#41  
Old 10-08-2010, 10:47 PM
jase (Jason)
Registered User

jase is offline
 
Join Date: Sep 2006
Location: Melbourne, Victoria
Posts: 3,916
Quote:
Originally Posted by pvelez View Post
I'm now playing with shorter and longer exposures to get the ADU count away from the 20,000 threshold I've been using so far.
Pete,

Where did you get a target ADU of 20,000 from? The full well depth of the SBIG 8300 is only 25,500! This may explain the over correction of the light frame data as I previously noted.

Go on the 1/3rds principle (~30%);
25,500 / 30% = 7650 target ADU, so round it up to 8000 and accept a difference of +/- 1000 ADU. When you combine a series of flat subs, data rejection with clip the highs and lows (extremes) and level the data (assuming you don't use sum!!). Average or median combines will work well, if you want to get fancy, CCDStack's clip min/max or poison sigma reject are great. If you're finding that its not correcting the frames well enough, increase the principle to 35%. Just set the target in SkyFlat assistant and let it work out the optimal exposure time. Too easy.

Quote:
Originally Posted by pvelez View Post
If I am binnng at 2x, do I need to target the ADU for a flat at half what I target for 1x binning? So if my target is 20,000 ADU, do I reduce that to 10,000 if I am binning 2x?
No, target ADU remains the same, its just that the exposure time to reach it will be shorter.

Cheers
Reply With Quote
  #42  
Old 11-08-2010, 08:07 AM
pvelez's Avatar
pvelez (Pete)
Registered User

pvelez is offline
 
Join Date: Aug 2006
Location: Sydney
Posts: 1,250
Quote:
Originally Posted by jase View Post
Pete,

Where did you get a target ADU of 20,000 from? The full well depth of the SBIG 8300 is only 25,500! This may explain the over correction of the light frame data as I previously noted.
Jase

I think there is a difference in nomenclature here. The full well capacity of the ST 8300 chip is about 25,500 e- which I understand converts to about 65,000 ADU. The Sky Flat Assistant sets by reference to ADU and I had thought that Maxim's graph function did the same. Feel free to correct me if I'm wrong on this.

Last night I tested flats with target ADU counts from 8,000 (co-incidentally) up to 30,000. It was late when I twigged to the error factor and dropped from 10% down to 2% and then 1%. Flats at 8,000 ADU yielded a very nasty vertical dark stripe down the image. The sweet spot (though still not that sweet) was at around 19,000 to 20,000 with a 1% error setting.

Its still not perfect - I am using 5 minute subs. I suspect that the uneven light will be less an issue at shorter subs. If so, applying the same flats will be worse rather than better I guess. Thats logical so its probably wrong.

Quote:
Originally Posted by jase View Post
When you combine a series of flat subs, data rejection with clip the highs and lows (extremes) and level the data (assuming you don't use sum!!). Average or median combines will work well, if you want to get fancy, CCDStack's clip min/max or poison sigma reject are great.
This bit is more of a challenge for me. I've set Calibration Wizard to the correct file and its identified in Set Calibration the relevant dark, bias and flat files. I've then created a master flat from the flats in that folder using median combine. The settings are as per your earlier post (albeit under a tab marked Advanced). I then open the light and calibrate it from the Process tab. I'm not sure how to level the data as you suggest. Is this an alternative to median combine?

Quote:
Originally Posted by jase View Post
No, target ADU remains the same, its just that the exposure time to reach it will be shorter.
Understood - thanks

Pete
Reply With Quote
  #43  
Old 11-08-2010, 10:10 AM
jase (Jason)
Registered User

jase is offline
 
Join Date: Sep 2006
Location: Melbourne, Victoria
Posts: 3,916
Quote:
Originally Posted by pvelez View Post
I think there is a difference in nomenclature here. The full well capacity of the ST 8300 chip is about 25,500 e- which I understand converts to about 65,000 ADU. The Sky Flat Assistant sets by reference to ADU and I had thought that Maxim's graph function did the same. Feel free to correct me if I'm wrong on this.
For 16bit cameras the maximum pixel value is indeed 65,535...however you're not calculating the flat saturation value from this figure!!! In doing so it would mean that every 16bit camera would have the same saturation value regardless of the CCD chips characteristics = incorrect flats!!! You need to calculate the value from the CCD chip's full well depth - 25,500 for the ST8300. Forget about the 16bit ADU for the time being.

Quote:
Originally Posted by pvelez View Post
Last night I tested flats with target ADU counts from 8,000 (co-incidentally) up to 30,000. It was late when I twigged to the error factor and dropped from 10% down to 2% and then 1%. Flats at 8,000 ADU yielded a very nasty vertical dark stripe down the image. The sweet spot (though still not that sweet) was at around 19,000 to 20,000 with a 1% error setting.
The vertical stripe you note is probably characteristics of the camera. Short exposures can exhibit this. Do your bias frames show a similar feature in the image? It is normally corrected during the bias/dark frame application to the flat subs prior to being combined.

Quote:
Originally Posted by pvelez View Post
Its still not perfect - I am using 5 minute subs. I suspect that the uneven light will be less an issue at shorter subs. If so, applying the same flats will be worse rather than better I guess. Thats logical so its probably wrong.
5min flat subs! Crazy. Long subs will work if your flats are controlled, i.e. you're using a lightbox or similar, but reaching the target ADU faster (shorter subs) is preferred especially if you're doing real sky flats as the longer the exposures the more changes you'll pick up stars in the dawn/dust sky and your time window to acquire the flats are short given the sky luminosity changes quickly at dawn and dusk. As a reference, the longest I'll go for sky flats is 120secs. I do reach this value when acquiring flats through the 3nm bandpass SII and OIII filters as the quantity of energy being passed is small compared to LRGB broadband filters. Long or short, you need to ensure you're scaling your dark frames to match the exposure times. Flats work best when the exposure time is altered to reach the target ADU - unless you have a way of controlling the luminosity of the FOV. Subsequently, I don't take flats shorter than 10 seconds. This is due to the Apogee camera's Melles Griot shutter leafs causing a star fish shape at the edge of the flat field. SBIG camera's use a different shutter mechanism, hence you wont have problems going shorter than 10 seconds for your flats.

Quote:
Originally Posted by pvelez View Post
This bit is more of a challenge for me. I've set Calibration Wizard to the correct file and its identified in Set Calibration the relevant dark, bias and flat files. I've then created a master flat from the flats in that folder using median combine. The settings are as per your earlier post (albeit under a tab marked Advanced). I then open the light and calibrate it from the Process tab. I'm not sure how to level the data as you suggest. Is this an alternative to median combine?
Yes, we're using different versions of MaximDL. I'm running 4.6, not version 5 hence the screen capture differences however the fundamentals are the same. You don't level the data manually, the algorithm used for data rejection is doing this for you. Its is determining which pixels should be considered outlier and rejecting them. Median combine is more than ample to produce calibration frame masters. Try getting around 8 flat subs per filter, but 3 will work. You want to ensure that you're not adding noise to the light frame subs as a consequence of bad flats.

Hope that clarifies.
Reply With Quote
  #44  
Old 11-08-2010, 10:45 AM
pvelez's Avatar
pvelez (Pete)
Registered User

pvelez is offline
 
Join Date: Aug 2006
Location: Sydney
Posts: 1,250
Thanks again for your time Jase

Quote:
Originally Posted by jase View Post
For 16bit cameras the maximum pixel value is indeed 65,535...however you're not calculating the flat saturation value from this figure!!! In doing so it would mean that every 16bit camera would have the same saturation value regardless of the CCD chips characteristics = incorrect flats!!! You need to calculate the value from the CCD chip's full well depth - 25,500 for the ST8300. Forget about the 16bit ADU for the time being.
OK, so I aim for about 8,500 as the target ADU in Sky Flat Assistant. I'll give that a go again.

Quote:
Originally Posted by jase View Post
The vertical stripe you note is probably characteristics of the camera. Short exposures can exhibit this. Do your bias frames show a similar feature in the image? It is normally corrected during the bias/dark frame application to the flat subs prior to being combined.
Until recently I didn't take bias frames. Now I do so that Maxim can autoscale the darks before subtracting them from the flats. To be honest I haven't looked too closely at them. I'll check this out too.


Quote:
Originally Posted by jase View Post
5min flat subs! Crazy.
It would be! I was only referring to 5 minute lights. My thinking was that the uneveness of illumination on the sensor becomes more pronounced the longer my light subs go. If my flats are too bright - the contrast with a shorter light sub will be more pronounced than if I applied the same flat to a long sub.


Quote:
Originally Posted by jase View Post
Long subs will work if your flats are controlled, i.e. you're using a lightbox or similar, but reaching the target ADU faster (shorter subs) is preferred especially if you're doing real sky flats as the longer the exposures the more changes you'll pick up stars in the dawn/dust sky and your time window to acquire the flats are short given the sky luminosity changes quickly at dawn and dusk. As a reference, the longest I'll go for sky flats is 120secs. I do reach this value when acquiring flats through the 3nm bandpass SII and OIII filters as the quantity of energy being passed is small compared to LRGB broadband filters. Long or short, you need to ensure you're scaling your dark frames to match the exposure times. Flats work best when the exposure time is altered to reach the target ADU - unless you have a way of controlling the luminosity of the FOV. Subsequently, I don't take flats shorter than 10 seconds. This is due to the Apogee camera's Melles Griot shutter leafs causing a star fish shape at the edge of the flat field. SBIG camera's use a different shutter mechanism, hence you wont have problems going shorter than 10 seconds for your flats.
Now this may be an interesting avenue too. I've been using a light box - luckily it has a potentiometer so I can step down the intensity. Using a target ADU count of 20,000, my flat subs are less than a second - in fact for the green filter its about 0.18 seconds - quite close to the spec minimum of 0.1s seconds. I can turn the intensity down a bit more and with the reduced ADU count I can probably increase the exposure time a bit. I wonder if the flat is affected by the shutter sweeping across the sensor so that I don't have even illumination across the field.

So I'll turn down the pot and take the subs as long as possible - the 8,000 target will make this easier.

Quote:
Originally Posted by jase View Post
Yes, we're using different versions of MaximDL. I'm running 4.6, not version 5 hence the screen capture differences however the fundamentals are the same. You don't level the data manually, the algorithm used for data rejection is doing this for you. Its is determining which pixels should be considered outlier and rejecting them. Median combine is more than ample to produce calibration frame masters. Try getting around 8 flat subs per filter, but 3 will work. You want to ensure that you're not adding noise to the light frame subs as a consequence of bad flats.
OK - got that (I think). I've been taking 10 flats and median combining them. I've relied on Maxim autoscaling the darks and subtracting them so I might try doing that manually.

Looks like I need to get the kids to bed early tonight so I have a chance to sort this out - all part of the learning curve.

Thanks again

Pete
Reply With Quote
  #45  
Old 16-08-2010, 07:52 AM
pvelez's Avatar
pvelez (Pete)
Registered User

pvelez is offline
 
Join Date: Aug 2006
Location: Sydney
Posts: 1,250
Sorted - I think!

I think I have this all sorted now. I spent a fair amount of time over the last few days playing with flats. Here is a summary of the results:

1. Error 1 - I found a few times that the FW moved to the wrong filter – usually to the Ha filter which is in the first position in the wheel rather than to the luminence filter as instructed. This doubtless happened enough to throw the results.

2. Error 2 - I was testing flats against lights taken a few days earlier. I couldn’t work out why all the dust donuts obvious on the flats were not taken out of the lights. The answer was that the lights didn’t have the dust on them at all. I suspect the dust accumulated on the filters between lights and flats. The most persistent of the donuts was in a dark section of the light. Had it been brighter in that section of the image, I might have realised that there was no dust there in the first place.

3. Error 3 – I was testing my flats against a single exposure taken through a green filter. I didn’t realise that the calibrated flat was displayed with a different stretch than the uncalibrated image. So in flicking between the calibrated and uncalibrated image using the undo button, I was actually seeing not only the effect of calibration, there was also a different stretch which gave greater prominence to the fainter sections of the image.

4. Error 4 – I don’t think I was properly applying darks to the flats. I now set up master frames so they are dark subtracted.

I suspect it was a combination of all of these things – as well as pilot error more generally – that had me in a tangle.

A couple of other points:

5. Shutter artefacts don’t seem to be an issue. Generally, the shape of the flat doesn’t change from a sub-1 second exposure to a 30 second exposure. I played with twilight flats over the weekend and so managed flats out to 30+ seconds. The pattern of illumination was the same.

6. I couldn’t work out why I had a strange colour gradient. The sky background at the left of the image always had a blue hue while to the right it was quite green. Looking at the blue filtered images, it was clear that there was a lot more blue in the background on the left side. However, when I upped the target ADU for the blue filter from 20,000 to 28,000 the difference was much less stark. To be honest I can’t work out why this might be the case – other than perhaps that the limited sensitivity of the ST8300 to blue light (and the cut-off of the blue filter) means that the response of the sensor is not linear at that wavelength (though that sounds v odd) or that Sky Flat Assistant responds too quickly to the peak intensity so it cuts short the exposure before there are enough photons registered by the sensor to give a truly flat field. Anyway, the result is that I take blues at 28,000 ADU and R, G and L at 20,000. I have yet to work out Ha – my Ha is not parfocal with the other filters and the cable for my electric focuser has yet to arrive so I’ll sort that out later.

I could be wrong about all this – however I managed to take about 1.5 hours of the Triffid last night and it looks great – without colour gradient or obvious uneven illumination. Maybe its not a drama as the target is quite bright. I’ll only know when I head back to galaxies.

Thanks to everyone who contributed to solving this. I suspect almost all of the errors were spotted by other contributors to this thread before I did. Goes to show there are some v smart (and experienced) imagers out there.

Cheers

Pete
Reply With Quote
  #46  
Old 16-08-2010, 04:51 PM
Exfso's Avatar
Exfso (Peter)
Registered User

Exfso is offline
 
Join Date: Sep 2006
Location: Adelaide
Posts: 1,699
I am glad we put the pot on the box pete.
Reply With Quote
  #47  
Old 16-08-2010, 08:56 PM
pvelez's Avatar
pvelez (Pete)
Registered User

pvelez is offline
 
Join Date: Aug 2006
Location: Sydney
Posts: 1,250
You betcha

Pete
Reply With Quote
Reply

Bookmarks

Thread Tools
Rate This Thread
Rate This Thread:

Posting Rules
You may not post new threads
You may not post replies
You may not post attachments
You may not edit your posts

BB code is On
Smilies are On
[IMG] code is On
HTML code is Off

Forum Jump


All times are GMT +10. The time is now 06:58 PM.

Powered by vBulletin Version 3.8.7 | Copyright ©2000 - 2024, Jelsoft Enterprises Ltd.
Advertisement
Bintel
Advertisement
Testar
Advertisement