Long time ago we have been doing astroimages with negative (or positive) film. We had one chance to take a long exposure photo. Then we have been given with CCD sensors – much more sensitive. And we also learn to stack many frames into final picture. Today we have also CMOS sensors. They are nor better neither worse. They work using the same phenomenon as CCD, they are similarly effective, and they differ in two aspects: CMOSes are faster and has low read noise. This read noise is sometimes underestimated and unfamiliar quantity. But it is changing astroimaging world. Of course there are other important factors – camera sensitivity (QE – quantum efficiency), thermal noise (dark current) that affects astrophotographers final work. But read noise affects the way we do astroimages.
How does it come?
It is a chain of dependencies. First CCD sensors were expensive and had really high read noise at the level of 100 electrons. Then it quickly went down to the level of around 10 electrons and stuck there for a long time. All popular cameras that astronomy amateurs loved 10-15 years ago had read noise at this level. And that implied the way we have been making astroimages for a long time.
What is read (readout) noise at all?
It is the noise that is added by camera electronics to every pixel value during readout proces. This value is constant, do not vary with exposure time and varies very little with sensor temperature. However for CMOS cameras (and DSLR as well) it varies with gain (ISO) setting.
How does astroimaging setup work?
Usual setup contains several components: sensor (camera – either dedicated, or DSLR), optics (telescope, lens) and mount that is able to track stars move (guided or unguided). Taking all the setup parameters we need to choose suitable target, exposure time and number to achieve desired outcome. There are quite a lot of degrees of freedom.
Large read noise way of imaging
For almost all CCD sensor cameras read noise is at the level of 10 electrons. This implies the way of imaging we got used to over last dozen or more years. We stack deep sky images with subframes, and the exposure time for each subframe is in the range from 1-2 minutes (for luminance with fast lens) to half an hour (for narrowband images with slow scopes). The reason for that is that we need to collect enough photons in each frame to make read noise insignificant.
Low read noise way of imaging
So this is for CMOS astroimaging cameras, but also in some extent for DSLR cameras. For this sensors read noise varies with camera gain setting. It is usually at level of 2-4 electrons for zero gain (low ISO) and can drop below 1 electron for high gain setting. The rule of collecting enough photons in each subframe to make read noise insignificant still applies, but the read noise is lower, so we can collect less photons and subframe exposure time can be shorter. Of course you can still use CMOS in “CCD way” – at zero gain setting and long subframe exposures.
So shorter subframe exposure time, what’s the big deal?
As I mentioned before – it is a chain of dependencies. Probably the reader is aware of so called planetary imaging, when we use CMOS cameras at high gain setting and ver short exposure times (in the milliseconds range) to collect large amount of subframes and then select best of them and stack to achieve high resolution image. This is the very opposite end of long exposed subframes with CCD cameras. But there are whole range of subframe exposure times between minutes and milliseconds that we can use. And in this range we can select some optimal values suitable for our needs. There are several aspects to consider.
Seeing and resolution
Short time subframes have several advantages:
- seeing is not constant – it varies over time. During short exposed subframe seeing varies less and it is possible to achieve overall better resolution
- assuming the same total imaging session time, when you do short subframes you have more of them. After session you can analyse frames and select best of them to create final image stack with better resolution
- tracking and guiding errors are less in short subframes
- stacking large number of subframes gives better outcome in terms of pixel rejection
Tracking and guiding
Using short subframes exposure time makes tracking error less visible, but also you can just give up on guiding. It is valid even for large pixel scale setups, when you use subframe exposure time less than 5-30 seconds (depends on the mount tracking ability). This way you can do astroimages with large instruments (like 10″ newtonian or SCT telescope) on entry level mounts like EQ6 without guiding.
It is too good to be true!
Of course, short exposed subframes also have drawbacks. Here are few of them:
- you need lot of space to store large number of subframes and also lot of CPU power to process them
- since read noise is not zero the signal to noise ratio will still be worse for large subframe number than for small amount of subframes. This way it will be harder to reach for faint galaxy halos or faint nebula framgents
- if you do not use guiding then some diagonal background patterns may show up when you stretch the image (even when you calibrate subframes). You can fight with them with dithering – but not every each subframe! You may split your session to do at least 10 dithering moves overall.
What about quantum efficiency and dark noise?
These attributes are completely different issues. Both QE and dark (thermal) noise do not care how you split your session to subframes. If you expose 2 hours of subframes and you split them for 20 and for 200 subframes you will collect the same amount of light photons (QE) and the dark current signal (thermal noise). The same statement is valid for background (photon) noise that you capture at your light frames.
Are CMOS cameras perfect?
Of course not! Perfect camera would have infinite pixel capacity, 100% efficiency, zero readout noise and zero dark (thermal) current. CMOSes are closer to be a perfect due to low read noise, but still not zero. If the readout noise value could be zero, then the final image effect would be identical for one 100 second frame and for stack of thousand of frames 0.1s each.
So how to choose optimal subframe exposure time?
There were whole lot of discussions on this subject already made for CCD sensors. These analysis are still valid for CMOS cameras. However there is some additional benefit for taking short subexposures. Some signal to noise ratio is sacrificed, but we gain some resolution. It is a compromise. Depending on the target, imaging setup and local conditions one need to compile its own best practices. Some hints may be useful:
- very short exposures (less than 0.1s) are mostly reserved for planetary and lunar imaging. But if you have large aperture you can also use this exposure range for bright and compact objects, like bright planetary nebulae, some globular cluster center areas, double stars, etc.
- short exposures (0.1-1.0s) are quite useful for doing high resolution images of brighter objects even with modest apertures (like 8 or more inches)
- another exposure range is 1 to 10 seconds. This is a compromise between resolution (at few seconds time span seeing is not frozen and also some mount tracking issues may occur) and signal to noise ratio (at few seconds exposure time and high gain read noise becomes less significant in the total image noise)
- over 10 seconds up to maybe 20-30 seconds is the range when you can still image at high scale without guiding on entry level mounts. 10 seconds may be still available for some mounts of EQ6 class with low periodic error. 30s unguided subframes are more suitable for better class mounts. However for the exposures in this range you should consider using guiding.
Anything more than 30 seconds will not provide better resolution than longer exposures. For the exposure times in this range there are other aspects you should consider: optimal subexposure time, quantization error (for CMOS cameras), mount tracking ability, etc.
So how to choose optimal gain setting?
It will depend on chosen exposure time and target. For very short times (1s or less) we should aim for high gain, more than 50% of the gain slider (50% ot the gain slider usually corresponds to the gain of 25-30dB, that means gain about 20-25x). For exposures in the range 1-10s we may be in the gain range below 50% but over unity gain. At longer exposure times we may image at camera unity gain. At high gain we make readout noise lower, but the side effect is also less pixel capacity, so brightest stars may be oversaturated. It is – as usual – a compromise.
In practice you can measure and estimate the best combination for your setup and location. Just make a series of images with different exposure times: 1, 2, 3, 5, 10, 15 seconds and analyse them later. Measure FWHM and star flatness/eccentricity. As target you should select some object with fine details – dense open cluster, bright globular cluster or tight double star. This way you will get reasonable information on your local conditions, mount tracking and optics quality and distribution of star FWHM over time series. But you need to remember that your local conditions may vary as well. And of course your imaging setup should have large enough pixel scale to sample image well. For setups with pixel scale 2-3 arcsec/px or above it makes little sense to do such high resolution attempts.
You can read more on this subject in related articles below, or in general in this section.
Cover photo by Nathan Anderson on Unsplash.