It's very exciting to be a (small) part of this, happy to answer any camera software questions (can't speak for the observatory's software though as I haven't seen it)
Are the cameras similar to what's in a consumer digital camera, that is, a single image sensor behind a bayer filer and a lens? Or does it use some other configuration, like an array of image sensors?
And does sensor readout work similarly to a consumer camera, sequentially reading out rows of sensor data? Is there any cool software processing during the capture, like decovolution?
Yes they're quite similar to consumer camera sensors, our sensors are usually from high quality production bins. We advertise this quality as "scientific CMOS" (sCMOS) to help highlight this. Consumer sensors can have a significant number of sensor defects which can be corrected so they aren't noticeable in casual photographs, but these defects are very detrimental for scientific imaging where quality is paramount. Another big difference is the noise and quantum efficiency characteristics of the sensor which is another key requirement for scientific instruments.
We don't supply lens', I think the logic is that scientific customer's know exactly what kind of optical setup they want so most customer's would tend to use their own optical equipment or buy it in.
Our camera's are monochrome (scientific cameras tend to care more about raw resolution than having a smaller res with bayer layer) so customers typically use different color/wavelength filters to get what they want and process them into true color images later if needed.
> Or does it use some other configuration, like an array of image sensors?
This particular camera, the Zyla has just one sensor. Though it is a little unique in our portfolio, in that the sensor can be read out from both halves simultaneously in various patterns. If your interested in the hardware we provide lots of info in our hardware manual: https://andor.oxinst.com/downloads/uploads/Zyla_hardware_use... I don't think we offer multi-sensor solutions, though I could be wrong.
> And does sensor readout work similarly to a consumer camera, sequentially reading out rows of sensor data?
Yes, there are two electronic shuttering modes we offer: rolling and global. Rolling takes a sequential row by row readout, and global does a readout of the entire sensor. The camera's used by the observator can only do rolling, but we have other Zyla models which also do global. There can be tradeoffs in choosing which one to use, typically framerate, noise and image distortion are the key factors in choosing. Global is available on some high end consumer cameras, but generally most consumer sensors will do rolling. Though this may have changed since I last looked.
> Is there any cool software processing during the capture, like decovolution?
In the camera side of the company, we try to leave the image as clean and raw as possible. We perform correction processing during acquisition on the camera; as high quality as the bins are, you still have to correct and characterize for various things to get the best performance in a scientific scenario.
In the applications side of the company we do all kinds of image processing: deconvolution (this is a big deal in the confocal microscopy world, we have our own patented deconvolution method: srrf-stream) https://fusion-benchtop-software-guide.scrollhelp.site/fusio..., AI analysis, 3d/4d imaging (https://imaris.oxinst.com/). Probably lots more I don't know about (I'm on the camera side).
[0] https://sites.google.com/inaf.it/shark-vis/instrument/detect...
Correct, it depends on the observation. Both sides have adaptive optics correction, but they work independently. This particular instrument (SHARK-VIS) is mounted on the "right" side, while SHARK-NIR is on the "left" side.
Or try: https://andor.oxinst.com/learning/view/article/introduction-...
Adaptive optics in particular requires very fast framerates and low latency to make rapid adjustments to the mirror's shape to compensate for the constantly changing atmosphere. It's really amazing that it's possible at all! I believe this is the method used here, though I can't say with certainty.
Lucky imaging is more akin to a brute force method, where you acquire lots and lots of images quickly and process the best ones when the atmosphere was being particularly cooperative at the time and not distorting the image very much.
Again, there are lots of experts out there on the topic, this is just my simple view into it.
When light passes through the atmosphere, it undergoes a convolution known as a point spread function (think of it as convolving the signal with a 2D gaussian that spreads the intensity out to neighboring pixels). If we know that PSF specific details, we can deconvolve the image, either computationally, or by modifying the mirror in real time.
From my understanding, you can project a laser into the atmosphere, where it gets affected by the PSF. When you look at that laser projection, you can find the PSF (because you know the input shape of the laser, and what it looks like after being affected by the PSF), and therefore use that in real time to deconvolve the astronomic images you are collecting.
This process can be done so quickly it can adapt to immediate changes in the atmosphere (turbulence). "Enhance" is definitely a thing- it's widely used in both telescopes and microscopes (and if you had the right priors for a blurry photo, you could do it there too).
I think this is a relatively simple read: https://en.wikipedia.org/wiki/Laser_guide_star along with https://www.llnl.gov/article/44936/guide-star-leads-sharper-...