Jump to content

  •  

CNers have asked about a donation box for Cloudy Nights over the years, so here you go. Donation is not required by any means, so please enjoy your stay.

Photo

How to take flats for CCD?

This topic has been archived. This means that you cannot reply to this topic.
39 replies to this topic

#1 anismo

anismo

    Fly Me to the Moon

  • *****
  • topic starter
  • Posts: 6,089
  • Joined: 30 Jan 2014

Posted 08 August 2014 - 09:20 AM

For DSLR, it seems pretty simple..  Just with a white t-shirt, using AV mode does the job. But for CCD what is required? Any specific technique.



#2 terry59

terry59

    Voyager 1

  • *****
  • Posts: 10,974
  • Joined: 18 Jul 2011

Posted 08 August 2014 - 09:39 AM

I use the t-shirt and keep my ADU count somewhere around 25000-30000 (I use SGP).

 

Edit: I take mine during the day with the scope parked (pointing north). My camera has an electronic shutter....that makes a difference :)


Edited by terry59, 08 August 2014 - 09:55 AM.


#3 Goofi

Goofi

    Cosmos

  • *****
  • In Memoriam
  • Posts: 8,137
  • Joined: 03 May 2013

Posted 08 August 2014 - 09:39 AM

Anis, I use an EL panel and a saved sequence for the filters.

Each filter has it's own time that gets me to 20,000 ADU.

I usually take 7 flats of each filter once I'm setup - takes less than 5 minutes.

 

A white t-shirt didn't work for me with the Ha filter ... each sub had to be about 4 minutes!

If you don't change your setup, there's a guy who says to do t-shirt flats during the day with the scope pointing north.



#4 CharlesW

CharlesW

    Long time member

  • ***--
  • Posts: 3,659
  • Joined: 01 Nov 2012

Posted 08 August 2014 - 09:44 AM

You'll of course want to get your ADU flat count from SBIG. Don't wait too long in the morning because these cameras haven't got the shortest shutter speeds and you'll never get an exposure short enough because the sky is already too bright. 



#5 anismo

anismo

    Fly Me to the Moon

  • *****
  • topic starter
  • Posts: 6,089
  • Joined: 30 Jan 2014

Posted 08 August 2014 - 09:55 AM

Thanks guys. How do you determine the ADU required?  And once you determine the ADU, is it trial and error with different exposures?



#6 Madratter

Madratter

    Hubble

  • *****
  • Posts: 13,277
  • Joined: 14 Jan 2013

Posted 08 August 2014 - 10:34 AM

For ADU level you want it high to avoid noise but low enough to avoid the non-linear portion of your CCD. About 50% of your max ADU is a good target for many CCDs. SGP has a nice flats wizard for finding the proper exposures to achieve that.

 

I use a white glad bag over the front of my scope, held in place by my dew sheild. I sometimes shoot the roof of my observatory, and sometimes use the sky. Dawn and Dusk are not necessary (and in fact are a pain because of the rapidly varying light level. It is important to make sure the bag is shadowed from the light source. I want no direct light from my lights or the sun falling on the bag.

 

I now use the sky when shooting flats for narrowband to keep the exposure times down in the reasonable level (under 1/2 minute) where I can use bias frames.

 

It is VERY important to stack them properly. That method will depend on what you are using as your processing program. In PI it is easy. You just use the batch preprocessing script.



#7 anismo

anismo

    Fly Me to the Moon

  • *****
  • topic starter
  • Posts: 6,089
  • Joined: 30 Jan 2014

Posted 08 August 2014 - 10:42 AM

Thanks MR. As usual, this shows me how much I dont know. In fact, I didnt even realize there is a flats wizard. So far, I didnt have any dust issues and didnt take any flats. The last night runs, I have few dust bunnies that I cannot get rid off using DBE..

 

I have to try taking the flats tonight .. or may be I can just take my iPad and stick it on the scope and take the flats now (another use of working from home :) ). 

 

On a related note, do you guys take Bias?  If I just take flats, won't it have bias signal ? (or take flats and do bias subtraction and then use both flats and bias to calibrate?)  



#8 Madratter

Madratter

    Hubble

  • *****
  • Posts: 13,277
  • Joined: 14 Jan 2013

Posted 08 August 2014 - 11:56 AM

You do want to take either bias, or flat darks. If the flats are short enough, you can just use bias. You definitely want to use at least one or the other. If the scope is small enough, using something like an Ipad is a trick that people use.



#9 terry59

terry59

    Voyager 1

  • *****
  • Posts: 10,974
  • Joined: 18 Jul 2011

Posted 08 August 2014 - 01:13 PM

Thanks MR. As usual, this shows me how much I dont know. In fact, I didnt even realize there is a flats wizard. So far, I didnt have any dust issues and didnt take any flats. The last night runs, I have few dust bunnies that I cannot get rid off using DBE..

 

I have to try taking the flats tonight .. or may be I can just take my iPad and stick it on the scope and take the flats now (another use of working from home :) ). 

 

On a related note, do you guys take Bias?  If I just take flats, won't it have bias signal ? (or take flats and do bias subtraction and then use both flats and bias to calibrate?)  

 

 

You do want to take either bias, or flat darks. If the flats are short enough, you can just use bias. You definitely want to use at least one or the other. If the scope is small enough, using something like an Ipad is a trick that people use.

 

 

I'm interested in more details on this. I do not take bias or dark flats and don't think I'm missing anything (with the CCD). If I get enough clear sky time I want to get my D50 out and experiment though



#10 Goofi

Goofi

    Cosmos

  • *****
  • In Memoriam
  • Posts: 8,137
  • Joined: 03 May 2013

Posted 08 August 2014 - 01:34 PM

I took bias and dark's during the day - regulated cooling is great! - and made master dark and master bias. I just use those, with plans to update every 6 months.

Flats, each time out (especially since up until today I've taken down my gear each night and set things back up the next night.)

 

MR, great tip on using the white glad bag, thanks!



#11 anismo

anismo

    Fly Me to the Moon

  • *****
  • topic starter
  • Posts: 6,089
  • Joined: 30 Jan 2014

Posted 08 August 2014 - 02:20 PM

 isn't dark flats  same as bias for CCD cause you simply close the scope and shoot at the fastest possible speed? since you cant get any ADU to increase when the cover is on?

 

Also , do you keep the temp same for flats and bias as well? I still am not sure if I want to take darks . I might end up creating the darks just like Goofi.. since I just shoot at -10deg .. with dark scaling in Pi, I expect a single master dark should handle a range of exposures?


Edited by anismo, 08 August 2014 - 02:23 PM.


#12 jbalsam

jbalsam

    Viking 1

  • *****
  • Posts: 978
  • Joined: 06 Jul 2012

Posted 08 August 2014 - 02:22 PM

Terry -

 

You should subtract darks (or biases) from flat frames because of the math that is going on with flat calibration. The signal in your Image is = [Bias + (thermal + photon)*time]. The signal in your flat image is also = [Bias + (thermal + photon)*time]. You can usually approximately assume that the thermal signal in your flats is negligible because of their short length (unless you are taking 5 minute flats for your narrowband filters...). The signal in your dark calibration files is = [Bias + thermal*time]. You are dark subtracting your Image already, so you are getting rid of Bias and Thermal offsets. So your image is now = photon*time.

 

Briefly, Flats are applied by dividing all of the pixel values in your master flat by the highest pixel value, so the resulting image has a maximum value of 1, and all other values less than 1. You then divide your image by the normalized flat image. Any areas of the flat that were darker than the brightest region will cause those areas of your image to get brighter. Those values in the normalized flat frame, those values that are less than one, are a record of the relative **rate** at which those pixels build up charge. So, the only information that is allowed to be in your flat frame is information related to the **rate** at which charge is built up over time. The bias information is a constant offset. It is not time dependent, so if you are including it in your calibration, it will skew the numbers. You will probably be able to disregard dark-subtracting (or bias-subtracting) your flats and still be able to correct things like vignetting and dust motes, but you will be unable to remove pixel response non-uniformity, and this will be the noise that ultimately limits what you can get out of your images.

 

For a much better and more exhaustive explanation, I defer to Richard Crisp: http://www.optecinc.... powerpoint.pdf


Edited by jbalsam, 08 August 2014 - 02:52 PM.


#13 Raginar

Raginar

    Voyager 1

  • *****
  • Posts: 11,176
  • Joined: 19 Oct 2010

Posted 08 August 2014 - 02:45 PM

I do sky flats using CCDCC. They're automated and automatically adjust based on the ADUs they see.

Of note, I recently changed how long my flats are by shooting for a significantly higher value. I read about it in the CCD sub forum and it really has improved my flat fields. I wish I knew the thread now... The value is 42000 Adu. Two guys mentioned it and I queried them and sure enough it worked! No more flat fielding issues.

I dunno why. I'm sure someone here has an idea?

#14 jbalsam

jbalsam

    Viking 1

  • *****
  • Posts: 978
  • Joined: 06 Jul 2012

Posted 08 August 2014 - 02:52 PM

Chris - that answer is given in very technical terms in that presentation I linked to above from Richard Crisp. Briefly, one of the main things you are correcting for with flats is pixel response non-uniformity, and that is only the dominant form of noise in an image when you are very far away from the read noise (low range) and shot noise (lower-middle range) limited regimes. For PRNU to be the dominant noise, you need to be much closer to the full-well capacity of your pixels. That's why you see people using values like 30,000+ ADU. Nothing wrong with using ~15,000 ADU, but you'll need to average a lot more flats to take care of shot noise.

 

edit - typo fixed


Edited by jbalsam, 08 August 2014 - 07:21 PM.


#15 Madratter

Madratter

    Hubble

  • *****
  • Posts: 13,277
  • Joined: 14 Jan 2013

Posted 08 August 2014 - 03:43 PM

Terry, jbalsam gave the explanation I would have but at a much more detailed level.

 

I take my bias at same temperature, but it probably makes little to no difference. I won't say no difference since you can get some interesting behavior in CCD chips. At any rate, I do it because it is so easy to shoot off 100 bias frame and I KNOW that way that they are matched.



#16 terry59

terry59

    Voyager 1

  • *****
  • Posts: 10,974
  • Joined: 18 Jul 2011

Posted 08 August 2014 - 06:59 PM

Terry, jbalsam gave the explanation I would have but at a much more detailed level.

 

I take my bias at same temperature, but it probably makes little to no difference. I won't say no difference since you can get some interesting behavior in CCD chips. At any rate, I do it because it is so easy to shoot off 100 bias frame and I KNOW that way that they are matched.

 

 

Do you see them as more necessary with the 8300 chip than the ones from Sony?

 

When I was using the D50 I didn't use them and I didn't dither. Dithering is still an issue because SGP hasn't been able to get the D50 SDK. I plan to try again with bias and no darks but it has really bad amp glow. Just trying to sort out an approach



#17 Raginar

Raginar

    Voyager 1

  • *****
  • Posts: 11,176
  • Joined: 19 Oct 2010

Posted 08 August 2014 - 07:14 PM

jbalsam,  thanks for the link to that.

 

Terry,

 

Per the math, it doesn't matter what chip you're using, you do it to remove all the noise that is inherent to the system regardless.  As soon as you decide not to do one of the steps (bias, flats, darks), you're skipping a step in reduction and your results are sub-optimal.  As people have shown, sometimes your data isn't good enough (see D50...) that it matters.  Or, the chip is awesome (Sony...) and again, the effects are less.

 

The guys at Pixinsight say if you want to get the most out of your pictures, take good calibration data and don't skip a step.  We spend hours gathering the data, we spend hours processing it... what's lost in taking a few minutes to gather 50-100 bias frames?  

 

http://pixinsight.co...g48211#msg48211


Edited by Raginar, 08 August 2014 - 07:15 PM.


#18 Madratter

Madratter

    Hubble

  • *****
  • Posts: 13,277
  • Joined: 14 Jan 2013

Posted 08 August 2014 - 07:56 PM

I don't see it as more or less necessary with the 8300 or Sony chipped cameras. Since I take Bias anyway, it just makes sense to use them.


Edited by Madratter, 08 August 2014 - 08:09 PM.


#19 Alex McConahay

Alex McConahay

    Hubble

  • *****
  • Posts: 13,379
  • Joined: 11 Aug 2008

Posted 08 August 2014 - 09:20 PM

You know, the original poster asked, essentially, what is the difference in acquiring flats when one is using a CCD instead of a DSLR.

 

The discussion has morphed into how to get the best flats. But most of what has been said is how to get the best flats whether using DSLR or CCD. Many of the issues relate to getting good Flats, regardless of DSLR or CCD.  

 

I would answer the original question something along the lines of:

 

1. Both require the proper exposure. For a DSLR this is often achieved by using AV Mode. For CCD, "proper" means in the linear range of the chip. We usually measure this in ADU's. My QSI 583 likes to have an exposure averaging 28,000. You can check the  ADU in most image acquisition programs. You can find out from the camera manufacturer what they suggest.

 

2. A DSLR is a one-shot color camera (generally). If you have a OSC CCD, you need just one flat. In Mono CCD (or with a DSLR using filters), you need a flat for each filter you used. (By "one flat" in both cases, I mean one "set" of multiple flats averaged statistically to eliminate most noise).  

 

3. How you acquire the flat is essentially the same. The original poster mentioned a white t-shirt. That works equally well (or not) regardless of the camera. As some have pointed out, there are better ways to take flats, or at least different, but they are available to both DSLR and CCD users.  

 

Alex



#20 anismo

anismo

    Fly Me to the Moon

  • *****
  • topic starter
  • Posts: 6,089
  • Joined: 30 Jan 2014

Posted 08 August 2014 - 10:08 PM

Thank you Alex!  I defintely a learnt a lot (Running flats as I type!). I have used a iPad over t-shirt and used SGP's flat calibration wizard set to 30000 ADU.  Lets see how it will work out. 

 

 

I am more than happy for the discussion as it adds more  relevant information as well :) 



#21 akulapanam

akulapanam

    Mercury-Atlas

  • *****
  • Posts: 2,912
  • Joined: 26 Aug 2012

Posted 08 August 2014 - 10:22 PM

I use a 24" by 24" flat led panel that I have hung in my shed.  It was about $150 from LED City here in Dallas.  I put a towel in front of the LED panel to dim the light a little.  Works very well.



#22 anismo

anismo

    Fly Me to the Moon

  • *****
  • topic starter
  • Posts: 6,089
  • Joined: 30 Jan 2014

Posted 09 August 2014 - 09:37 AM

The flats I took seem to "sorta" ok.. I see the dust donuts in the right places.. but for some reason the dust donut is not a complete circle.. it is like clipped on on the side. I wonder if I have to bring the whole OTA (with the camera still attached) in and try using my monitors to re-shoot flats..  I will also shoot bias/darks this time around I guess.



#23 terry59

terry59

    Voyager 1

  • *****
  • Posts: 10,974
  • Joined: 18 Jul 2011

Posted 09 August 2014 - 10:07 AM

I don't see it as more or less necessary with the 8300 or Sony chipped cameras. Since I take Bias anyway, it just makes sense to use them.

 

 

jbalsam,  thanks for the link to that.

 

Terry,

 

Per the math, it doesn't matter what chip you're using, you do it to remove all the noise that is inherent to the system regardless.  As soon as you decide not to do one of the steps (bias, flats, darks), you're skipping a step in reduction and your results are sub-optimal.  As people have shown, sometimes your data isn't good enough (see D50...) that it matters.  Or, the chip is awesome (Sony...) and again, the effects are less.

 

The guys at Pixinsight say if you want to get the most out of your pictures, take good calibration data and don't skip a step.  We spend hours gathering the data, we spend hours processing it... what's lost in taking a few minutes to gather 50-100 bias frames?  

 

http://pixinsight.co...g48211#msg48211

 

Thanks...I'm sorry I twisted this thread.

 

Really though, I just am unconvinced that bias is that important in every situation. I look at the recent "discoveries" that darks aren't always the best approach....



#24 Madratter

Madratter

    Hubble

  • *****
  • Posts: 13,277
  • Joined: 14 Jan 2013

Posted 09 August 2014 - 10:24 AM

What darks are good for is removing hot pixels. If you have a camera with very few (a sony) or you have other ways of dealing with them (dithering with modern stacking algorithms) they may end up being unnecessary. There is nothing I am aware of that will do the job of bias other than bias. But at the end of the day, I'm all about experimental results. Do it both ways and then compare. Measure the signal to noise and see if one is better than the other.



#25 terry59

terry59

    Voyager 1

  • *****
  • Posts: 10,974
  • Joined: 18 Jul 2011

Posted 09 August 2014 - 11:28 AM

What darks are good for is removing hot pixels. If you have a camera with very few (a sony) or you have other ways of dealing with them (dithering with modern stacking algorithms) they may end up being unnecessary. There is nothing I am aware of that will do the job of bias other than bias. But at the end of the day, I'm all about experimental results. Do it both ways and then compare. Measure the signal to noise and see if one is better than the other.

What are your standard results from the NoiseEvaluation script?

 

Edit: My results on the IC1396A Ha image

 

* Channel #0
σR = 1.908e-002, N = 103398 (7.28%), J = 4

* Channel #1
σG = 1.908e-002, N = 103398 (7.28%), J = 4

* Channel #2
σB = 1.908e-002, N = 103398 (7.28%), J = 4


Edited by terry59, 09 August 2014 - 11:29 AM.



CNers have asked about a donation box for Cloudy Nights over the years, so here you go. Donation is not required by any means, so please enjoy your stay.


Recent Topics






Cloudy Nights LLC
Cloudy Nights Sponsor: Astronomics