DSLR Shooting - How to manage Darks, Flats, Bias, etc. - Forums

Hi,

Does anyone have any good tips on acquiring Darks, Flats, Bias, etc?

I am very new to this all, shooting with DSLR, and would love some insight to improve my photos by using such.

Thanks.

Darks(and lights): Cover your viefinder, in the beginning I had some artifacts because of light leaking in. Always done at the same ISO as your lights. Also, the temperature should be at least similar (in the range of 3°C) to the one you are shooting the lights at. If you image outside in a cold night and take the darks, say, in a dark heated room you will get overcorrection. In this case, you could take your darks in a fridge instead (supposedly it's dark in there once the door is shut ;-)

Flats: white t-shirt flats do work, but to get a relly good flat build or buy some flafield box. for a small Apo, you can easily an inexpensively build one from a piece of electroluminescent foil, opaque acrilyc glass, and an inverter plus the 12v-plug of your choice. Shooting with a DSLR, pay attention to the histogram: The RGB peak should be roughly in the middle, and no colour channel further than 2/3 of the way to the right. Always done at the same ISO as your lights

Also: If youe are using a flatfield box, be careful that the expose times don't get too short, rather dim down the light source and take longer flats. The reason is once you get below 1/50s, you might get rolling shutter effects due to the light sources flickering. Above that, it's averaged out.

Bias: Some people say for DSLR it's best to take flat darks instead, which is to say darks which are the same exposure time as your flatframes. I always did it that way and it works well for me. Use the same ISO as your lights.

Hope this prevents you from stumbling into some of the pitfalls I stumbled in ;-)

Clear skies, Marc

Also: take as many Darks as possible, roughly the same number as your Lights, but at least 25 or so. In the beginning I would take 5 darks, but in this way you will introduce more noise into your pictures than you will improve it.

for the Flats and Flat Darks: I tend to take 25-30 on my DSLR

Hi,

In addition to what Marc has written :

Marc Agostini:
Always done at the same ISO as your lights

Well in fact flats don't have to be at the same ISO as your lights, and it is both more convenient and more efficient to go down to ISO 100 for flats : as one flat at ISO 100 is equivalent to 16 flats at ISO 1600 for example, you will have to take much less of them and they will be improved (compared to higher ISO). For example, 3 to 5 flats are enough for deep sky (if they are nicely done obviously !)

Same remark for bias, you can do them at ISO 100, as most - if not all ! - cameras don't have an average bias signal depending on the ISO setting : convenient too!

Clear skies !

Once you have them, organize them so you can find them and re-use them.

Darks and bias frames are good for months, or even years. Flats have to be re-shot more often.

Don't delete your old flats. As you learn new processing tricks you may want to re-process the old data, and you will need the old flats for that.

Regarding flats, I used a white t-shirt covering the lens and a rubber band to keep it stretched. I shot the flats after dawn, using the blue sky. Right at dawn the brightness of the sky changes pretty quickly, but once the sun is over the horizon, brightness stays steady enough for a series of flats. Another VERY important point when shooting flats is NOT to touch the focus (or anything else). The focus must be exactly as it was shooting the object of the night. Otherwise donuts etc will move slightly making the flats unusable.

Bias is easy and quick. Just cover everything with a dark blanket or whatever, and shoot as short exposures as you can. Shoot many bias frames and make a master bias stack for your later calibrations. Bias and flats are bang for the bucks.

Darks are nice, but they are time consuming. If you're in a hurry, you can simply use bias and flats only, and then replace darks with software pixel rejection for a decent result. Otherwise you can shoot darks any time you like. The camera doesn't even have to be mounted on the telescope. Some cameras tend to "drift" over time, meaning that darks will have to be reshot after some time (months), but normally it makes sense to keep a library of darks with different exposure times.

CS

Hi everyone,

I did my first attempt at integrated flats and darks (but forgot to do dark flats and honested don't yet understand bias and would love more information on that).

Here it is: https://www.astrobin.com/p45ndjI would really love to achieve less noise and greater clarity in detail. My total integration of lights is 3.5 hours. I used about 40 flats and about 70+ darks.

Would the use of dark flats and biases help me with reducing more noise or do I simply need more exposure time overall to achieve this?

Thanks for your help!

Hi everyone,I did my first attempt at integrated flats and darks (but forgot to do dark flats and honested don't yet understand bias and would love more information on that).

Here it is: https://www.astrobin.com/p45ndjI would really love to achieve less noise and greater clarity in detail. My total integration of lights is 3.5 hours. I used about 40 flats and about 70+ darks.

Would the use of dark flats and biases help me with reducing more noise or do I simply need more exposure time overall to achieve this?

Thanks for your help!

Actually these calibration frames (darks, flats, bias) will not help you reduce noise or improve detail.

Bias frames remove the DSLR sensor's "read" signal, which is usually columnar and present heavily on the bottom of the frame. Although you'll probably benefit from them, I don't see those issues in your image. Flat frames calibrate out uneven sensor illumination, especially vignetting. Although critical to obtain even fields, they won't help with noise either. Darks remove thermal signal that gets introduced predictably in time as you integrate your data. None of this is particularly relevant to reducing noise. Only more integration time (or better sensors- you likely don't need one, or noise reduction algorithms) really help this.

I don't use dark frames whatsoever with decent success. The bias frames are critical to establish "lows" in the automated algorithms during preprocessing, but I find that dark frames eat up a ton of time I could be using to collect data. DSLRs seem fine without dark calibration as a rule. Although this does have me thinking if I should reevaluate what my results would be with them! I haven't seen evidence of the need (I use the Nikon D5300).

Post your integrated image, as well as a single light frame, and we could see if the noise is inherent in your image (to that extent), or if it's from your post-processing/clipping the darks.

It is very easy to take dark frames, bias frames and flat field images with your DSLR. When you in the evening prepare your equipment it is often bright enough in the sky to take the flat field images, mostly they are blue and make 30 images but keep in mind that you use the same ISO as the other images. For instance, if you choose ISO 800 with taking flat field images, the other images must be ISO 800. For focussing at the scope with your software program you can go further than ISO 800 to focus the star. I use EOS Utility because the program can taken very sharp images to focus and when you have a Newton scope chose a very bright star so that at focussing the spikes fall togethet until they are one line and you are ready to take images. And forget not to attach a light pollution filter if your location is situated in a light pollution area. Choose the exposure time when it is fully dark in the sky. I have a Epsilon 180 ED af F 2,8. I can go with ISO 800 and a filter to 2 minutes exposure time, the background is still dark and you can see the wanted object. Make many images so much as you can special for nebula and other objects such as galaxies, planetary nebula. Star clusters do not need many images, but when you are capture NGC 6992 you needed 90 images of 2 minutes and when you have a F 5 system, you need at least 270 images. The bias Frames can be taken after the session, chose a exposuretime of 1/4000 seconds and cover before the camera for incoming light and use ISO 800. Dark Frames of 2 minutes must be taken after the session at the same night and shoot at least 30 images of ISO 800 so much as you can. You see it is a lot of work, And then at a evening you will start the imaging process and mostly the most users use DSS stacking or a other program. That cost you a lot of time to create the wanted image. I wish you succes with imaging.
Kind regards, Hans Verheijen/ The Netherlands.

Actually these calibration frames (darks, flats, bias) will not help you reduce noise or improve detail

Yes, and worse, calibration frames add noise.

To reduce noise, a very efficient method is called dithering, which can be done manually if you don't control your mount with a computer. By doing so, you also reduce the number of dark frames that have to be taken, 15 for example become sufficient.

If you don't dither, the more dark frames, the better.

More global exposure time will also improve your signal/noise ratio.

When it comes to bias frames, they are easy to take. They remove the offset signal. Just select the shortest exposure time of your DSLR, and take a shot in the dark. Now you have a bias frame. Even if they are quick to take, taking hundreds of them will not really help you much with noise reduction, as the effect would be invisible (they are substracted from flat frames). But if you really want to limit noise addition coming from a bias frame, with DSLRs, you can make a bias frame totally free from noise by mesuring the average offset signal of your DSLR and making a synthetic bias.

On light frames, you can limit the read noise by finding the optimal ISO setting. It depends on several parameters, such as the type of target you are shooting and the light pollution level. Usually, on deep sky objects, a good compromise is when the read noise starts to stabilize (after decreasing) when the ISO sensitivity rises. See here for read noise charts on common DSLRs.

Be careful with over processing too.

Hans Verheijen:
For instance, if you choose ISO 800 with taking flat field images, the other images must be ISO 800

No, and it is even counterproductive to take flat frames at ISO 800 for example...

I never needed darks with the D3300. I need them sometimes with the D7500. Usually when it is very hot, the sensor has been overused and the background is pitch black (they are not needed in the milky way region or around Orion for example where the sky background is yellowish or red because of dust and start clouds). Even 3-4 darks will work wonders for correcting the (very limited) amp glow and banding due to sensor heating and I find anything taken within 10 degrees is good enough for me. The main reason behind the conventional wisdom that darks must match lights temperature is the fact that dark current is strongly affected by temperature, but modern CMOS sensors used in DSLR cameras have dark current suppression.

Flats, I guarantee you are better off using a sensor cleaning kit to correct dust motes (iow keep your sensor clean

) and software to correct vignetting. That takes care of the two reasons for needing flats without introducing noise to your pictures (which is something flats absolutely do).

Bias frames are of course the easiest ones to obtain and can be reused as many times as you want. But I just don't bother with the D7500, the master bias signal is so close to 0 it is completely pointless to use it. With the D3300 it was a little higher but also insignificant compared to your average CCD or my ZWO ``178MC.

I agree with AwesomeAstro: in modern DSLRs calibration frames are used for different reasons, not noise reduction and most of them. The only way to reduce noise is to take many, many dithered subs. Think 6 hours or more.

Romain Perron:
Yes, and worse, calibration frames add noise.

Absolutely true, my flat frames make my image ever so slightly noisier. But for anyone concerned about that, they are more than worth it in removing countless awful gradients that render my images unusable otherwise.

Dithering is definitely important, but more for fixed-pattern noise in the sensor, not general noise from many standard sources. The image linked above seems to be suffering more from the latter, not much of the former. I suspect the biggest issues include ISO value choice, not clipping the dark pixels (important), and by far the most, post processing with masks, both in noise reduction (to slightly help the fainter parts of the galaxy) and in stretching. Masks are absolutely priceless in post-processing.

Hi everyone,I did my first attempt at integrated flats and darks (but forgot to do dark flats and honested don't yet understand bias and would love more information on that).

Here it is: https://www.astrobin.com/p45ndjI would really love to achieve less noise and greater clarity in detail. My total integration of lights is 3.5 hours. I used about 40 flats and about 70+ darks.

Would the use of dark flats and biases help me with reducing more noise or do I simply need more exposure time overall to achieve this?

Thanks for your help!

Hi,
which ISO did you set for this image? Looks like the signal is « flooded » by the noise... 3.5 hours should give you a decent image.
I take darks at same ISO as lights, and bias and flats using the lowest ISO of my camera. Only darks will help to minimise the amount of thermal noise generated by the sensor. Flats are requested to remove optical artifacts, so it’s not necessary to « add » noise with them, to my humble opinion.
Christophe

Christophe Perroud:
Flats are requested to remove optical artifacts, so it’s not necessary to « add » noise with them, to my humble opinion.

Don't worry too much about that. Flats have a much better signal/noise ratio than the lights they correct, they are definitively not a limiting factor ;). By the way, this is why 3 or 5 flats are quite enough for deep sky.

Christophe Perroud:
Only darks will help to minimise the amount of thermal noise generated by the sensor.

No, darks will never help reduce noise.

I can not agree that Flat field images taken with ISO 800 are not neccessary but I have the expierience that you are taken the main images wich have a another ISO such as ISO 1600 DSS stacking gives a message that the flat field images not correspondending with the other images with a another ISO and at the imaging proces they will not be used. Therefore I use Always the same ISO for ALL the images who are taken in the session. Perhaps you use no DSS stacking to create the image. But this point is useful to remember and thanks for your comment.

Thanks, everyone.

Any tips for taking Bias frames (with DSLR)? It's the same as taking darks but with a faster shutter speed, correct?

Also—can someone explain how to dither, please? I will be upgrading to a guided system in the near future (ASIAir Pro most likely), in case that is helpful to know.

Additionally, for my flats, I shot at ISO 100 (as per a suggestion above) while my lights and all else was at 3200. I wonder if I need to do more integration time at a lower ISO; maybe 1600 or 800?

Here's my integrated image (I know, I need a guide scope):

Here's a single light frame:

I suspect the biggest issues include ISO value choice, not clipping the dark pixels (important), and by far the most, post processing with masks, both in noise reduction (to slightly help the fainter parts of the galaxy) and in stretching.

Can you tell me a bit more about “clipping the dark pixels”?

Also, post processing with masks. Are you referring to within photoshop?

Can you tell me a bit more about “clipping the dark pixels”?Also, post processing with masks. Are you referring to within photoshop?

Clipping the darks means you've brought the left slider of your histogram too far to the right. This makes pixels that actually have some "brightness" become zero when they shouldn't, meaning you've "clipped" that detail. This is tough to avoid if you don't extract your background well, as it seems is necessary from that final image above.

Masks are incredibly complicated, challenging to master and detailed, but yup it's the same concept as photoshop masking (only astronomically suited- Pixinsight by far has the best tools for this). They take some time to get right but definitely can help with this.

Any tips for taking Bias frames (with DSLR)? It's the same as taking darks but with a faster shutter speed, correct?

Fastest shutter speed, ISO 100. But with DSLRs, the simplest and most efficient way to manage bias frames is to create a synthetic bias based upon your DSLR's offset level, see my message above. Then you have a bias that you can use forever without addind read noise.

Also—can someone explain how to dither, please? I will be upgrading to a guided system in the near future (ASIAir Pro most likely), in case that is helpful to know

Just shift the fov, but slightly (a few pixels amplitude) and randomly between frames. Ideally, do it between each exposures, but if you do it manually (hand controller) you can wait 5-10 frames between each shift not to go crazy.

I wonder if I need to do more integration time at a lower ISO; maybe 1600 or 800?

Looking at the "read noise versus ISO" chart for your Canon 60D, the read noise difference between ISO 800 and ISO 3200 is less than 1 electron. You may benefit from reducing a bit your ISO setting, maybe 1600 or 800 if your sky is quite light polluted. This way you would gain in dynamic range at a relatively small cost in terms of read noise. Tests have to be done depending on your shooting conditions.

But apart from these considerations, you would really benefit from working on your processing techniques to improve the rendering of your final image. It is no small task, over-processing is common when beginning. As AwesomeAstro mentionned it, be careful not to clip the darks for example. It results is the loss of signal, even if you think you are hiding background noise.

Thanks everyone—this is all very helpful.

I use APP currently. Does anyone have examples of what their image looks like at the point they finish in APP and bring it into Photoshop?

I will work on everything mentioned and keep improving.

I never use darks, flats or bias frames with my D3300. Although I experimented a lot in the beginning, I could never find any advantage. In all cases I could remove the gradients in my stacked image by applying Pixinsights "Dynamic Background Extraction". I take a lot of time to do it and place up to 300 samples manually, which sometimes takes about 2 hours. But the result is always pleasant. Below you find an example.

1.) Stretched stacking result of my Orion widefield image (50% crop). You see massive gradients.

2.) Same image with the same stretch after one iteration of "Dynamic Background Extraction" in Pixinsight.

3.) Final result.

Does anyone have a resource/instructions about how to create a synthetic bias? I shoot with the Canon 60D currently.

Also, any little tricks that may be worth looking into for APP processing would be appreciated!

Thanks again for all the help.

What I call a synthetic bias is simply a constant image filled with the offset level of your DSLR. It can be done with a lot of softwares.

If you want precise instructions, I can tell you what to do on Siril, a free software that I use for pre-processing my images (far better than DSS by the way). Take a master bias. Open it with Siril then right click -> statistics. Note the median value of the image (it is a power of 2). Now type the command « fill *value* » where *value* is the median level of your image (ie the offset level of your DSLR). Now you have a bias that will make its job (substracting the offset) without adding any king of noise (sigma = 0).

This procedure can be generalized to any image processing software able to mesure and modify pixels’ values.

Note that this is nothing extravagant. Substracting the offset level on a DSLR is something really easy and it is the first thing to be done by a software processing raw files : it reads the offset level in the EXIF data and substracts it from the image. The synthetic bias is just there to reproduce this offset level so that the pre-processing software can substract it from the image without adding any noise.

Now I think you can stop worrying about calibration frames. You have a bias reusable forever, try to dither and take about 3 flats/15 darks per session, and you are good to go.

Most importantly, keep on improving your processing skills and accumulate as many hours of data as you can. I can't help you with APP, but I'm sure many will.

Clear skies.