#1  
Old 25-06-2016, 01:08 PM
Shiraz's Avatar
Shiraz (Ray)
Registered User

Shiraz is offline
 
Join Date: Apr 2010
Location: ardrossan south australia
Posts: 4,918
ASI1600 performance/processing

Hi. thought it might be worth starting a thread on practical experience with ASI1600 imaging. To start off, the following is a summary of the "amp glow" issue.

Regards Ray

the ASI1600 produces some "amp glow" that is different from the CCD experience. Although it looks a bit worrying under a heavy stretch, it is of no practical significance in almost all imaging.

First off, if you trust images rather than numbers, the attached image shows a very heavily amplified part of a 300s sub taken through a Baader O3 filter using an f4 scope (the full dynamic range of the displayed linear image corresponds to only 100 electrons). The sky that night was dark with no moon and the O3 nebulosity should have been negligible.
The measured sky noise in the region of "amp glow" (on the right hand edge) was very similar to that in other parts of the scene (you can just tell where the region is from the slight fixed pattern noise, but you have to look hard).
The sub was not calibrated in any way, but the FPN and the "amp glow" could be calibrated out if required, without leaving any significant additional noise. If you use an oversampled scope under completely dark skies and with a 3nm narrowband filter, the "amp glow" may just possibly result in slight additional noise in the edges of an image - but in all other circumstances, the "amp glow" is not an issue at all.

If you want numbers - from my camera, in the centre of a 300s dark frame taken at -15c and gain 100, the excess signal above bias is ~7-15 ADU. At gain 100, the camera produces ~10 ADU per electron (1.6e/12bitADU and x16 to get to 16bit data), so the excess signal is 0.7-1.5 electrons, which corresponds to a dark current of 0.002-0.004 e/s/pix (very low).
Assuming shot noise, the total noise (read noise and dark noise) should have been ~2.5eRMS and it was measured to be ~2.3eRMS - the good agreement means that there is no fixed pattern noise (FPN) in the centre of the field.
In the brightest part of the "amp glow", the excess signal above bias is ~70 ADU or 7 electrons, which corresponds to a dark current of 0.02e/s/pix - still very low.
Again assuming shot noise, the (dark noise and read noise) should have been ~3.3eRMS, but I measured ~4eRMS - so there is a little bit of FPN in the "amp glow" region (it can be seen at extreme stretch).
ie, even in the worst glow region, the total camera noise from the 1600 is still way below the (dark noise and read noise) of most CCDs and in the central region it is almost non-existent.

The fact that the "amp glow" can be seen at all is an indication of just how low the noise in this chip really is - in any CCD, this level of glow would be submerged below the combined dark noise and read noise. So, you can dark-calibrate out the "amp glow" if you wish and it should not add any noticeable noise to your images. Alternatively, you could dither to break up the FPN and then ignore both the "amp glow" and dark calibration altogether and rely on non-linear rejection to remove any abnormal brighter pixels - you might find this to be as effective as full calibration for broadband imaging and possibly also for NB imaging under moonlit sky.
Attached Thumbnails
Click for full-size image (ampglow.jpg)
184.5 KB218 views

Last edited by Shiraz; 27-06-2016 at 08:15 AM.
Reply With Quote
  #2  
Old 25-06-2016, 02:31 PM
glend (Glen)
Registered User

glend is offline
 
Join Date: Jun 2013
Location: Lake Macquarie
Posts: 7,033
Thanks for starting this thread Ray.
Reply With Quote
  #3  
Old 25-06-2016, 02:34 PM
Atmos's Avatar
Atmos (Colin)
Ultimate Noob

Atmos is offline
 
Join Date: Aug 2011
Location: Melbourne
Posts: 6,980
Very informative post Ray. It definitely sounds like a very clean sensor in general purpose, as you say, amp glow is noticeably under heavy stretch but can easily be calibrated out. The overal noise that you're dealing with is very low that's for sure.
Reply With Quote
  #4  
Old 26-06-2016, 10:32 PM
Shiraz's Avatar
Shiraz (Ray)
Registered User

Shiraz is offline
 
Join Date: Apr 2010
Location: ardrossan south australia
Posts: 4,918
further to the earlier post on "amp glow", here is a composite image showing small segments of a 694 dark frame and a 1600 dark frame. Both were taken at -15, 300s and both have had bias subtracted. The 1600 was run at gain 100 and bias 21. The two images have been stretched so that the total image dynamic range corresponds to 20 electrons - ie we are comparing apples with apples and are looking right down in the noise.

http://astrob.in/full/253351/0/?real=&mod=None

Even though the 694 is a very low noise CCD, it displays a lot more read noise that the 1600 - as expected from the specs. One of the "amp glow" regions of the 1600 is shown - but even that is no more noisy than the average of the 694. The amp glow region incorporates some dark current offset, which can be calibrated out. I have done that on one of the darks and the resulting image is very low noise, with no obvious sign of too much excess noise from the glow - however, much of the read noise in the calibrated dark is clipped at 0 electrons (there is no dark current pedestal to keep it from going negative), so the calibrated dark is not visually comparable to the other two images - to avoid confusion have not posted it.

In summary, the 1600 has lower noise than the 694 over most of the chip. With dark calibration it has lower noise over the whole chip. Regards Ray
Reply With Quote
  #5  
Old 27-06-2016, 02:38 AM
glend (Glen)
Registered User

glend is offline
 
Join Date: Jun 2013
Location: Lake Macquarie
Posts: 7,033
Ray thanks for that comparison. The amp glow in the 1600 is, as i understand it, dependent on the driver and perhaps on the camera firmware revision being run - they may not be the same thing. Can you tell me which driver version applies to your dark test below?
Re Firmware, I believe that my 1600 is running the ASI1600CoolFpga_V1.104.vme firmware upgrade (dated 15/06/16), as Sam installed that during my Teamviewer session. I also have what appears to be the previous file ASI1600Cool_V1.102.vme (dated 9/3/16), in a release folder he left on my laptop. I do not know if this firmware makes any difference to amp glow.

For comparison purposes I can supply you with a 300 second Unity dark file (@-25C), which I have been using and which was shot after the upgrades applied by Sam. I also have a Master Dark built with the same dark files if that is preferred. Interestingly, the individual 300sec dark FIT files, viewed in FITS Liberator at pixel level, show eleven white pixel spots (hot pixels) scattered around the frame, which are removed during DSS processing to Master Dark (with hot pixel detection/correction turned on).
A review of my other individual FIT Dark files shot at 60, 90, and 120 seconds show significantly less noticable hot pixel presence. Unity seems to have an influence as well as compared to less Gainy settings. For LRGB imaging the hot pixels would not be a concern but for narrowband they do show up but are easily removed during Master Dark processing. Do other people have these in their individual Dark files?

When I stretch my 300s dark files they do seem darker on that right side boundary, which is where it had appeared previously, but obviously a histogram comparison might tell more than my old eyes and an undetermined stretch.
I am unequipped to evalutate the differences in the firmware releases but will ask Sam for some notes about what any differences might be. If you would like a copy of one of my dark files just PM me your email address and I will send it to you (31meg), or I can load it into the file sharing area of Astrobin..

Last edited by glend; 27-06-2016 at 03:58 AM.
Reply With Quote
  #6  
Old 27-06-2016, 07:05 AM
Shiraz's Avatar
Shiraz (Ray)
Registered User

Shiraz is offline
 
Join Date: Apr 2010
Location: ardrossan south australia
Posts: 4,918
Thanks Glen. I am using driver ....5. I have no idea what firmware, but whatever it is, it is working fine.

would be good to compare darks - Astobin shared would be best.

I have 3 of those stuck-on pixels - easily dealt with (eg a hot pixel map, dither + rejection). Also, at high gain, there is a small number of transitory hot pixels that appear on only a few frames in a run - looking at a sequence of darks at high stretch, these pixels turn on and off and are present on maybe 1/3 of the frames typically. They cannot be removed from lights with conventional dark processing, because their average in the master dark is less than the value in any of the lights where they are switched on. Hot pixel mapping is needed to manage these if you really want to do a full calibration. They only have low signal levels and I prefer to use dither plus average stacking with rejection to completely remove them. Come to think of it, they behave like regular cosmic ray hits, but they are associated with particular pixels (or pixel pairs occasionally).

Last edited by Shiraz; 27-06-2016 at 08:09 AM.
Reply With Quote
  #7  
Old 28-06-2016, 05:35 PM
Shiraz's Avatar
Shiraz (Ray)
Registered User

Shiraz is offline
 
Join Date: Apr 2010
Location: ardrossan south australia
Posts: 4,918
Flat calibration with 1.25 filters

Hi

used sky flats and bias to calibrate a single sky flat taken at a different time from the calibration flats.

without flat calibration, there was 30% drop from the centre to the darkest corner (right hand image below). After calibration, the test image was flat to better than 1%. The noise was (naturally) greater in the darker corners, but not excessive.

At f4, I can live with the 1.25 filters - but they work best if they are as close as possible to the chip.
Attached Thumbnails
Click for full-size image (flatterdemo.jpg)
70.7 KB85 views
Reply With Quote
  #8  
Old 19-07-2016, 11:57 PM
Shiraz's Avatar
Shiraz (Ray)
Registered User

Shiraz is offline
 
Join Date: Apr 2010
Location: ardrossan south australia
Posts: 4,918
some exposure strategies

The attached image shows a table of "optimum" broadband sub exposures for the ASI1600MM-cool for a range of gain settings. The optimum is that exposure where read noise adds only 5% to the total noise - ie read noise is effectively submerged under the sky noise. A few assumptions were made in these calculations (for example the QE was guessed at), so the results may not exactly match with reality. However, these suggestions should provide a useful starting point for deciding on sub exposures for broadband imaging.

To use the table, select the column that applies to the Fno of the system and the type of sky - 21.5 is very dark and 20 is moderately dark (the sky values have units of Vmag/arcsec2 or ~SQM units). The column values show the suggested sub exposure lengths (in seconds) for the gains shown in the left hand column.

The graph shows how the dynamic range (indicated by the maximum electron count in a pixel after stacking), varies with gain setting, assuming that the suggested optimum sub exposures are used. The DR values shown are specific to a chosen system, but the shape of the underlying curve will apply to any system - and the message is that if there is a need for as much dynamic range as possible, avoid high gain (ie anything above about 100). If high gain is used to enable short exposures, you can do so if the loss of dynamic range is acceptable.

For narrowband imaging, there are too many variables to produce a useful summary table. The best advice may be to use an intermediate gain as a starting point (eg maybe 70) and then choose the longest sub exposures that fit within the mount's capability and that keep star saturation at acceptable levels. Short subs may be possible on very bright nebulae, but in general, it might be a reasonable approach to start at 5 minute subs and increase from there as much as possible. If looking for very dim NB structures, minimum read noise will help and it may be useful to run at high gain (eg 200 or more). You will then need to choose subs that are long enough to retain the advantage of the low read noise, but short enough to stop too many stars from saturating.
Attached Thumbnails
Click for full-size image (asi1600.jpg)
77.3 KB170 views

Last edited by Shiraz; 20-07-2016 at 09:14 AM.
Reply With Quote
  #9  
Old 20-07-2016, 03:19 AM
glend (Glen)
Registered User

glend is offline
 
Join Date: Jun 2013
Location: Lake Macquarie
Posts: 7,033
Excellent work Ray, thank you. I will give it a try when the moon gets out of the way and if the clouds ever go. I will print it out and laminate it for the wall of the observatory workstation. I, for one, would be happy to pay for a little ASI1600 Android app that calculates sub lengths based on present conditions, i guess it would need to interface with the SQM app ( which i believe exists now) , or receive manual input of SQM reading from a handheld unit. Sort of an ASI 1600 exposure metre app.
I have been using 60" at Unity gain (139), but seems i need to try the mag20 f5 numbers now.

Last edited by glend; 20-07-2016 at 03:30 AM.
Reply With Quote
  #10  
Old 20-07-2016, 06:59 AM
Atmos's Avatar
Atmos (Colin)
Ultimate Noob

Atmos is offline
 
Join Date: Aug 2011
Location: Melbourne
Posts: 6,980
Very interesting Ray, I am assuming that the Gain 0 decrease in total count is caused by the 3.6e- read noise?
Reply With Quote
  #11  
Old 20-07-2016, 07:39 AM
Shiraz's Avatar
Shiraz (Ray)
Registered User

Shiraz is offline
 
Join Date: Apr 2010
Location: ardrossan south australia
Posts: 4,918
Quote:
Originally Posted by glend View Post
Excellent work Ray, thank you. I will give it a try when the moon gets out of the way and if the clouds ever go. I will print it out and laminate it for the wall of the observatory workstation. I, for one, would be happy to pay for a little ASI1600 Android app that calculates sub lengths based on present conditions, i guess it would need to interface with the SQM app ( which i believe exists now) , or receive manual input of SQM reading from a handheld unit. Sort of an ASI 1600 exposure metre app.
I have been using 60" at Unity gain (139), but seems i need to try the mag20 f5 numbers now.
Thanks Glen - hope it turns out to be useful. I have not the slightest inkling on how to write an app, but I used to do assembler programming, so I guess it might be do-able.

Quote:
Originally Posted by Atmos View Post
Very interesting Ray, I am assuming that the Gain 0 decrease in total count is caused by the 3.6e- read noise?
Hi Colin. I think that the kick up in read noise at zero gain causes the drop, but it is a good question to ask. Might also be some compounded reading error in getting data across from the ZWO graphs - will go back and check. thanks.

EDIT: checked with a little more care on reading the ASI data - the results look a bit more consistent, but are essentially the same, with the dynamic range dropping a little at gain zero because of the read noise profile. Have upgraded the original post with the "better" data and also rounded off the suggested exposures to better reflect the inherent inaccuracy of the underlying data. thanks again.

Last edited by Shiraz; 20-07-2016 at 09:20 AM.
Reply With Quote
  #12  
Old 26-07-2016, 11:42 AM
glend (Glen)
Registered User

glend is offline
 
Join Date: Jun 2013
Location: Lake Macquarie
Posts: 7,033
I had a chance last night to test image with Ray's chart as a guide. Having previously been imaging at Unity Gain settings I had become familiar with the camera's behaviour at that setting. I quickly noticed that it is going to take a lot more subs at say 35" and Gain 50 than subs captured at 60" at Gain 139 (Unity).

Not suggesting Ray's chart is incorrect, far from it, but user of the numbers should take into consideration the influence of Gain reduction, and greater sub numbers required for data build, as a trade off for the ultimately increased dynamic range.

Last edited by glend; 26-07-2016 at 12:13 PM.
Reply With Quote
  #13  
Old 26-07-2016, 12:30 PM
Atmos's Avatar
Atmos (Colin)
Ultimate Noob

Atmos is offline
 
Join Date: Aug 2011
Location: Melbourne
Posts: 6,980
Quote:
Originally Posted by glend View Post
I had a chance last night to test image with Ray's chart as a guide. Having previously been imaging at Unity Gain settings I had become familiar with the camera's behaviour at that setting. I quickly noticed that it is going to take a lot more subs at say 35" and Gain 50 than subs captured at 60" at Gain 139 (Unity).

Not suggesting Ray's chart is incorrect, far from it, but user of the numbers should take into consideration the influence of Gain reduction, and greater sub numbers required for data build, as a trade off for the ultimately increased dynamic range.
If you're capturing 60" at unity then you should be taking considerably longer subs at Gain 50 than 35", longer than 60" in fact.
Reply With Quote
  #14  
Old 26-07-2016, 01:46 PM
glend (Glen)
Registered User

glend is offline
 
Join Date: Jun 2013
Location: Lake Macquarie
Posts: 7,033
Quote:
Originally Posted by Atmos View Post
If you're capturing 60" at unity then you should be taking considerably longer subs at Gain 50 than 35", longer than 60" in fact.
On the surface Colin that would seem to be true, but i believe the 35" sub recommendation is based on the effect of skyglow and thus it is implied that you need many more short subs than longer subs to obtain the same amount of signal (assuming Gain was the same). When you reduce Gain as well it further compounds the need for more subs, at least that is my understanding of the chart.
Reply With Quote
  #15  
Old 26-07-2016, 02:35 PM
Atmos's Avatar
Atmos (Colin)
Ultimate Noob

Atmos is offline
 
Join Date: Aug 2011
Location: Melbourne
Posts: 6,980
Following the chart that Ray posted the 35" recommendation is for 250 gain and not 50.. Typo?
Reply With Quote
  #16  
Old 26-07-2016, 03:35 PM
glend (Glen)
Registered User

glend is offline
 
Join Date: Jun 2013
Location: Lake Macquarie
Posts: 7,033
Quote:
Originally Posted by Atmos View Post
Following the chart that Ray posted the 35" recommendation is for 250 gain and not 50.. Typo?
Colin, the f5 column for 20.5mag sky says 35" at Gain 50 on the chart Ray posted below.
Anyway i am reshooting tonight and will try a few different Gain settings between 50 and 139. It seems logical to assume that 120" might be a better choice at Gain 50. I might mess around with different sub lengths at Gain 50 and see what i can find out.
Reply With Quote
  #17  
Old 26-07-2016, 03:49 PM
Shiraz's Avatar
Shiraz (Ray)
Registered User

Shiraz is offline
 
Join Date: Apr 2010
Location: ardrossan south australia
Posts: 4,918
All that is needed is enough sky signal so that the associated shot noise overwhelms the read noise. The read noise of this camera is so low that you don't need much sky noise to do that.

if you use high gain, the read noise is lower and you can use shorter subs to overcome that lower read noise. High gain with appropriate short subs will give exactly the same SNR results as fewer, low gain, longer subs, provided the total exposure time is the same. It is counter-intuitive, but the best sub exposure will produce subs that look pretty crappy individually, but that is not the final goal - take no notice of what the subs look like - just the final stack. If you insist that you must have subs that look good individually, use longer subs and accept that you will lose some dynamic range.

If the sky is bright, the sky noise will overwhelm the read noise with shorter subs. It isn't that the short subs give you better results than longer ones - it's just that they do not make the bad situation (high sky noise) noticeably worse.

So in summary -
high gain > use more, shorter subs
bright sky > you can choose to use more, shorter subs

I have tried 30 second and 1 minute exposures with my f4 system under fairly dark sky - a final stack of 15x1 minute subs was slightly better than that of 30x30seconds, as suggested by the table, but the difference was not extreme. Glen, with your f5 system at gain 50 under mag20.5 sky, I would be using maybe 1 minute rather than the 35 seconds (which is recommended for a significantly brighter sky).

Since the signal is low in the subs, floating point stacking and processing is required - if the stack or process software has fixed point internal representation, there might be a loss of signal resolution leading to poor outcomes. I use Nebulosity and PixInsight for all processing - these are floating point. Anyone had experience with other stack or processing packages? Photoshop for example seems to be 16 bit fixed for most operations, but I really don't know much about it.

Last edited by Shiraz; 26-07-2016 at 05:23 PM.
Reply With Quote
  #18  
Old 26-07-2016, 04:03 PM
Camelopardalis's Avatar
Camelopardalis (Dunk)
Drifting from the pole

Camelopardalis is offline
 
Join Date: Feb 2013
Location: Brisbane
Posts: 5,425
Dropping the gain from 139 to 50 is just shy of 9dB, so you'd want to increase the exposure length by about 3x to get roughly the same pixel values.

Experimentation is key...my strategy is to expose long enough to get good signal in the object of interest, but not blow out the stars...although there's not always a balance point.
Reply With Quote
  #19  
Old 26-07-2016, 04:33 PM
Shiraz's Avatar
Shiraz (Ray)
Registered User

Shiraz is offline
 
Join Date: Apr 2010
Location: ardrossan south australia
Posts: 4,918
Quote:
Originally Posted by Camelopardalis View Post
Dropping the gain from 139 to 50 is just shy of 9dB, so you'd want to increase the exposure length by about 3x to get roughly the same pixel values.

Experimentation is key...my strategy is to expose long enough to get good signal in the object of interest, but not blow out the stars...although there's not always a balance point.
thanks Dunk. the complication is that the read noise also decreases as the gain increases, so you don't need the same pixel values at high gain to overcome the lower read noise.

Agree that experimentation is the best way to sort it out, but this camera is so flexible that there is no single "best" and the suggested optimum at high gains may produce subs that look really crappy - but that still produce as good a stack as that from much better looking longer subs taken at lower gain. ie, you cannot tell by looking at the subs how well you are going - it is only on stacking that the results become clear.

Last edited by Shiraz; 26-07-2016 at 05:22 PM.
Reply With Quote
  #20  
Old 26-07-2016, 05:39 PM
Camelopardalis's Avatar
Camelopardalis (Dunk)
Drifting from the pole

Camelopardalis is offline
 
Join Date: Feb 2013
Location: Brisbane
Posts: 5,425
Thanks Ray, it's good to get feedback that it's a versatile imaging machine

From ZWO's plot it looks like read noise is about 40% greater at gain=50 than unity gain. Surely that'd push the exposure time higher, all things being equal? Of course, that's my conventional thinking

Can't wait to get mine out to a dark site...
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 11:29 PM.

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