Calibration in DSS..... Is it doing it wrong?....RESPONSE FROM LUC OF DSS

[jgs001]

That's how I'm doing it Lewis

[dph1nm]

that would be utterly pointless if DSS did that. It makes absolutely no sense to do so, and would increase the random noise contained within the subtracted frame......

Well I agree it is somewhat pointless, and a waste of processing power, but as far as I can see mathematically

Light - dark

is identical to

Light - (dark-bias) - bias

and does not add noise. I have always assumed this is what DSS does. My only concern might be if you use some fancy median combine for your master dark/bias frames, but even then I expect the added noise is trivial.

NigelM

[RikM]

Stacking images using DSS:

1) Load flats, bias frames, and a dummy light.

2) Stack

3) Clear the list

4) Load lights, darks and master flat (from 2)

5) Stack

6) Job done

This makes sense to me so I will do it this way in future.

I really like DSS and would not be able to combine frames without it as I can't afford fancy software. If this simple 1,2,3 list is the way to get the best from it, it should help plenty of DSS users. What a smashing thread:hello2:

[davew]

An updated response from Luc - Yahoo! Groups

Dave

[RobH]

I've been in contact with Luc, who wrote DSS, and he has kindly explained how DSS approaches the suubject of calibration.

Here are my questions, and any relevant replies.......

It seems to me that the importanty question here is whether using multiple bias adds noise ior not.

NBP says it does, Luc says not.

I don't know....I'm a snapper, not a scientist!!!

BTW.....Can we please try to keep any responses civilised pleased

Rob.

Hello.

You may know that there's a thread active on Stargazerslounge at the moment regarding calibration in DSS.

I started the thread as there have, on occasion, been questions regarding calibration that have come up, and often DSS has seemed to go against the standard methods of calibration in the way it applies bias frames.

From what is being said, DSS subtracts bias from lights, darks and flat darks..

Bias doesn't need to be subtracted from darks, or flat darks, as the bias information is already in them, and will be removed once the light is dart subtracted, or the flat is flat dark subtracted.

If bias is removed from everything, it means that there is multiple bias subtraction going on, which adds noise.

The understanding of the experienced imagers on SGL, who use Maxim, or CCDSoft, is that calibration is thus....

Lights should be either dark subtracted, or bias subtracted, but not both.

Flats should be bias subtracted, or flat dark subtracted, but not both, before being applied to the lights.

If you don't mind, can you let me know how the DSS algorithms deal with the calibration process, and do they, in fact, multiple subtract bias?

I'd like to clear up the confusion on this matter, and put an end to what is now turning into speculation on the thread, and which badly needs input from yourselves.

Thanks for writing a program that is used by so many with great success.

You have my greatest respect for doing this, especially for no financial gain.Now, let me explain how DSS is doing calibration and you will understand that all is good and soundly done.

Hello Rob.

First, let say that you have the full set of calibration files (lights, darks, flats, bias, dark flats).

In this case, DSS is subtracting the bias from every other frames. (this is the first workflow diagram on the web site)

So:

Dark’ = dark-bias

dark flat’ = dark flat-bias

flat’ = flat-dark flat’-bias (which is strictly identical to flat-dark flat)

light’ = (light-bias-dark’) (which is strictly identical to light-dark) / (flat’)

So the bias is not subtracted twice from the light because it has already been removed from the dark.

Indeed : light-bias-dark’ = light-bias-(dark-bias) = light - bias - dark + bias = light - dark.

As a side note, as at least one contributor has noted, subtracting the bias frame is *adding* the noise of the bias frame to the light.

This is one important reason to create a master bias as noise free as possible by using a lot of bias frames (they are cheap to come by).

Of course, there is also a bias signal to be removed and it would be a shame not to get rid of it.

Now, if you don’t have the full set, you have two choices (and only two choices).

If you are using flats (and not using them will create vignetted images) you still need to remove the bias signal from the flat frames.

You can do it by using bias frames or dark flat frames. Either way is working since the dark flat frames also contain the bias signal.

The second workflow diagram is for when you are using only dark flats, and the third workflow diagram is for when you are using only bias.

Both are perfectly sound

BTW, using dark flats is only making sense when the dark signal in your flat frames is strong enough to justify its subtraction.

Some CCD camera have a significant dark signal (and even hot pixels) even with very short exposures and/or they need a minimum exposure time.

In both cases the use of dark flats is mandatory.

PS: DSS is using the only calibration methods there is, the same as recommended by Iris and I still you like to see another mathematically sound working method.

Clear skies,

Luc

Thanks Luc.I'd figured out that you are effectively neutralising the use of double bias.

However, as you say, you do add noise, the more calibration steps you use.

What your method does is exactly the same as calibrating your lights with just darks, and bias (or flat dark) subtracted flats, but with more stages, where a small amount of noise is added at each stage by using bias subtraction more than once. (If noise is added at each stage....I make no claims to be a theoretician).

You don't need to bias subtract your lights when using darks, as the bias signal is there in the dark, and is subtracted when the light is calibrated by the master dark.

The only other bias component that needs to be removed is that contained in the flat, so this needs bias subtraction.

Consequently, would it not be easier to simplify the process, and cut out the extra bias stages?

On the other hand, I understand that in a lot of ways, DSS has been designed to be simple to use, and if you just throw everything into the mix, and do it the way you describe,the results are, to all intents and purposes, the same.

I agree completely that with all calibration frames, you need to cut down any possibility of adding noise, which is why I personally have bias masters made of 50 or more subs, and I use not less than 30 each of darks and flats.

For the reason you state, I almost never use flat darks, as they are of such short exposures that bias does the job just as well.

Cheers

Rob.

Hi Rob,

In fact subtracting the same master bias from the darks and the light frames is not adding any noise compared to just subtracting the master dark (with its bias signal and noise) from the lights.

On a theoretical point of view, the noise is added at each stage only if you are using *different* master frames in the stages.

Thus removing the same bias signal/noise is only adding the noise once.

Moreover, since you can easily achieve a noise free master bias by using a lot of bias frames (100+), the end result is that using a good master bias is often better than just using a good master dark unless you are using a equally large number of dark frames to create it (and it does come cheap ).

However, like I said before, if you insist on not using bias subtraction from light and dark frames, you can just tell DSS to view your bias frames as dark flats and only the flat frames will be bias subtracted.

In this case, for all intent and purpose, the dark flats subtraction will be the same as a bias subtraction.

Clear skies,

Luc

[Shibby]

Thanks very much Rob, I think I can conclude from what Luc is saying that I can just throw all my frames in and it will handle it; if I want the process to be a bit more streamlined, load the bias frames as dark flats. (regardless of noise, it seems pointless to do a load of unnecessary processing)

[Psychobilly]

Which is the way I have always used it....

[adamsp123]

Which is the way I have always used it....

Ditto

[namreg]

Ye gods! I wish I hadn't started reading this thread!

I think I'll just continue to chuck everything into DSS and let the program sort it out. It seems to work for me.

[RobH]

Ye gods! I wish I hadn't started reading this thread!

I think I'll just continue to chuck everything into DSS and let the program sort it out. It seems to work for me.

And I wish I hadn't started it :D:D

Think I'll stick with Maxim

[namreg]

Think I'll stick with Maxim

Having said I'll stick with DSS, I do have Maxim. Just out of interest, you couldn't point me in the direction of a good online tutorial on Maxim could you Rob? It's hardly intuitive!

Thanks

[simon hicks]

It would be interesting to see if you get any difference between a stack in Maxim and a stack in DSS. To draw any conclusions from this whatsoever, it would take a well designed, well thought out experiment with identical parameters and identical images and a well designed means of 'measuring' the result. And this would probably be a complex analysis requiring some quantitative analysis of different parts of the image (darkest regions, bright regions, hot pixels, other defects etc).

My hunch is that they would come out exactly the same and preferences between different packages are mainly due to the users experience and the choice of calibration/aligning/stacking parameters. In other words if you know how to use whichever package you use and you are using it correctly, then that's probably very close to optimum and there's nothing meaningful to be gained from trashing the other packages.

I'm really glad this thread finaly got round to asking the DSS author for clarification.

[roundycat]

I still have a foot in the camp that says subtracting the bias from the lights and the darks is pointless. The main reason for this is that a subtraction is just that - a subtraction. Noise, whether from bias or darks adds in quadrature. My gut feel is that if you subtract these components piecemeal you end up with a different result. One that is noisier. Just for the record, I measure noise in a 'flat' part of the picture using a 10 pixel radius cursor and reading the standard deviation.

If I can dig out some old lights and cal frames I might just sit down with someone that uses DSS and see what happens when we calibrate a group of lights to make a master using DSS and Maxim.

Dennis

[Dangerous-Dave]

OK....Two things now need to happen.

Firstly, any untoward bashing of DSS needs to stop.

For the record, I think that the guys who write DSS are to be applauded.

They have come up with a program that many use very successfully, at no financial benefit to themselves, and it must be a heck of a job writing a program like that.

Secondly, for the authors of DSS to be brought in on the discussion as we are now going round in circles, and they may well be able to provide answers that will finally clear up any confusion.

Without knowing how the DSS algorithms work, we are shooting in the dark.

Cheers

Rob

Who in your opinion is practicing "untoward bashing of DSS"? If that was in reply to my post, it was no worse than your comment at 1:52pm today.

The best moderator is an invisible one.

[Shibby]

I'd say this thread has probably come to a conclusion now.

[RobH]

Thanks to everyone who has contributed to this thread.

At last we seem to have an understanding of how DSS works, and the confusion that has been seen in the past should now be just that, a thing of the past.

It seems that the DSS method, and the method used by myself and others, end up at exactly the same point, the only difference being that one goes the long way around to get there

Again, thanks all.

Cheers

Rob

[Psychobilly]

Cheers Mr H for the Dialogue with Luc...

Billy...

[martin_h]

Sorted!!

[Psychobilly]

I somehow suspect this wont be the last of it

Cheers to Martin_h for his initial forays into the Yahoo_DSS group as well ...

[x6gas]

I must admit I am one of those people who throws lights, flats, darks and bias into DSS, and after selecting preferred options, lets it get on with it. I can see now that this may not be the right thing to do, and illustrates my lack of understanding of this process and some of the terms used. I think it would be really useful, and benefit a lot of users, if someone could compile a simple summary step by step process list specifically for DSS users based on the discussion here. Maybe it could become a sticky for 'preferred' DSS processing.

As a noob, I'd find that really useful...

[philming]

Hi all sorry for my silly question, but I dont get all this topic.

I mean, DSS is a software where you clic "open picture files", give it the batch of subs, then clic"dark frames", and give it your darks and so on (I usually never do any flats for now, just subs and darks). The hit compute offsets, then stack, and here you go...

What am I missing ?