Day 25

I spent today writing code to analyze our data from the past week or so. Originally we tried mapping Gaussians to our data, but because we were on a logarithmic scale, most of our data points were negative, because our spot is very small and dim. Unfortunately, the Gaussian fit function in MATLAB does not deal well with negative numbers, so it mapped a Gaussian function to only the very few pixels around the center of our PSF and ignored the rest of it. Fortunately, there are many ways to acquire the data that we need, so we were able to calculate the encircled energy of each of our datasets. Encircled energy is, essentially, how much of the total energy (as a percentage) is contained within any given radius from the center of the function. We can calculate the approximate radius of the spot by finding the radius within which 90-95% of the data is contained.

Without a window in front of the sensor, the function looks like this, with the radius being about 15 counts wide, about 1.5 pixels. The function looked quite similar with a coverslip and a microscope slide, but looked a fair amount wider when it came to the thickest window.

Originally we saw a dip in the plot, which was odd, because we should not have the energy decreasing as more of the function is included; that could only happen if we had negative pixels.

Upon closer investigation, it appeared we did have negative values in an oddly uniform pattern around the spot. These negative values occurred because we magnified our data, multiplying our x and y axes by 10 each, leaving the algorithm to interpolate data for the new points. It turns out the default setting takes values from the nearest 4x4 group of pixels, which can lead the graph to believe there is a larger downward trend than actually exists in the data, and can give us values outside our original data range (in this case, meaning negative values). I was able to fix this by altering the settings to take only data from a 2x2 pixel neighbourhood for interpolation purposes.

Day 24

This morning we walked around the RIT campus, trying to find a microscope slide to use for our medium-width piece of glass. We first tried to find one in the building where the construct is, but then Dmitry remembered that the guy we were looking for had already graduated, so he was gone. We then tried to borrow one for the engineering building, but they refused to let us, on account of us "being from another college." Luckily we were able to find one in Gosnell Hall, so we brought it back to the lab. I was able to run our program after homing and focusing our setup, and we were pleasantly surprised to find that the resulting surface function was quite similar to that of the function from the thinner glass. It was so similar, in fact, that Dmitry had to take a couple minutes to figure out which was which when I asked him if he could tell.

Unfortunately, our fancy new camera will not be coming for another week, so I won't actually get to use it since presentations are due a week from today. While that is a little disappointing, it's alright, if we make good progress tomorrow, we should be able to take a few images with the Kepler camera, which I'm excited to use.

Tomorrow we'll be fitting curves (most likely Gaussians) to the data I took today. Besides the data from the images with the microscope slide, I also took two more sets of data with no glass cover and with the thin glass (a coverslip) as a second trial where I was much more careful to have our spot of light be in as exact focus as I could manage. Tomorrow we'll compare the two sets of data, and determine the effect of the window thickness and the differences in accuracy of focus.

Day 23

We managed to fix our light source today; I was able to find a new bulb with the help of Matt, and not only does it work, it's quite a bit brighter than the first one was before it blew out. I then took a new series of images with the thinnest piece of glass we could find, but unfortunately the newly increased brightness of the bulb needed to be accounted for in my code, so the plots and images I received on the other end were entirely dominated by noise. Eventually, after some jiggering with the code, I modified the SNR range, and managed to get rid of most of the noise. Our new camera should be coming tomorrow, so our next step will be performing the same tests with the fancier camera and comparing the results to our current ones.

Day 22

The Undergraduate Research Symposium was today, and there were so many interesting talks available that we couldn't go to all of them, but we did get to go to a couple. We heard a talk from a student who has been working with Professor Ninkov; he's developing and testing a quantum dot film, which would cover a CCD. The film takes in the light from a source (i.e. the sun) and fluoresces lower energy light into the camera, this means that sources with shorter wavelengths can be observed using CCDs with quantum dot films across them, as the decrease in energy results in longer wavelengths, which fall in the band that CCDs detect most efficiently. We also got to go see Lee's talk, which was about polarization in astronomy and the array he and Dmitry have been testing, which they'll be bringing up to Oregon to observe the solar eclipse on the 21st.

Unfortunately, the bulb in our light source blew out today, and we had to find a replacement, which Matt was able to locate for us. The new bulb worked at first, for a few minutes, while I was homing and calibrating the stages, but I didn't realize it had gone out when I was trying to find and focus the beam of light with the camera, so it took a while before I figured out that there wasn't any light to focus (oops). On Monday we're going to have to switch back to the original light source we were using before Peter and Ashley borrowed it for their setup.

Day 21

Today I finished writing the code to automate the image acquisition and creation of point spread functions (PSFs). The code allows us to take 260 images, 20 at each of 13 exposure times, and average each set of 20 exposures to get rid of noise in each frame. Then, taking this new set of averaged data, we can remove the saturated pixels and get rid of interference from the dark current, and after normalizing the data by dividing each point of information by its correlating exposure time (in seconds), we can map a surface function to the normalized data. We will be doing this several times with different thicknesses of glass between the beam and the sensor, which will allow us to determine the effect of glass (which keeps our Kepler camera in a vacuum chamber) when it interferes with the beam path.

We visited a couple of the other labs on the basement level this morning, and it was really interesting to see what they've been doing. Peter, Ashley, and I wrote a little card for Lee, an REU student who's been really great and who we've gotten to know pretty well, thanking him for everything he's done to help us (he's given me some really helpful advice for programming and debugging), as he's heading back to California tomorrow after the undergraduate research symposium. We're definitely going to miss him.

Day 20

Today I had the opportunity to hear three talks on various topics within the field of remote sensing. We got to hear Yue present on stray/scattered light with TIRS and TIRS GOES. He described the effects of the errors within an image and a couple ways to mitigate them, or reduce them at least. He's over in California, and will be presenting the same information at a meeting there. We went up to support Chris, an REU student who we've gotten to know during our lunches in the reading room. His presentation was excellent, he explained everything clearly, and I came away from it understanding his research pretty well. He was taking remote, hyperspectral images of large areas of land using LANDSAT to determine soil moisture content, which has potential applications in drought monitoring and agricultural management.

After the first two talks, we went to the final Tech Talk of the summer, where Rolando Raqueño talked about DIRSIG and its incredible ability to model virtual reality environments with imaging. He also mentioned my Dad, and his lab (GRIT), as they have collected a fair amount of data which has gone into DIRSIG, which was really cool to hear about.

I spent the rest of the day debugging my code in MATLAB, and at the very end of the day, right before I left (and after much messing about with timeout values). I managed to finish all our debugging and we'll have a chance to run the code tomorrow with some proper images instead of our test ones.

Day 19

Today I wrote and finished the code which will allow us to automate photo acquisition. This code will be incredibly useful, as it allows us to take many frames for each exposure time and to average them, eliminating background noise in the data. With this code, we can take, say, 20 frames for each of the 13 exposures, giving us 260 images, which would be far too many to take by hand. Furthermore, we will need to perform this test with various exposure times, several times over, covering our sensor with different thicknesses of glass to simulate a window across it (as there must be for our Kepler sensor, which is held in a vacuum-sealed chamber).

After completing this first code, I began writing a secondary code, which will (once it is up and running) read in all our data from a given set, and save the image files, creating objects which we can alter, view, and analyse in whichever ways we want. The secondary code is nearly done now; I just need to debug a few lines and everything should be working fine.

Peter and Ashley managed to cut the sheets of polycarbonate for the ceiling today. We put one panel in to check, and it fits now, which is great, but unfortunately, our cleanroom won't be able to get any power for quite a while. We won't be able to use it until we're approved to hardwire it into the wall, which they've warned us will be quite a long process.