PFS Upgrade Series, Day 4: Rollercoaster!

This is part of a series of posts about upgrading an instrument at Las Campanas Observatory. If you want to start at the beginning, it's here.

As you may recall, on the first day of PFS upgrades, we had to move the instrument from a building at the top of a hill down the hill to another building, where we are working now. This was a literal down-hill move, and we'll have to go back up again in a week or so to (hopefully) use the instrument on the telescope. But today, we started at the top of a figurative hill, fell off and mulled around for a few hours, and slowly climbed our way back up. 

Caveat: This is my impression of the day. I'm not as experienced in this work as Jeff and Steve, so I'm certainly simplifying things, but hopefully not getting things really wrong. 

We started the morning thinking that we had defined the optical axis of the instrument, and were thus ready to start measuring the alignment of the collimator/camera optics of the instrument, pictured below. Jeff and I bolted this piece to the optical bench inside the instrument this morning, using the hoist to help lift and position it. Steve arrived and we thought we were all ready to go, but we soon discovered that (1) the pupil image was offset from center on the wavefront sensor when using the objective lenses that came with the sensor [and our optical set-up], and, relatedly, (2) when we changed the focus of the set-up by moving the wavefront sensor, it introduced significant aberrations (astigmatism and coma). The expectation was that this shouldn't happen.

Above and below, our set-up this morning. The paper on the flat mirror (seen below) is to restrict the area we are analyzing to match that of the real instrument on the telescope. I think this is before we put the objective lenses on the wavefront sensor; I can sort of make out our metal pinhole attached to the wavefront sensor on the left.

Observing something one doesn't expect is sometimes exciting, but in this case was very puzzling, such that we had to ponder it over lunch. After we had fun talking about a possible new speckle camera for Magellan, we decided to put a metal pinhole in front of the wavefront sensor and try to get the light to come out of and go back into the pinhole, after going through the lenses, off the mirror, and back through the lenses again. We planned to do this by adjusting the wavefront sensor and *not* the flat mirror on the other end of the optical bench; we wanted to preserve that well-defined optical axis. For the same reason, we were reluctant to tilt the wavefront sensor stage -- moving it in x, y, and z was easy to "undo", because the knobs have gauges, but the two tilt knobs do not, and are very sensitive. 

We tried this set-up, with no tilting, but were still having the same focus problem, and (I think?) the pupil image was still not centered on the wavefront sensor. At this point I stepped out to get some sun and then call my mom. When I came back into the clean room, the plan had changed. Steve and Jeff thought they had worked out how to interpret the alignment procedure in the wavefront sensor manual, and were executing it. We thought we had done it correctly before, but really had not. Go figure! The true alignment procedure consisted of installing the pin hole provided with the wavefront sensor in front of the sensor, and moving the wavefront sensor stage in the (x,y) directions until we could align the point of light coming back to the sensor with its center. It was kind of like trying to play skee-ball and get the ball right in the center of a hole to get the most points. 

Then we took out the pin-hole and used the tilt screws on the wavefront sensor stage -- yes, we changed the tilts! -- to fully center the pupil image. We iterated this procedure a few times, also trying to minimize the tilt and curvature values (by adjusting in the z-direction of the wavefront sensor mount). In the end, we got a very nicely centered pupil image, a small focus offset, and small optical aberrations. 

Centered image (large gray/black graphic and pink intensity map), yay! Also, it maybe hard to see, but the middle lower window lists the Zernike coefficients, which correspond to different deviations in the wavefront (light path).

We think this is the correct alignment procedure to use, such that any aberrations we measure are now due to the instrument optics and not the wavefront sensor itself. But -- second unexpected thing of the day -- the aberrations were very small, tenths of a micron! This means that the original optical alignment of PFS that Steve and Jeff did almost ten years ago, without the fancy-shmancy wavefront sensor, was very good, good enough to not change any of the lenses now to try and improve it. It took about an hour for Steve and Jeff to convince themselves that they really did do such a good job aligning the instrument originally, given that we do see some aberrations in the images PFS produces when in use. That is, something must be causing those aberrations, but it appears not to be the lenses. 

I was honestly very excited (and impressed!) by this news, but it was also kind of a "whomp-whomp" moment, since this source of aberrations would have been the easiest to fix. Tomorrow we will examine one of the other potential sources, inhomogeneities in the prism, but this isn't really fixable; we can't change anything about the prism other than getting a new one. If the prism is obviously the cause of the aberrations, we could try tweaking slightly the lenses to correct for the prism, but that's dicey. A third source of aberrations could be in the exact position of the dewar (holds the CCD detector), but to test that we have to get the new CCD ready to mount. 

So, all in all, a roller-coaster day!