A flat-field data product shows “typical” amount of starfield variation across most of the PUNCH
field of view, and photon noise level (associated with the solar F corona) near the center.
PUNCH is doing something very ambitious: merging together images from four separate instruments mounted on four separate spacecraft. Doing so requires very, very precise determination of the conversion between each pixel’s value and brightness on the night sky. That’s because, in one 90-minute orbit, the same feature on the sky is imaged by at least eight different pixels across different regions of at least three different cameras. Combining those data requires us to know exactly how to convert the value of each pixel to brightness on the sky.
On the ground, we measured the “flat field” sensitivity of each camera, using smoothly illuminated optical diffusers. But those are only good to something like ten-percent precision, which is not nearly precise enough. PUNCH uses the sky itself. Initial measurements of the “optical gain” function used the brightnesses of individual stars. But those run into difficulty, because of something else – optical aberration, which moves photons from the core of a star’s image to a smeared-out “coma”. The aberration is typical of most optical systems (the first paper to complain about it was written by some guy named Johannes Kepler, together with his Italian buddy Galileo Galilei, in 1611). We correct it on the ground, using complicated mathematics. Tracking individual stars also leads to a host of other difficulties. So we’re checking that work using a simple cross- comparison process.
We took each of about 1,500 PUNCH images of the sky and projected them from the camera frame to the heliocentric output plane of the PUNCH constellation, and then measured how much each neighborhood varied in brightness. Merging all of those produced the image above: a σ- map (“sigma map”) of the typical amount of brightness variation in each neighborhood of the overall PUNCH field of view, averaged across all of the cameras in PUNCH.
The next step is to remap all of those average brightness variation levels back to the focal plane of each of the 1,500 images. Then we can compare the variation in each neighborhood of each camera, to the PUNCH consensus average on the sky itself. That gives us a map of how each camera responds, and allows us to cross calibrate them to within a small fraction of a percent.
There’s one issue: see that bright yellow “doughnut” in the middle of the figure? That is the part of the image that is dominated by photon counting noise: quantum uncertainty in the amount of energy that hit the detector at that location. For most of the figure, the brightness of the figure is proportional to the sensitivity of the camera, but photon counting noise isn’t. It varies more like the square root of the sensitivity of the camera. So we can’t use this method to measure the flat field too close to the Sun. But it works like a champ over most of our field of view.
Another minor issue is that you can see some horizontal banding near the top and bottom of the image. Those are statistical imprints left by bright stars as they cross through the field. But they fade when we scale up to the full analysis, using the over 80,000 clear images (and over 160,000 images for each polarizer position!) we have collected so far in the mission.
PUNCH data analysis is simple in concept – but when high precision is involved, almost everything grows more finicky and complicated. We’re making great progress, and rolling out new versions of the data every couple of weeks. They are only growing more spectacular with each release.
Version 0h of the data, which will start appearing in NASA’s Solar Data Analysis Center archive over the next few days, achieves roughly 1% relative photometric precision over most of the field of view. We still have work to do, but we’re getting there!
This is PUNCH Nugget #23. PUNCH nuggets are archived at the PUNCH mission website. You can sign up to receive PUNCH nuggets by email.
