisting system that will let us load and unload receivers at-will. This presented a few challenges as the receiver is pretty heavy, we’re very constrained on space, and we are restricted to loading at one Azimuth orientation.

The solution we came up with was to use three separate hand winches connected to the cryostat with wire rope.  The rope for each winch runs up to one of the three pulley stations that surrounds the crysotat envelope, then back down to the ground where the receiver will be when ready to load. The pulleys are affixed to the theta axis which means they will rotate with the cryostats, so we’ll end up with four sets. The winches however, are fixed to elevation instead (and therefore fixed in Azimuth). This means that we can spin our theta axis without moving the winches away from the loading position, allowing us to load in a receiver and rotate it away opening up another slot for the next receiver.

The photos above show the winches attached to the mount and threaded through their respective pulleys. The two wooden platforms give us a flat surface to walk around on during the loading process but don’t interfere with any hardware. One thing you might notice from the photos above is that the winches don’t align with the beams that they sit on. This is intentional, in fact the winches and the pulleys are all twisted away from this alignment. The reason for this is because the wire rope needs to approach the drum of the winch in a very specific way, otherwise it doesn’t want to wind correctly which can cause problems. However, the pulleys also need the wire rope to approach them in a specific way, if the rope ends up out of the plane of the pulley wheel it can ride up on the side of the pulley and begin to abrade which weakens the rope. All the angles are purposely designed to achieve approach angles that are close to ideal.

 

For our first loading test of this system, we went back to using our “Dummy” receiver. This has the same weight as a real receiver but is slightly shorter. We passed the rope coming from each winch through an eyebolt in the top flange, then hooked it around an eyebolt in the bottom flange. We need to lift from such a low point because we have very little space up at the top end, and the ends of the ropes themselves create conflicts right at the end of the loading process if we have them higher up. Lifting from the bottom flange gets around this problem but presents another complication, we have to lift from below the center of mass. If it begins to tip too far it can become unstable and try to flip over. Passing the rope through the upper set of eyebolts fixes this and helps stabilize the receiver for the loading process.

 

When we finally attempted to load the receiver we came across one last complication. With three separate winches all being activated at different times, the receiver began to sway back and forth by quite a bit, enough that it would hit the mount at the points of tightest clearance. In order to fix this we ended up synchronizing the three winches using an electronic metronome. Interestingly with the winches all synchronized we observed very little sway in the receiver as it was being loaded.

The above two photos show the receiver once it has been loaded and is ready to be attached to the mount with the blue harness arms that can be seen next to it. The left photo above shows just how tight of a space we have up at the top which is the reason we pass the rope through the top set of eyebolts and connect them at the bottom.

 

Finally, with mention of the synchronized winching of course we have to provide a video. We haven’t quite figured out what the optimal speed it yet but experimenting with it should be interesting.

VID_20190415_171349 from Mike Crumrine on Vimeo.

3 – Axis Motion

With successful integration of the Theta motors, we quickly moved into installing the elevation motors. Installation itself was relatively simple, and since we’d already laid the groundwork with our control software we were able to get it up and running in only an hour or so.  Once running, we needed to check that the limit switches functioned as expected. Since the elevation axis isn’t capable of 360 degree rotation, we install limit switches on both sides. These switches act as safeties that prevent us from driving the axis too far in either direction. In the photo below, the blue box with “H” and “S” written on it holds the “Hard” and “Soft” limit switches.

IMG_20190214_163609.jpg

As the mount rotates counter-clockwise in elevation, the screws just to the left of the limit switches get closer and closer to the flippers on the underside of the labelled box. When the “Soft” switch is tripped by these screws, the controller prevents further commanded motion in that direction and the mount slows to a stop. If the mount continues to travel too far past the soft limit, it will trip the “Hard” limit switch. This switch trips all power to the drive amplifiers, causing the motors to break hard and stop the mount essentially in place. A second set of these limit switches is on the other side of the elevation gear and prevents too much motion in the opposite direction.

 

Once we’d confirmed the operation of the limit switches, we went ahead and drove all three axes of the mount simultaneously for the first time.!

 

We next need to fine-tune the control system to give us the precise motion we want on all axes. Now that all the motors are on we can also attach the Azimuth and Theta encoders which enables high precision tracking of the exact position of the mount on each of these axes. We’ll do this using encoder tape and magnetic read heads for these two axes. The elevation axis uses a more compact encoder that is already attached right next to the bearing.