4

This answer to How can I avoid star trails without an expensive tracking mount? is consistent with this answer to What is synthetic tracking, and why would a 35 cm Earth imager be 10-30x better than Pan-STARRS or LSST for interstellar asteroid discovery? in that CMOS sensors can preferable to CCD sensors in high cadence astrophotography applications.

This is backed up by the large number of amateur photographs of moving artificial satellites in Earth orbit from the ground by Ralf Vandebergh (https://twitter.com/ralfvandebergh?lang=en) including the GIF shown below of atmospheric seeing effects in a series of photos of the International Space Station.

Question: Why exactly do "CMOS astrophotographers" prefer CMOS sensors? Is it possible to explain exactly what it is about CCD sensors that either prevent them from performing as well as CMOS imagers for astrophotographers photographing objects moving within the field or make them less desirable? Based on reading answers to the linked questions below, I assume that it's something to do with electronic shutter and readout, but I can't quite understand where the fundamental difference lies within the readout scheme.


Animated gif made from frames of the same imaging session as posted before. Taken with fully manually tracked 10 inch F4,8Newtonian reflector (equatorially mounted), tracked at crosshairs at 6x magnification. ALccd 5L-11 mono CMOS camera.

For more images & animations see: http://www.ralfvandebergh-astrophotography.simpsite.nl/home

Source: http://spaceweathergallery.com/indiv_upload.php?upload_id=153718 (other examples spaceweathergallery, spacesafetymagazine, badastronomy)

Ralf Vandebergh ISS GIF atmospheric effects

  • 2
    Just going by the title of the question... They wouldn't be "CMOS astrophotographers" if they preferred CCD (or really any other non-CMOS) sensor... – twalberg Jul 8 at 16:52
  • @twalberg thank you for your valuable insight ;-) – uhoh Jul 8 at 17:03
5

With a CMOS sensor, the analog-to-digital converters (ADCs) is on the same die as the imaging sensor.

With a CCD, you get analog signal out of the chip, and an external analog-to-digital converter (ADC) needs to be used.

There are benefits and drawbacks of each approach. For example, you can pair CCD with a really good ADC having high dynamic range, and then, you can have higher dynamic range than with CMOS.

With a CMOS sensor, you get less noise because ADC is on the same chip and the analog signal needs to travel a shorter distance to the ADC.

Furthermore, CMOS sensor can have multiple ADCs easily, meaning readout speed is higher than with CCD.

Astrophotography is all about noise. You want to use high ISO in astrophotography to capture the minimal amount of light that stars provide to you.

  • Thanks, this is just the kind of answer I was hoping for! Being able to have multiple ADC on-die sounds like a big advantage for high-cadence astrophotography, and I'm still going to have to get used to the idea that CMOS imagers can have comparable noise to CCDs. They didn't used to (okay that was during development back in 199x...) – uhoh Jul 7 at 6:36
  • A shorter path to the ADC does not IMPLY lower noise ... and having both sensor and ADC made on the same die could even force you to use a process that is not optimal for either.... – rackandboneman Jul 7 at 19:26
3

The CCD (Charged Coupled Device) sensor consists of row after row of photosites called photodiodes. During the exposure, photons bombard causing the photodiode to gain an electric charge. The more photon hits at any given photosite, the greater the accumulated charge. When the shutter is closed, the charge is then moved within the site to a storage area. Next the charge is moved again into an area of the chip called a transfer register. Here the magnitude of the charge is read out and converted to a voltage. This voltage is incredibly weak so next it is amplified. This movement followed by amplification is completed one row at time. Once a whole row has been operated on, all charges in the row are delated. Now the next row is now marched into the transfer register and the process begins again. This coupling of rows results in an analog voltage that represents the image. The analog voltage is then transferred to adjacent chip where it is converted to a digital signal.

The CMOS (Complementary Metal Oxide Semiconductor) sensor also contains rows of photosites arrange in a grid pattern. However, in this design much of the processing occurs within the photosite itself. To accomplish, each photosite contains a converter and amplifier. This method induces far less noise as the data is worked on directly, no need to transfer data out row-by-row. With many functions built into the CMOS chip, the circuitry is reduced, thus less electrical consumption resulting in more economy of operation.

Because the CCD must shift charges around the CMOS has a speed advantage. The circuit of the CMOS results in power saving allowing for higher sensitivity to light in each sight. The chief disadvantage of the CMOS is, each site has its own amplifier. Each will have slightly different efficiencies thus the CMOS has higher fixed pattern noise. Because the CMOS amplifies in the sight, there is less possibility for Blooming, a charge leak can induce cross-talk into adjacent sights.

3

The other answers cover CMOS vs. CCD. There is one additional consideration that has not been mentioned in either the question nor any of the existing answers that is a significant consideration to serious astro-photographers: cooling capacity.

Most serious astro-photographers prefer using sensors not encumbered within typical consumer cameras. This allows them to use more powerful cooling methods than those employed by cameras designed to be handheld for more general use. Whether CMOS or CCD, the higher cooling capacity allows keeping chip temperatures lower for extended periods of time required to do astro-photography.

  • This is a good point, thanks! I wonder if there are any differences in the benefits of cooling between CMOS and CCD, or if one can run colder than the other? fyi I noticed that the imager mentioned in the question has a recommended cooler but only a fan removing excess heat from the electronics – uhoh Jul 10 at 0:51
3

The key factor with CCD sensors when it comes to high speed (or what we call Lucky) imaging is the readout speed. CCD cameras due to the nature of their technology, which moves charge (electrons) from one pixel to the next down the columns, one row at a time, favors SLOW readout. Further, due to this charge-shifting nature of CCD sensors, there is usually no room on the sensor die for the more complex kind of readout logic that CMOS sensors employ (per-pixel amps, per-column ADC and CDS units, etc.) As such, all pixels on the sensor are usually actually "read" by a single, or possibly pair of, off-die amp and ADC units. This also favors SLOW readout, as at a high frequency these components also have more noise (input referred noise, the primary sources of read noise.)

So for clean, low noise results, CCD sensors generally must be operated at lower frequencies, which leads to slower readout. In fact, many popular larger-frame CCD cameras can require many tens of seconds to read out a single frame! Obviously for high speed imaging, having to spend 20-30 seconds just reading a frame is untennable.

CMOS sensors differ from CMOS in that rather than moving charge around the sensor, then off the sensor, they immediately convert charge to voltage right in the pixel. That voltage is then simply "applied" to the rest of the readout logic.

Further, with all of these components on the sensor die, they can be highly parallelized. Pixel amps are usually shared among small groups of pixels, either 2 or 4, so for sensors with tens of millions of pixels, you have millions of amps. Each amp can operate concurrently.

Most modern CMOS sensors use per-column ADC units, also on the sensor die. For sensors with thousands of columns, there are thousands of ADC units. Again, these can all operate concurrently.

enter image description here

With all of this high parallelism on the sensor die, this allows every pixel of a given row to be read out simultaneously...to the tune of thousands of pixels at once concurrently, in contrast to CCD where each pixel must, in fact, be read out serially one (or maybe two) at a time. Because each readout circuit is replicated per shared pixel or per column, they do not need to be clocked each at a high frequency. Speed comes from parallelism, not clock. So you maintain the benefits of low frequency (slower clock) for reducing noise (same as with CCD), but can still read out immense volumes of data very quickly.

With all of this parallel performance, reading the full sensor frame in a fraction of a second is possible. This allows very high frame rates (with some CMOS cameras that have hardware ROI features so you can reduce the area of the sensor actually read out), frame rates can range from 20-40fps for the entire frame to as high as 700-800fps for smaller ROI. For the most part, true lucky imaging, where you are operating at the frequency of scintillation (form of atmospheric turbulence caused by the jetstream), occurs with exposures around 10ms, or 1/100th of a second, 100fps. At this frequency you can discard poor quality subs due to seeing, keep only the high quality subs, and achieve high resolution, high detail images of moving objects like ISS, satellites, planets, etc.

It should be noted that the slower readout of CCD is not necessarily always a bad thing. A lot of astrophotography, long exposure stuff, is still done with CCD cameras. And the quality can be exceptional. CCD is slowly diminishing over time largely due to cost. A larger frame (36x24mm, 37x37mm, 52x52mm, etc.) monochrome CCD sensor is easily in the several tens of thousands of dollars range, up into the hundreds of thousands of dollars.

There are some new 36x24mm monochrome CMOS sensors coming onto the market from Sony soon that will be finding their way into cameras that are sub-$10k, and one may even be closer to $5k. While this may seem crazy expensive, in relative terms to similar larger frame CCD cameras, these are actually huge savings. The price of smaller CMOS sensors so far has been giving CCD cameras a run for their money, and have opened up the world of more advanced astrophotography to a much, much broader range of amateurs than ever before.

Cost is also a huge driver of CMOS vs. CCD.

  • This is really helpful, thanks for being so thorough! I started with this question which asks about this telescope. Sections 2 and 4 say the focal plane has three Fairchild/BAE sCMOS detectors (with) 2560x2160 pixels each, and they record video at 50 Hz both for Earth observation (0.9 m resolution at 500 km) and for astroid observation. Every 3.3 seconds they re-point the telescope to move the asteroid from on sensor to the next (identical?) sensor, not sure why. – uhoh Jul 12 at 22:46

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.