Hello! I have been following the CMOS thread with interest.
Since it has been discussing the mechanics of using CMOS for photometry, I thought I would ask this question in a different thread.
Given a choice, how might CMOS be superior to CCD for photometry, especially for the next 1 to 2 years when both types of astronomy cameras are available?
1) Price? CMOS seems less expensive
2) Chip availability - I am not sure what the CCD chip supply is, but I presume it is good for a couple of years?
3) Ease of repair once CCD chips are not available? Is this an issue in fact? How often do CCD chips need to be repaired/replaced as opposed the a camera's electronics and mechanical features?
4) The ability to identify individual problematic pixels? In CCD, a bad pixel would cause a column defect? How often would this cause a problem with photometry over the life of the CCD chip/camera?
Downside of CMOS compared to CCD?
1) Is the downside of CMOS truly the small pixels? I presume many folks use scopes with focal lengths longer than those whose scopes are used for astrophotography. Are the downsides of binning and oversampling small?
2) Might the large FOV cause a problem with air mass extinction for high precision photometry? This would not be a problem with CMOS per se since large chips are available in both. However, large format CCD chips are VERY expensive. Given the the lower cost of CMOS, more folks might start using full size chips. Over about 45 arc-minute to 60 arc-minute field of view, might air mass extinction start to become an issue from one side of the field to the other at 30 degrees altitude and focal lengths under 2000mm for high precision photometry? (Though how often do we approach millimag precision?) I would think those using DSLRs have thought through this issue.
Thank you for guidance. This will be very helpful for the period of time when both types of astronomical cameras are available. Best regards.
Mike
First off, there seems to be a lot of confusion, which IMO is at least in part due to the fact that "CCD" is an architecture, and "CMOS" is a technology. In what is referred to as "CMOS" the architecture involves an amplifier per pixel, whereas in CCDs there is one amplifier per column. CCDs are a relic of the early days of CMOS when adding a few transistors per pixel was a very big deal, as transistors were big and dies were small.
A reason why CCD chips are so much more expensive than their competitors is that the market for them is so tiny. I'd bet that even a high end specialty camera company like Leitz sells more CMOS cameras per year than all the CCD vendors combined; and their market share is miniscule when compared to Canon, Nikon, Sony, . . . But of course the commercial imaging chips, though of high quality, suffer the unique disadvantage of using a Beyer array, and the pass bands of those filters differ significantly from those of any of the myriad of astronomical photometry "standards". For precision photometric applications, yet another disadvantage is that no two manufacturers use precisely the same filters; there are even differences within the product lines of individual manufacturers. There are now some specialty "CMOS" camera vendors who offer cameras without the Beyer arrays, at stiff markups, and one can also find vendors who will remove the Beyer array from your commercial camera. If you take that path, be sure to pick someone whose process includes replacing the passivation layer on the chip.
There is no reason whatsoever why CMOS pixels have to be small, and no reason whatsoever for anyone ever wanting to repair a CMOS or CCD chip.
If you refer to the DSLR photometry guide, you'll see answers to some of your other questions. For instance, since the response of each pixel is measured individually as part of the flat field and dark field calibration processes, it is determinable which pixels are "hot" or "dead". Photometric software performs primary extinction corrections across fields of view of several degrees extent, even at relatively high air masses.
With even APS-C chips and focal lengths in the 40-50 cm regime, uniformly accurate photometry is possible over fields of view of several degrees. Full frame chips, now widely available in mid-range to high end DSLRs, permit even larger FOVs in principle, but one needs to beware of excessive vignetting. It's not so much that the vignetting cannot be corrected as part of the field flattening process, but rather that one loses SNR that flat fielding does nothing to restore.
I don't think this is a correct representation of the marketplace if you look at cameras in general, not DSLRs.
It are CCD sensors and cameras that are increasingly becoming a "specialty" niche as more and more vendors are phasing out their CCD offers.
Monochrome (w/o color filter arrays) CMOS sensors, and cameras, are mass produced for some time now, coming from the industrial and surveillance camera segment (with advances in artificial intelligence image recognition, the demand for those sensors grows a lot). As for astronomy cameras, those CMOS sensors are used in consumer product lines like the popular ZWO ASI cameras for some time now, and in the low to medium price segment mostly. Only lately you can find popular CMOS (astro) cameras in the higher price range.
HB
Hello! Thank you all for your comments.
Just curious. Which programs might automatically correct for extinction for wide field images?
During a typical run on a clear night, I might have seveal hundered images over three filters. With my narrow FOV, I do not need to correct for differential airmass. However, this situation with wide FOV. The DSLR manual suggests that anything over 0.5 degrees might need extinction correction, especially near 30 degrees elevation from the horizon.
So, as I think about wife field imaging with CMOS chips, I would be curious to know any programs that might automatically correct for air mass extinction. Thank you and best regards.
Mike
First, a couple of notes:
(1) Let's understand that changes in high-end camera chips are being driven by the utterly dominant biology/medical demand, not by astronomy. It's not CCD that is the niche--they are still as sensitive, productive, response-linear, and widely used as they ever were. It's astronomy altogether that is the niche market, and astro photometry not even that. If astronomy accounts for as much as 5% of high-end CMOS demand, or if biology/medical accounts for less than 90% I will be surprised. Consider who bought FLI, and why. Let's ask: is medical research or amateur astronomy a more promising future market?
(2) I'm not sure what CMOS has to do with wide fields of view and extinction. CCDs are and have long been routinely available in wide fields of view, so the within-fov extinction problem is far from new. Even so, no I do not know of any software that does differential extinction within fov. One could write one's own scripts, but it would have to be from scratch. For example I don't see extinction handling in astropy at all. Myself, I would worry more about good flat corrections (including inaccurately corrected vignetting) than about cross-fov extinction gradients at 30 degrees elevation.
-------------------------
Since this topic's title is very general, let me give a CMOS/CCD outlook from inside my very active photometry program:
As of right now, CMOS fails to perform as I need for photometry, where CCD succeeds. I am obviously not alone. That hardly renders CCD a relative niche or "relic". Yes, this could well change. Whenever the camera vendors decide to actually make a CMOS camera that's fully photometry-ready right off the shelf (as CCD cameras have been for many years), I'll eagerly consider it. But the camera I choose will be as performant and as boring (tinkering-free) as possible. My camera is just one citizen of a complex scope rig at a remote location, and it will behave itself night after 400-image night or I chuck it out.
I'm not sure how CMOS can solve any of these current photometry limitations of mine: exposure time, noise, filter wheel reliability, mount tracking, extinction, transforms, backfocus, field-of-view, and certainly not image calibration. Yes, I'm very open to improvements in sensitivity (both in wavelength range and peak QE), but so far the improvements aren't worth the pain. I would also love to extend to much shorter exposures, but only if they are stacked together ON-CAMERA and downloaded as a single image already stacked. This latter condition is non-negotiable. For heaven's sake, the GPU hardware to complete on-camera full-frame stacking within milliseconds is already available for less than $150 and draws less than 5 watts peak, so I don't understand why the CMOS manufacturers are dragging their feet on this obvious potential advantage to astro photometry (unless, again, our niche is too small to worry over).
Again this is all simply my view from inside an ongoing high-throughput multi-filter photometry program that I am reluctant to interrupt. I'm grateful for the brave efforts of CMOS pioneers in this forum thread and others, and I'll be happy should CMOS turn out to be the future. Let's just hope it doesn't turn out that "CMOS is photometry's chip of the future--and always will be".
A slight disagreement: AIP4WIN will produce airmass estimates on a star and frame basis if you choose the Instrumental Magnitude report format. That makes calculation of first order extinction on a per star basis pretty simple.
As to my statement about CCDs being a relic, I was attempting to interject something about the evolution of microelectronics technology into the discussion, to further understanding of why the CCD architecture for CMOS process imaging came into being. And, I stand behind what I said.
Eventually, CMOS cameras will be infinitely superior to CCD cameras because CMOS cameras will be available on the general market while CCD cameras won't.
CMOS advantage over CCD
CCD advantage over CMOS