Digital SLR Imaging

Using DSLRs for Astronomical Imaging

David Ratledge


The arrival of affordable digital SLR cameras on the consumer market has begun a revolution in astronomical digital images. It has brought the ease of use of an SLR camera - optical viewfinder, visual focusing and simple attachment of lenses - to the astro-imager. However, it is the sheer size of the CCD/CMOS chip in these devices that imakes the difference. They began at t6 megapixels and in colour of course. Their general sensor size is in the region of 24mm x 16mm - a size previously only dreamt of by the richest of amateur astronomers. Even full frame (36mm x 24mm) are now on the market and I would expect more. To puchase an astronomical CCD camera with a chip this size needs a serious investment in cash

Despite what an article in December 2005's Astronomy Now said, red nebula do not come out a funny blue/green colour! Red nebula just need either longer exposures or better still, the filter swapping for a clear IR block one. Even with the cost of swapping/removing the filter you will still have a cheap colour camera with a massive chip.

Note somewhere around the 40D Canon introduced double filters in front of the chip. One blocks deep red the other UV/IR. With these later cameras a filter swap is not actually needed - just the deep red blocking filter needs removing. This will change focus postion so camera lenses will not focus according to the distance scale on them and some will not even reach infinity but apart from that it seems the best solution if using it at the telescope.

First generation Digital SLR Cameras - Canon and Pentax. The Canon 300D was one of the first modified cameras in the country (from Hutech)

First to break the 1000 barrier was Canon but all manufacturers, including Nikon, Pentax and have now followed. This puts them in price bracket of the cheaper (cooled) astronomical CCD cameras with comparatively small chips. The chip that was used by Nikon and Pentax in their first DSLRs (a Sony chip) also appeared in the Starlight Xpress dedicated cooled CCD camera at around 4000 although the price did drop later. With DSLR cameras currently well under 500 and conversion costs under £200 it is easy to see why DSLRs offer so much.

But do they work? They are uncooled so will they be bedevilled by noise? Some have automatic dark frame subtraction (a duplicate dark frame is taken straight after the image and subtracted. However it is better to switch off automatic dark frames - they waste imaging time and a single dark frame has too much noise. Much better is to take plenty of them on cloudy nights.

I have used Canon's 300D/Rebel, 20Da, 40D, 60D and the Pentax *ist D. Nikon continues to have a problems with automatic noise reduction (even for raws), which cannot be switched off and deletes faint stars! See link. Currently Canon appears best for Astronomical Imaging with a very definite edge in lower noise - they seem to have built-in dark current adjustment/subtraction. They also have RAWs that, whilst they have been "adjusted", are capable of being calibrated correctly.


QE Graph

Quantum Efficiiency. Canon 40D and Kodak E-series CCD compared.


read noise

Quantum Efficiency

Where dedicated CCD cameras score over DSLRs is their higher quantum efficiency (QE). From the graph on the left, it is clear that a dedicated CCD camera, based on the popular Kodak e-series chips, has only marginally higher QE in the blue but considerably more in the red and H-alpha in particular. The Kodak E-series have a QE of 65% for H-alpha whereas for a modded Canon 40D it is around 25%. However, comparing the QE of mono CCDs and one-shot colour DSLRs is like comparing apples with pears - it is not as simple as it might appear!

In the early days of CCD imaging using mono cameras, colour images were produced by the RGB method ie 3 images sets were taken in turn with a R, G and B filters. Using this simple technique then there is little to choose between a dedicated mono camera and a one-shot colour type such as a DSLR. In fact some R filters used by mono CCD camera imagers reduce the amount of red light transmitted to balance exposures and so virtually wipe out the QE advantage they theoretically should have.

However, modern colour imaging with a mono CCD camera will more likely use the LRGB method. Here the L, luminance, has the maximum signal-to-noise ratio possible as it is using every pixel at every wavelength at full efficiency. However, having taken his luminance data, the mono CCD camera user will have to spend time collecting RGB data - time that does not improve the signal-to-noise of the L image - it only colourises it. The DSLR on the other hand does not have to waste time doing this and therefore, all his imaging time is productively spent improving his image. Swings and roundabouts!

For narrowband imaging, the mono CCD camera will use every pixel to collect the signal irrespective of what wavelength is being imaged. For example, for DSLR cameras, when taking H-alpha images only one pixel in 4 is recording a signal. However, narrowband imaging is still possible - see below for details.



Read Noise

Because we will be shooting many images each with their own read noise and stacking them together it is essential that the camera we choose has low read noise. Canon appear to be leaders in this respect. Left is a log scale of the read noise from one offset (bias) frame from a Canon 40D. Note it exhibits some fixed pattern noise (undesirable) but the histogram is close to a classic bell-shaped gaussian distribution (desirable). In reality this is a pretty good result for a non-specialist astronomical camera.

Note: a bias or offset frame is not the same as a read noise frame. A bias or offset frame actually measures the bias offset in a chip caused by the bias current. A bias (offset) frame contains read noise as do all frames read from the camera. The easiest way to produce a read noise frame is to subtract the master bias from a single bias frame.


The area around the Flaming Star Nebula - Pentax 300mm lens + Canon 40D

Andromeda Galaxy Pentax *istD un-modded and 300mm telephoto lens. The early Pentax cameras were very noise limited - the latest could have improved.

Andromeda Galaxy: Canon 300D modded and 300mm telephoto lens

M31 Canon



Early Days Nebula Test

Flame Nebula/Horsehead Nebula

Pentax un-modded and 300mm lens (left) - Canon 300D modded and 530mm lens (right)

These were relatively short sub-exposures. With experience it became apparent that the Canons could easily handle 10 to 15 minute sub-exposures with consequently much superior results.

The difference shooting longer and using longer sub-exposures makes:


The following are intended to show the versatility of a DSLR:-  

Pleiades Hutech Canon Rebel/300D and Takahashi 106FSQ

This was 18 x 4mins. An early shot and today I would use much longer sub-exposures.

Perseus Canon 20Da with 28mm lens at f/8

This was 6 x 4 minutes.

California Nebula - Canon 40D modded and Pentax 300mm lens at F/4


Star Trails - 2 hour total exposure. Shooting on Christmas day means no aeroplane trails! Canon300D processed in IRIS using add max.



Lunar Geology - Canon 20Da

Sun in CAK light, August 2006 - Canon 300D modified and Baader CAK filter.
Note CAK image not recorded in standard 300D or 20Da


Narrowband Imaging with DSLRs

Using narrowband filters with DSLRs is possible and can give good results. In the case of H-alpha, as only I in 4 pixels sees red, they are somewhat inefficient but good results are possible. Note: Flat fields can be a problem with H-alpha and DSLRs - certainly with IRIS you need to make the master flat manually.

Orion Ha

Orion and Barnard's Loop. Canon 300D (modded) with Baader H-alpha filter.

In the case of OIII, then both the blue and green pixels will record a signal but not at peak transmission. Still 3 out of 4 pixels will be recording so it is more efficient than H-alpha.

However, there is an alternative! That is to use a (visual) UHC type filter which passes both OIII (with H-beta) and H-alpha at the same time. Then the red pixels will record the H-alpha at the same time as the blue and green pixels are recording OIII. Modern DSLRs don't pass IR (even when modded) so visual UHC filters, which transmit in the IR, are not a problem. Using a UHC filter on a modded DSLR is a highly efficient way of maximising exposure time.

The Veil Nebula shot with an UHC filter (Canon 40D + Pentax 300mm lens)


What ISO?

For the 14-bit Canon cameras, then an ISO of 400 seems to work best - this is my experience over many years. This is also the closest ISO setting to unity gain ie 1 photon = 1 ADU number so there is a logic to it. For the 12-bit cameras then increaseing this to 800 would appear to be sensible. This value certainly works well on the 350D although I always preferred ISO400 on the 300D. Note these comments apply to RAWS - if shooting JPEGs then there is a benefit is using higher ISOs but there are so many other drawbacks with JPEGs that they should not be used for astro imaging.

What exposure? As I said I use ISO 400 as this preserves star colour yet is still fast enough for deep-sky objects. In my eperience the longer the exposure the better with perhaps 15 minutes being a noise limit for the Canon. I routinely use 10 minutes with my 40D which works well especially if temperatures are low.


The images here have have been processed using IRIS. I have found this the best and quickest for dealing with these large colour images. I tried shooting jpegs when they first came out but raws are the only way to go. IRIS handles the whole image processing operation including full calibration using darks, flats, offsets and cosmetic defects. Make sure you tick the dark optimisation box when pre-processing - this overcomes the inevitable differences in temperature and that internal RAW "adjustment" that Canon carries out.


Many with astronomical CCD cameras dengrate DSLRs. Put it down to ignorance. I use both and both types excel in their own ways. Both deserve their place in our imaging tool-kit. I wouldn't want to manage without either.

Which is the best DSLR camera? If you want a camera that doubles for everyday use as well as astrophotography then the Canon 60Da is probably the simplest although it is somewhat pricey. For emission nebula then a modified camera (IR filter replaced) is best. I would personally recommend the Canon 300/350D/450D/40D modified by swapping/removing the filter . Two firms in the UK now offer Canon conversion in the £100 to £250 range. The Pentax can be modified by Pentax Europe but is filter removal not filter swapping. This camera then becomes sensitive to IR which may or may not be an advantage.

Finally, one thing to remember: bigger pixels equals better signal to noise ratio - all things being equal. As pixels inevitably get smaller then cameras will not necessarily get better because they have higher pixel numbers. For this same reason the Canon 5D (MkII or III) and the 6D, with the filter swapped, could well be the best currently available - they have big pixels as they have full frame sensors (36mm x 24mm). So check those pixels sizes before you buy - this is one place where size matters!

Canon 40D, 450D: Despite pixels only 5.7 microns these cameras seem more a step forwards than backwards! They have 14 bit A-D converter giving raw files with a dynamic range of 16,000. They also have live view for easy focusing and this can be displayed live on an attached computer. It could be the best yet! The 50D had a reputation for fixed pattern noise and the 60D may be a better bet if you need resolution i.e. for short focal lengths.

Current recommended modded cameras: 40D, 450D, 1100D (all live view) - or at the budget end: 350D but no live view and the 300D is still worth a punt if under £100 already modified.

Sub-exposures and Signal-to-Noise

An article in Astronomy Now (Jan 2008 - Tech Talk) erroneously explains the benefits of stacking many digital SLR images together. The author has confused how signal to noise increases with increased exposure and states 100 1-minute exposures equates to one 10 minute exposure. Oh dear - there is a square root in there but not like that! Signal to noise actually increases in proportion to the square root of the exposure whether this is a single exposure or a stack of multiple exposures. The single exposure is better because it has a single read-out noise component whereas the 100 1-minute ones will have 100 read-out noise components. To minimse read noise then the longer the sub-exposure, without burning out bright details, the better.


Some manufacturers are now introducing back-illuminated chips. Professional CCD chips use this technology to boost quantum efficieciency up to 95%. Hopefully this technology will appear in mainstream DSRLs soon. The future is bright!


Full colour book which includes DSLR imaging:
"Digital Astrophotography - The State of the Art" edited by David Ratledge
Order Online


Return to main menu