We passed a big milestone today with the ASM working in closed loop with 400 modes at 1 kHz (the most complex AO mode)! This 400 mode interaction matrix has been made possible by the excellent trouble shooting from our friends at Arcetri Observatory, Simone, Enrico, Alfio, Armando and Marco!
It was such an exciting event that Alan Uomoto made a movie:
How did this happen? Well, yesterday the AO loop was struggling to close on the bumps we were referring to as Viscachas:
The Viscacha is a bump causing a dark spot in the Pyramid pupils at around 2:30 o'clock.
When the loop tried to close on this, we would get a higher and higher unstable patch of actuators trying to correct it:
Here is the ASM display after closing the loop yesterday on 200 modes at 600 Hz. You can see the commands on the bad patch at ~2:30 o'clock; the mirror is working hard to correct something there, but we didn't know what.
So Simone and Enrico figured out that we were actually getting cross-talk from the Pyramid, because the phase bump was so high. This is similar to a quad-cell Shack-Hartmann without a guard band, where a subpupil may wander into an adjacent subaperture. Here is Simone’s drawing where he works out the solution:
Simone's drawing to solve the problem of the viscacha
So the solution is kinda a hack, whereby we applied a negative sign to the interaction matrix for that patch — and the bump and the viscacha disappeared!
Evolution of the Viscacha -- Final rendition, all corrected!
And so tonight we were able to close the loop with our new interaction matrix, and get a nice flat wavefront!
Result: Closed loop, flat wavefront, no viscacha!
New arrivals today: T.J. Rodigas (Steward) and Runa Briguglio (Arcetri).
Sorry for the tardiness of this post. The internet was down on the mountain when the night shift went to bed this morning. The day was a struggle, but we finally got at least something working, and then did some testing through the night.
The issue was a feature in our pupil which Simone has dubbed the Viscacha.
Our Pyramid pupils from this afternoon, with a Viscacha head clearly visble.
We tried several things. One quick and dirty modification was to remove the field stop for the Pyramid. Simone and Enrico climbed into the NAS to do this – after first convincing Laird that this was a good idea.
Simone and Enrico perform minor surgery on the NAS.
The day shift ended with a somewhat working 200 mode interaction matrix. So the night shift did some closed loop testing.
Katie at the AO operator's workstation.
Things are hopeful. With only 200 modes we have a fairly high Strehl ratio PSF. The problem is keeping it stable. These two videos, taken with the VisAO CCD 47 at 32 fps, show this.
Youtube noted that my image was shaky, and asked if I wanted to fix it. F*&!$^ you Google. Here’s the same data set, but keeping only the best 5% of the images (Lucky imaging style).
After breakfast Laird reduced this data by “shifting and adding”. Here’s what our images will look like soon:
Our SAA z' PSF with a Vischacha contaminated 200 mode IMAT.
To try to control the vibrations that even Google noticed we put the Clio rack on some foam pads. This improved the power spectrum, at least removing the ~2 Hz spike.
The CCD 47 spot x position power spectrum after putting Clio2 electronics on some foam to isolate the pump.
Quotes:
Simone: “Laird, take a seat. You want to take your seat. I think we need to remove the Pyramid . . . field stop.”
Tyson: “Why don’t you just say you’re a carnivore plus vegetables and cheese.”
Laird: “I thought you were trying to take the frisbee away.” (To Phil, after realizing that Phil was playing frisbee with grad students and the post-doc.)
Phil: “I was just trying to help them meet their 100 hours for the week.”
Simone, on the viscahcha in the PWFS pupil images:
“The viscacha again!”
“There is a clear viscacha.”
“It is not a dynamic viscacha, it is a static viscacha.”
Laird: “It looks like a viscacha that’s been run over by a car.”
Simone: “Yes, but it’s still a viscacha!”
Laird: “I’d really like a headless, tail-less viscacha.”
"Wild" viscachas from the backside of the telescopes. Wild as opposed to the "tame" viscachas who live in the ASB.A group shot of the first wave of the MagAO team. Armando left us this morning, so yesterday we gathered the team for a group shot. We'll repeat these as people come and go. From the left, front row: Marco Xompero, Alan Uomoto, Laird Close, Katie Morzinski. Back row: Armando Riccardi, Enrico Pinna, Alfio Puglisi, Simone Esposito, Jared Males, Tyson Hare, Phil Hinz. (Not pictured: Derek Kopon)Ya-Lin and Kate joined us today. Time to get to work.Phil, Laird, Ya-Lin, and Kate wait for the green flash.
Usually an AO system has the opposite problem: There are aberrations you can correct but not measure. And of course, there are all sorts of aberrations you can neither measure nor correct, like the very highest spatial frequencies. (And if you don’t have enough stroke, you can saturate your attempt to correct an aberration — but that wasn’t the case today.)
Contemplating and debating the riddle
But today while we were making our interaction matrices, we found we had 3 high spatial frequency dots in our pupil images (a couple times the size of an actuator) and we spent some time trying to track down whether they were phase or amplitude, and whether it was a dirty optic, scattered light, or misalignment.
Pato takes Enrico up in the scissor lift to inspect the optics
Jared had to power-cycle the CRO controller, and so he and Armando went up on the scissor lift to check if the CRO was in the right position and to inspect the optics.
Jared and Armando on the scissor lift, checking if the CRO motors are working properly
When we closed the loop, these “dots” could not be corrected. We tried shifting some optics to see if we could move the spots, and we concluded they weren’t an amplitude error (too bright). Finally, we tried setting the ASM back to an older flat shape — and the dots disappeared! This means that these phase errors were somehow introduced to the system at the time we were calibrating the interaction matrices. So then when we closed the loop, we were driving toward that shape — not toward a true flat.
So the answer to the riddle is, the shape you can measure but not correct is an error in your null!
Now that we’ve solved the riddle, tomorrow we’ll re-do our interaction matrices with a flatter null.
Vizzy contemplates the calibration riddle from her perch
Yesterday:
Simone: “Hey, I think we should try taking the pyramid out.”
Laird: “No.”
Armando: “No.”
Phil: “No.”
Jared: “No.”
Katie: “What? Is he serious???”
Today:
Laird: “Hey, maybe we should try taking the pyramid out.”
Simone: “No. The last thing we want to do is take the pyramid out!”
Our CRO is a very fast f/1 optic, and our ASM makes an f/16 beam. So motions of the CRO are amplified by a factor of 16 in our focal planes. We can see this on the VisAO CCD47 as a dancing image due to small mechanical vibrations of the telescope.
The MagAO project is adding a little bit to the vibrations of the telescope, mainly with two cooling pumps. One pump circulates glycol for our CCDs and the Shutter, and the other keeps Clio2 cold. With the 16x CRO amplification, we really see the impact of these pumps. We recorded 60 sec time series of the CCD47 operating at 32Hz in a 64×64 subarray mode. Here are the results:
With both pumps off:
Unnormalized power spectrum of the x position of the VisAO image position. Note that I'm being lazy and not fully normalizing the PSD. With both pumps off, we had an rms jitter of 2.2 pixels.
With the Clio2 pump off, and the CCD pump on:
Here we see more white noise, but no strong resonances and the image is fairly stable by eye. Rms jitter was 5.4 pixels.
With the Clio2 pump on, and the CCD pump off:
The Clio2 pump excites a ~2.3 Hz resonance. We think this is due to the rack holding the pump sitting on a beam connected directly to the telescope. Rms jitter in this configuration was 9.5 pixels
We are actually very happy with the vibration performance of the system attached to Clay. Taking into account the factor of 16 for the CRO reflection, we expect to have only a few milli-arcsec of jitter when we observe actual stars. We have also taken measurements with our internal artificial star without the CRO and confirmed this. Good news, especially for VisAO.
Another experiment we conducted was having the PI bounce around the control room. He has a noticeable effect.
Time series of the VisAO centroid x position, with some input from the PI.
Today we are going to explore the MagAO pupils and their corresponding transforms in the image plane, courtesy of Fourier optics.
So let’s have a look at the pupil. Here is a photo of the ASM, taken with a digital camera. This was from before Clio was mounted, so that we just stood on the Nasmyth platform and put the digital camera where Clio is now. The light source is the sky, and the light path is primary + secondary + tertiary.
ASM image, from before Clio was mounted. You can see bird poop on the tertiary, a splotch at about 11:00 in this image, and the lollipop-shaped "slot" is at 9:00.
The main features of the pupil are the outer diameter of the mirror, the inner diameter of the secondary obscuration, the support spiders holding up the secondary, and the slot. Here, then, is the pupil mask:
Pupil mask
Since we know what the pupil looks like, we can create simulated images of the focal plane by taking the Fourier transform modulus squared:
What the PSF will look like in the image plane (log scale)
If we really stretch the color table, you can see the diffraction off the spiders, but it is not a big effect. Also, I couldn’t find the diffraction off the slot, so it is negligible:
Stretching the color table to saturate the inner part of the PSF allows you to see the diffraction spikes off the spiders, but they are very faint
Now, Clio is an infrared camera, going out to 5 um, and so it has its own pupil mask, a cold stop. So let’s look at the pupil through Clio, by taking a pupil image (which we did after Clio was mounted). Here is an image of the pupil plane through the whole system, taken with Clio by putting in a powered lens to the focal plane to make a pupil image:
Clio pupil image, 3.4 um
It’s pretty cool because you can see the 2 spiders holding up the secondary obscuration on the cold stop, but you can also see the 4 telescope spiders and the ASM slot! Here’s just the Clio cold stop pupil mask:
Pupil mask - Clio cold stop
And here is its Fourier-transform-modulus-squared: The simulated PSF:
Simulated PSF for Clio cold stop (log scale). Diffraction off the spiders is a little bit visible here, since they are slightly wider than the telescope spiders.
And here is the zoomed-out, saturated version so that you can better see the diffraction spikes:
Clio cold stop PSF -- scaled to bring out the diffraction spikes.
