| << Back a page |
CCD V's CMOS
(& Planetary Webcam V's DeepSky DSLR)
CCD or CMOS ?
The main advantage of CCD techynology is that it is very immune to noise (indeed, it can be up to 10x better than CMOS). The main disadvantage is that CCD's are expensive to make and consume more power (typically 4 to 5x CMOS).
The main advantage of CMOS sensors is that they are cheap. The main disadvantage of CMOS is that it suffers from more noise than CCD's, especially 'thermal' noise.
Initially, CCD's were a lot more sensitive that CMOS sensors, however CMOS has, to some extent, 'caught up' when it comes to sensitivity. The sensitivity of the latest DSLR camera's (as measured by the maximum supported ISO) now rivals that of dedicated professional astroimaging sensors.
Planetary imaging
For planetary imaging, we typically have lots of light to play with .. and the size of the 'image' in the eyepiece (or on the sensor) is typically very small (eyepiece projection can improve the size, as can the use of Barlows).
However the effect of atmospheric distortion is much more significant when imaging the planets. Distortion causes planetary 'disc' to distort and features to 'swim' about.
But exposure times can be sort (even sub-second) so the best approach is to use a web-cam (whihc ttypically ahas a sensor no more than 1/2" (12mm) square) to take a 'movie' of thousands of frames and throw away the worst distorted during 'stacking' (using, for example, RegiStax ).
Deepsky photography
Here we need large sensors to capture nebula and galaxies .. and exposure times will be much longer (typically 5 minutes at a time) - and whilst atmospheric distortion still causes stars to 'swin around', these are 'point objects' and stacking software (such as DeepSkyStacker) have evolved to cope ...
In amateur deep sky astrophotography, cost and number of pixels outweighs the noise so when Canon started selling a low cost CMOS sensor DSLR camera (the 300D, in Aug 2003), it was quickly adopted for amateur deep sky imaging. As a result, the Canon range of DSLR camera's (especilaly the 350D) became the 'de-facto' standard.
The presence of the IR blocking filter is a big disadvantage when it comes to imaging faint nebulae, and although Canon did manufacture an astrophotography version of the 350D (without the IR filter), it was much too expensive for 'amateur' use (and the 'semi-professionals' stuck with their bespoke CCD's) .. so it failed to find much of a market and has since been discontinued. However, after Canon 'pointed the way', a number of enterprising individuals started to modify their standard 350D's and these modified camera's occasionally appear on eBay. If you seach via google you may still find some-one that offers a midification service. Most will replace the IR filter with optically clear glass (so the camera can still be used with it's normal lens), and, if a custom 'white balance' is used, even take 'mormal' daylight photo's.
Nikon CCD sensors camera's were initially a lot more expensive than the Canon's, so despite the advantages of CCD V's CMOS they never realy caught on to the same degree. It is to be noted that the higher picel count Nikon's are now equipped with CMOS sensors as CCD's seem to have reached a limit at about 10 mega-pixels.
| Approx date of introduction | Camera | Sensor mm & type | Mega Pixels (effective) | Pixel area (µm2) | Max ISO |
|---|---|---|---|---|---|
| 2003(Aug) | Canon 300D / Digital Rebel (& 10D) | 22.7 x 15.1 CMOS | 6.3 | 1600 (10D 'H' = 3200) | |
| 2004(Apr) | Nikon D70 | 23.7 x 15.6 CCD | 6.1 | 58.7 | 1600 |
| 2004(Nov) | Canon 20D | 22.5 x 15.0 CMOS | 8.2 | 3200 | |
| 2005(Feb) | Canon 350D / Digital Rebel XT | 22.2 x 14.8 CMOS | 8.2 | 40.1 | 1600 (3200**) |
| 2005 (Jly) | Nikon D50 & 2006(Dec) D40 | 23.7 x 15.5 CCD | 6.3 | 58.7 | 1600 (D40 'HI 1' = 3200) |
| 2005(Aug) | Canon 5D | 35.8 x 23.9 CMOS | 12.8 | 3200 | |
| 2006(Aug) | Canon 400D (Rebel XTi) & (2008) 1000D & () 40D | 22.2 x 14.8 CMOS | 10.1 | 31.3 | 1600 (3200**, 40D H=3200) |
| 2006(Sept) | Nikon D80 | 23.6 x 15.8 CCD | 10.2 | 34.5 | 1600 (boost=3200) |
| 2007(May) | Nikon D40X | 23.6 x 15.6 CCD | 10.8 | 34.5 | |
| 2007(Aug) | Nikon D3 & 2008(Oct) D700 | 36 x 23.9 CMOS | 12.1 | 67.0 | 6400 (b=12800, HI2=25600) |
| 2008(Jan) | Canon 450D / Digital Rebel XSi | 22.2 x 14.8 CMOS | 12.2 | 1600 | |
| 2008(Mar) | Nikon D60 | 23.6 x 15.8 CCD | 10.2 | 34.6 | 1600 (boost=3200) |
| 2008(Oct) | Nikon D90 / D5000 (2009 Jun) | 23.6 x 15.8 CMOS | 12.3 | 28.9 | 3200 (boost=6400) |
| 2008(Dec) | Nikon D3X | 35.9 x 24 CMOS | 24.5 | 1600 (boost=6400) | |
| 2009(Feb) | Canon 5D MkII | 36 x 24 CMOS | 21.1 | 6400 (H1/H2=12800/25600) | |
| 2009(Mar) | Canon 500D / Digital Rebel T1i | 22.3 x 14.9 CMOS | 15.1 | 22.0 | 3200 (H1/H2=6400/12800) |
| 2009(Jly) | Nikon D300s & D5000 | 23.6 x 15.8 CMOS | 12.3 | 3200 (boost=6400) | |
| 2009(Oct) | Nikon D3s | 36 x 24 CMOS | 12.1 | 12800 (ext=102400) | |
| 2009(Dec) | Nikon D3000 | 23.6 x 15.8 CCD | 10.2 | 3200 (boost=6400) | |
| 2010(Apr) | Canon 550D / Digital Rebel T2i & 7D | 22.3 x 14.9 CMOS | 18.0 | 6400 (H=12800) | |
**higher figure only supported if the CHDK firmware 'hack' is used.
Webcams (and the famous Philips ToUcam)
The very first amateur digital (planetary) astroimaging was done using a webcam !
Over the years, considerable effort has been put into optimising the performance of these 'low end' camera's by making simple circuit changes (to reduce amplified noise, control exposure times etc).
Unfortunately, some individuals efforts were turned into a business, with the result that specific webcam modifications have became 'commercial secrets'.
So, to avoid trouble with those wanting to keep their "secrets" I'll say no more about webcam's (other than to suggest you look on eBay for a bargain rather than line the pockets of the commercial vendors).
CCD security cameras
One source of good low-light performance cameras is the security industry.
Many of the 'low end' CCD 'security cameras' found on ebay are specifically designed to operate in the IR region - indeed many have IR LED illumination so they can operate in total darkness. Unfortunmatly, because of their ability to operate in 'zero light', their low-light sensitivity is often very poor. 1 Lux is typical (1/4" SHARP chip), although the1/3" inch Sony Super HAD Color CCD goes down to 0.5 LUX. Some using a 1/3" SONY chip can operate down to 0.001 Lux in PAL (but presumably B&W) mode, for about £30.
The main drawback with the simple eBay style is that they are fully automatic and will auto-switch to IR mode even if you don't want them to. To be usefull, it is necessary to gain control over the CCD by modifying the circuitry (which may be almost impossible given the higherly integrated / minimal chip count nature of theses units).
More useful are the camera's designed to work in low light conditions (without switching into IR mode) and equipped with control interfaces (usually serial link) - for example, the PC164CEX-2 delivers 600 line resolution at 0.0001 lux for $140. Such cameras fit neatly in price below the 'semi-professional' astrophotography cameras.
Dedicated astronomy Cameras
Dedicated cameras are almost all B&W only (although some come with an 'integrated' filter wheel), and, despite using sensor that have up to 10-50 times fewer elements than that of DSLR sensors, prices are, indeed, astronomical :-)
As mentioned already, smaller chip sizes and pixcels counts mean the camera is really only suitable for planetary imagimg.
| Camera (chip) | Sensor mm & type | Pixels (mega) | Pixel area (µm2) | Sensitivity (lux) |
|---|---|---|---|---|
| DMK 21AU04.AS (Sony ICX098BL) | 4.6 x 3.97 CCD | 0.33 | 31.36 | |
| DMK 31AU03.AS (Sony ICX204AL) | 5.8 x 4.92 CCD | 0.8 | 21.62 | |
| DMK 41AU02.AS (Sony ICX205AL) | 7.6 x 6.2 CCD | 1.4 | 21.62 | |
Optimum pixel sizes (see also link)
There is no advantage in having pixels smaller than the resolution limit of your telescope's optics. This is given by the diameter of Airy Disk (in mm) = 2.43932 x λ x Focal Ratio (where λ = Wave Length in mm (e.g. 546nM = 0.000546mm)
A typical 5" Refractor (127mm dia lens, 1000mm focal length) = f8 & we have Airy Disc dia. approx 10 um (= area of about 31 µm2 ). A typical 10" Reflector (250 mm dia. mirror) will be f4 and have a limit 5um = area 15 µm2.
As will be seen from the tables above, many modern camera sensors are getting close to these limits.
Colour 'sensors'
All sensors simply measure the intensity of light falling on them. To generate a colour image it is necessary to use RGB filters. With a B&W webcam or B&W security camera, you exposure 3 (or more) images using a different filter for each exposure.
Note - most astronomy filter wheels have 4 positions - this is because any filter will reduce the available light falling on the sensor so it is normal to expose one image in 'B&W' mode i.e. 'without' a filter (to get the overall 'intensity' of light) and then 3 filtered images (r, g, b) to get the overall colour.
In a DSLR a colour result is obtained in a single exposure. To achieve this, the final colour image is generated from sets of 4 'sub-pixels' arranged in a 2x2 array known as a bayer matrix. Individual sub-pixels have built in colour filters (part of the 'micro-lens' structure above each sensing element that is used to maximise the light falling on it).
The 2x2 sub-pixel array is arranged in a 'Bayer Matrix', with 2 G's, 1 R & 1 B as follows :-
| G | R |
| B | G |
Needless to say, an '8 mega pixel' camera that outputs an 8 mega pixel RGB image actually has 'only' 8 mega (sub-)pixels in total i.e. it has 2 million Bayer Matrix 2x2 sets (and thus 2 mega pixels are Red, 2 mega are Blue and 4 mega pixels are green !)
To 'convert' the Bayer Matrix pixels into RGB, the camera performs INTERPOLATION to 'guess' the color / intensity of the RGB from the nearby Bayer Matrix pixels ...
Each (3x8 = 24bit) RGB pixel in the resultant image is thus derived from 3 to 5 'sub-pixels' in the sensor. Cameras that allow the user access to the "Raw" sensor data will output the actual individual (10, 12 or 14 bit values) found at each of the underlying sensor sub- pixels....
Needless to say, this is not what we really want when trying to maximise the sharpness of point objects (such as stars).
| << Back a page |