another dark question

[Daniel-K]

when taking darks do you still leave a time gap in between?

[Sam]

I take my darks in as close as possible to the conditions that the lights are captured. I usually allow about 30 sec between lights and I do the same for the darks. The theory being that the darks will be at the same temperature as the lights. This is with a DSLR so I don't have much control over the temperature, I imagine it's a bit easier with a cooled camera.

[IanL]

For a temp controlled CCD it should not be necessary to leave a gap between dark exposures (or light ones for that matter) since the camera will rapidly reach a stable temperature and stay there.

For a DSLR it is more difficult. The aim is to try to have a set of lights that are all taken at the same temperature (the sensor temperature that is), and then a set of darks that are also at the same temperature. If you can do so, this will ensure that the dark current/noise in both sets of frames is the same and thus can be subtracted from the lights.

Problem is that:

- Measuring the temperature of the sensor is not that reliable. Most newer DSLRs will report the temperature and using astro-capture software (APT, BYE, etc.) you can record that with the file and also view it on screen, but most of the experts don't think it is a 100% reliable reflection of the true sensor temperature or that it necessarily responds to changes in sensor temperature immediately.

- How do you keep the sensor temperature constant over time? For one thing the ambient temperature varies from night to night, and will (typically) fall during the course of a night as well. This has a significant effect on the sensor temperature. During the summer months I find I am getting temps of 28 deg C or more, and in the winter on a cold night I can keep it nearer to 16 deg C. The effect on noise in the images is really obvious. For another thing (most importantly I think) the sensor heats up as you use it to image, and cools down again when you are not imaging, so the temperature is constantly changing as you work.

This leads to a dilemma:

- Either you leave reasonably long gaps between exposures in a bid to allow it to cool a bit. I find this is necessary in the summer especially otherwise it can get so high the camera shuts off.

- Or you keep exposing constantly so that the sensor reaches thermal equilibrium, i.e. gets to a point where it is cooling at the same rate as it is heating and the temperature stays constant. This may take up to two hours according to some test results that I have seen online. This is more viable in winter, but still you can end up with a high sensor temperature vs. allowing it to cool between exposures.

I haven't found the right answer (other than saving up for a cooled CCD). Some brave souls mod their DSLRs to add cooling direct to the sensor, but this is not for the faint hearted and I cannot risk wrecking my camera. Less intrusive is to build a cooler box (basically a min-fridge) that sits around the camera when it is on the scope, which is less efficient but again there are results that show you can then reach thermal equilibrium at a lower temperature than you otherwise would. Some people report that just blowing a small computer fan across the camera when it is on the scope will also keep the temperature down.

[Daniel-K]

what if my darks are cooler then my lights? how will this affect the image?

[IanL]

If you really want to get in to the weeds about darks and DSLRs, read this, but it is heavy going and you'll need to persevere/read a lot of other background stuff to get a good handle on the topic:

http://www.cloudynights.com/item.php?item_id=2786

The upshot is:

- Basically when a photon of light hits a sensor element, it 'bumps' an electron in to the sensor element's electron well. Measuring the voltage of the sensor well at the end of the exposure allows us to 'count' the electrons it contains, and therefore the number of photons that hit the sensor element, i.e. how bright the light source was at that point in the image. Problem is that heat in the sensor also causes electrons to displace in to the sensor wells in the chip as they are just another source of energy. There is no direct way to tell the difference between electrons displaced by light and electrons displaced by heat (the 'dark current', literally because the process will generate a current in darkness).

- Dark current accumulates over time. In a CCD camera it is a predictable manner that is a function of the CCD characteristics, sensor temperature and time. It is easy to measure the dark current, just by taking an exposure with the sensor covered up so that the only source of electrons must be heat. Provided you expose at the same temperature (same heat available in the sensor) and for the same exposure (same amount of time for the heat to displace electrons), then you can get a very good estimate of the dark current and subtract it from your light frame, therefore leaving behind only the electrons that were displaced by light not heat.

- We literally subtract one from the other by taking the measured pixel brightness of the light frame's pixels and subtracting the corresponding pixel value from the dark frame. (We have converted the electrons via their total voltage to numbers using an analogue to digital converter circuit in the camera).

- The problem is that dark current accumulates in a noisy manner. The electrons do not displace in a nice steady stream, but in fits and starts due to quantum mechanical processes. Actually the photons from the target accumulate in a similar noisy manner for the same reason (they are emitted by the same quantum mechanical processes in stars and nebulae). In the case of the light, this is called 'Photon Shot Noise' because the photons arrive like pellets from a shotgun. In the case of dark current is it simply called dark current noise.

- The way we deal with dark current noise is by taking a fair number of dark frames and then averaging them together. This literally smooths out the noise and the more dark frames you have the better estimate of the actual dark current you get. When you subtract the 'master' dark frame from the light, you do not remove all of the dark current noise, but you can show (using statistics) that the more darks you have averaged in to your master dark, the less dark current noise remains in the single light frame after dark subtraction. It is a game of diminishing returns though, and by the time you get to maybe 30 dark frames you'd be better off spending time doing something else. (If you can make a set of temperature-matched dark frames using a cooled CCD, you can re-use them in future and it is feasible to build up a big library of re-usable darks).

- By they way, that is also how we deal with photon shot noise in the light frames; by exposing for a long time, and then stacking multiple exposures we reduce the noise (or more properly we increase the signal faster than we increase the noise, leading to a better signal-to-noise ratio and a smoother image). There are other sources of noise in the light beyond photon shot noise which are introduced by the camera, but long exposure/stacking deals with all these sources at once.

- Okay so how does this work for DSLRs? Unfortunately not as well as we would like. For one thing we already know it is hard to control/match the temperature of the sensor between lights and darks. This will definitely change the amount of dark current, but much more importantly it also affects the amount of dark current noise. By using temperature mismatched darks and lights, you can easily end up increasing the amount of noise in your image, not reducing it as you wanted to.

- SImilarly, as Craig Stark discovered in his article I have linked to above, DSLRs do all sorts of funky processing to the image before writing it to the RAW file. You absolutely are not getting the raw data from the sensor as you would be with a proper CCD camera. For example he measures the dark current and sees that it actually goes down for longer exposures, whilst the dark current noise goes up (which would only happen if the dark current was also going up).

The camera actually reduces the dark current before writing out the RAW file, but doesn't deal effectively with the dark current noise. There is no easy way to decide how we deal with that problem I'm afraid.

So how do you know if your darks are making the image more or less noisy? Well most astro processing packages have the ability to calculate statistics from your image. (If you haven't got one that does, take a look at the free ImageJ package, http://rsbweb.nih.gov/ij/ ).

What you need to do is take your image before dark subtraction, select an area of the sky background (with no stars, galaxies or nebulosity) and measure the variance of that area. Now subtract your master dark, select exactly the same area in the resulting image and measure the variance again. If the variance in the new image is a smaller number than the original, then congratulations, you did some good! If the variance is more than the original, you made the image more noisy so go straight to jail, do not pass "Go", do not collect £200.

Try to get some better darks, or try to eliminate those darks that are not as close a match before you create your average master dark.

You can evaluate individual dark frames using the same method, by comparing the variance (or the standard deviation which is just derived from the variance) of the whole dark frame. The bigger the difference in VAR or STDDEV, the less well matched the frames are in terms of dark current noise. It is more important to have matching dark current noise than it is to have matching dark current, though you should aim for both if you can. (Some DSLR capture programs will allow you to write the STDDEV of the whole frame in to the filename so you can even dispense with the stats package for comparing darks).