#1  
Old 01-08-2012, 07:54 AM
troypiggo's Avatar
troypiggo (Troy)
Bust Duster

troypiggo is offline
 
Join Date: Nov 2008
Location: Brisbane, Australia
Posts: 4,846
8300 sensor RGB binning questions

This came up in the QSI Yahoo! Group and thought I'd also post here.

Does anyone here shoot L unbinned and RGB binned with 8300 sensors? I'm trying to find out if it's a sensor thing, normal for RGB binning, or something I'm doing in terms of processing or perhaps even image capture times or number of subs.

Quote:
I'm keen to hear other experiences as well. I have a 583ws, and have found that the binned RGB is quite "blocky". I know the pixels are double the size and would expect them to be blocky in that respect, but it's more than that.

See these cropped images. The L is unbinned, the RGB is 2x2.

L: http://flic.kr/p/cHivQ5
RGB: http://flic.kr/p/cHivts

See how the binned RGB stars have a dark pixel perimeter, then lighten up again? When I do the L and RGB combination, this is evident. ie the smooth L doesn't smooth that out. The little brighter diffraction spike looking pixels are there and noticable.

As a result, I've taken to shooting 1x1 for RGB as well, but would love to be able to cut exposure times down by using binned.

I'm not sure if it's my processing, a property of the sensor (I'm sure I've read somewhere that this is common on the 8300 sensors, but haven't been able to research further), normal for combining 1x1 L and 2x2 RGB, or something else.

Perhaps it's my image capture technique? As I understand it, binned 2x2 has 4 times the gathering capability, so 1/4 of the exposure time is correct?
Reply With Quote
  #2  
Old 01-08-2012, 08:50 AM
RickS's Avatar
RickS (Rick)
PI cult recruiter

RickS is offline
 
Join Date: Apr 2010
Location: Brisbane
Posts: 10,584
I've only ever used 1x1 binning with my SX H-18 after seeing horizontal blooming at 2x2. This is caused by the shallow horizontal shift register on the KAF8300 and can be reduced somewhat by dropping gain (some cameras appear to do this automatically when binning is used). There are some examples here: http://www.pbase.com/wjshaheen/horiz...he_kaf8300_ccd

Looks like a different issue to what you're talking about, Troy, but you may run into this as well. I don't think the KAF8300 design is well suited to binning.

Cheers,
Rick.
Reply With Quote
  #3  
Old 01-08-2012, 08:56 AM
troypiggo's Avatar
troypiggo (Troy)
Bust Duster

troypiggo is offline
 
Join Date: Nov 2008
Location: Brisbane, Australia
Posts: 4,846
Aah, yes. I've seen that effect on really bright stars if binning and exposures too long. That is different to what I'm experiencing, though.

The QSI recommendation is to use high gain for 1x1 and low gain for binned. Even then I've seen that effect you mention on really bright stars.
Reply With Quote
  #4  
Old 01-08-2012, 01:34 PM
cventer's Avatar
cventer
Registered User

cventer is offline
 
Join Date: Jul 2005
Location: Melbourne Australia
Posts: 957
Interesting. I binned my stf8300 2x2 with 5 min subs recently and had this horizontal blooming. It was realy annoying. I don't think you can lower gain on sbig drivers though.

Any other suggestions other than shorter subs?
Reply With Quote
  #5  
Old 01-08-2012, 09:34 PM
gregbradley's Avatar
gregbradley
Registered User

gregbradley is offline
 
Join Date: Feb 2006
Location: Sydney
Posts: 17,877
Never seen that with my FLIML8300 so it must be controllable.

Greg.
Reply With Quote
  #6  
Old 01-08-2012, 10:11 PM
Paul Haese's Avatar
Paul Haese
Registered User

Paul Haese is offline
 
Join Date: Jan 2009
Location: Adelaide
Posts: 9,937
Never seen that and have only recently (last two or so images) started doing 1x1 on all my channels. All other stuff has been 2x2 for colour and 1x1 for detail layer.
Reply With Quote
  #7  
Old 01-08-2012, 10:49 PM
allan gould's Avatar
allan gould
Registered User

allan gould is offline
 
Join Date: May 2005
Location: Brisbane
Posts: 4,485
Not seen the effect with my QSI583
Reply With Quote
  #8  
Old 02-08-2012, 12:16 AM
DavidTrap's Avatar
DavidTrap (David)
Really just a beginner

DavidTrap is offline
 
Join Date: Jun 2009
Location: Brisbane
Posts: 3,032
I've not seen it with my QSI583.

Paul - why did you change to 1x1?

DT
Reply With Quote
  #9  
Old 02-08-2012, 01:13 AM
troypiggo's Avatar
troypiggo (Troy)
Bust Duster

troypiggo is offline
 
Join Date: Nov 2008
Location: Brisbane, Australia
Posts: 4,846
Ok, thanks guys. As usual, must be me.
Might do some more experimenting with the binning.
Reply With Quote
  #10  
Old 02-08-2012, 08:19 AM
gregbradley's Avatar
gregbradley
Registered User

gregbradley is offline
 
Join Date: Feb 2006
Location: Sydney
Posts: 17,877
Quote:
Originally Posted by DavidTrap View Post
I've not seen it with my QSI583.

Paul - why did you change to 1x1?

DT
I know you asked Paul but if I can butt in. There is a regular debate about 1x1 on all channels veruss 1x1 on luminance only and 2x2 on colour.

The usual reasons for 2x2 on colour is that colour does not have sharp edges and more fading edges whereas luminance has sharp edges and requires extra resolution.

So 2x2 speeds up the acquisition process and also if you have poor seeing you are not losing anything anyway.

But for maximum resolution and in areas of good seeing and on objects where you want maximum detail (galaxies for example or globs or structures in nebs) 1x1 gives that extra. Its more suited to someone with a permanent observatory as the extra time required to get noise free channels is not so much of an issue.

Additionally some desaturate the RGB and add it to the luminance for a slight bit extra omph. I have not noticed that that technique adds much but every little bit counts.

1x1 is a moot point if you have poor seeing and you may as well shoot everything 2x2 especially a small pixelled camera like the 8300.

Not all cameras get the same gain on 2x2 but there always seems to be a gain. It isn't 4X though more like 2.0X. Richard Crisp did a paper about this recently measuring an 8300 chip for 2x2 performance. As I recall it ended up about 2X better signal to noise ratio using 2x2.

Greg.
Reply With Quote
  #11  
Old 02-08-2012, 08:23 AM
Paul Haese's Avatar
Paul Haese
Registered User

Paul Haese is offline
 
Join Date: Jan 2009
Location: Adelaide
Posts: 9,937
Quote:
Originally Posted by DavidTrap View Post
I've not seen it with my QSI583.

Paul - why did you change to 1x1?

DT

David I simply changed to 1x1 to accommodate narrow band blending into my imaging. I have for a long time wanted to undertake comprehensive imaging projects. Using 1x1 just makes it simple.
Reply With Quote
  #12  
Old 02-08-2012, 08:48 AM
troypiggo's Avatar
troypiggo (Troy)
Bust Duster

troypiggo is offline
 
Join Date: Nov 2008
Location: Brisbane, Australia
Posts: 4,846
Quote:
Originally Posted by gregbradley View Post
...
Not all cameras get the same gain on 2x2 but there always seems to be a gain. It isn't 4X though more like 2.0X. Richard Crisp did a paper about this recently measuring an 8300 chip for 2x2 performance. As I recall it ended up about 2X better signal to noise ratio using 2x2.
....
I found something similar, although nothing actually measured. Instead of 2x2 exposure times being 1/4 those of the 1x1 lum, I was going with 1/2 the exposure time. eg L at 1x1 was 4mins, I'd use 2x2 RGB at 2 mins instead of 1min.

Is that what you're saying?
Reply With Quote
  #13  
Old 02-08-2012, 01:52 PM
RickS's Avatar
RickS (Rick)
PI cult recruiter

RickS is offline
 
Join Date: Apr 2010
Location: Brisbane
Posts: 10,584
Richard's analysis is here:
http://tech.groups.yahoo.com/group/F.../message/11264

The benefit of binning on the sensor is that you only get the same read noise for a binned super-pixel as you would get for a single pixel in 1x1 mode. So, for example, with 2x2 binning you are getting 4 times better SNR just considering read noise.

Richard's point is that you also need to consider shot noise and the improvement in that increases as the square root of the number of binned pixels - with 2x2 binning you only get a 2 times better SNR just considering shot noise.

In practice, the benefit is going depend on which of these factors is dominant. For very faint signals the benefit will be closer to 4 times and will decrease to approximately 2 times as the signal increases.

Richard also mentions the problem of saturation of the horizontal shift register which is what causes horizontal blooming on the KAF8300.

Cheers,
Rick.
Reply With Quote
  #14  
Old 02-08-2012, 02:02 PM
troypiggo's Avatar
troypiggo (Troy)
Bust Duster

troypiggo is offline
 
Join Date: Nov 2008
Location: Brisbane, Australia
Posts: 4,846
While we have full moon at the moment I'm doing some experimenting.
Reply With Quote
  #15  
Old 02-08-2012, 02:10 PM
DavidTrap's Avatar
DavidTrap (David)
Really just a beginner

DavidTrap is offline
 
Join Date: Jun 2009
Location: Brisbane
Posts: 3,032
Quote:
Originally Posted by Paul Haese View Post
David I simply changed to 1x1 to accommodate narrow band blending into my imaging. I have for a long time wanted to undertake comprehensive imaging projects. Using 1x1 just makes it simple.
Ta for that

So how long at your typical RGB subs at 2x2 vs 1x1?

DT
Reply With Quote
  #16  
Old 02-08-2012, 02:20 PM
Paul Haese's Avatar
Paul Haese
Registered User

Paul Haese is offline
 
Join Date: Jan 2009
Location: Adelaide
Posts: 9,937
I was using 15 minutes at 2x2 and for some objects 10 minutes. At present I have been doing 10 minutes at 1x1 but that is likely to change at some point.
Reply With Quote
  #17  
Old 05-08-2012, 09:51 PM
troypiggo's Avatar
troypiggo (Troy)
Bust Duster

troypiggo is offline
 
Join Date: Nov 2008
Location: Brisbane, Australia
Posts: 4,846
Today I did some experimenting and got to the bottom of the issues I was referring to above. Turned out to be the algorithm I was using in PixInsight to do the upscaling from 2x2 to match the 1x1 size.

If you're interested, if you use the BatchPreProcessing script, for the image registration section they use StarAlignment with the default algorithm of "auto" which is actually, I believe, is one of the Lanczos algorithms. Changed that to some others and tested, and the best result I found was Bicubic B-spline.

The above would also apply if you were using StarAlignment, not just the BPP script.
Reply With Quote
  #18  
Old 05-08-2012, 10:06 PM
RobF's Avatar
RobF (Rob)
Mostly harmless...

RobF is offline
 
Join Date: Jul 2008
Location: Brisbane, Australia
Posts: 5,716
2x2 ok on my QHY9. In a perfect world I wouldn't resort to binning any of the RGB, but in reality its often tempting especially if seeing is questionable or shooting narrowband (S in particular).
Reply With Quote
  #19  
Old 10-08-2012, 01:23 AM
Tandum's Avatar
Tandum (Robin)
Registered User

Tandum is offline
 
Join Date: Apr 2008
Location: Wynnum West, Brisbane.
Posts: 4,161
Troy the problem with the 8300 sensor, no matter which camera, is that the readout buffer is a row of standard pixels with 25K well depth. If you read out a heavily exposed and binned image through those buffers, they will overflow and flood the Analogue/Digital Converter on the camera.

What Rick see's on his SX camera is an overload of the A/D converter which causes lines coming out from bright stars when you bin. The gain on the camera is set too high for binning and in SX cameras that gain is set via a pot on the cameras main board. In the QHY cameras they have a feature called clamping which switches in a diode on the A/D converter input and simply lops off hi levels above the clamping point.

Not sure what the others do but the problem is in the sensor itself and the readout buffer. I found it worked best to do 1x binning for everything but I know that's often not possible so next best is to figure out the binning multiplier and reduce binned subs by that factor. I mean, binned x 2 should be 4 times but it's not, it's probably closer to 2.5times. Take some shots of an G2V star and measure it.

So you might end up with 10 minute Lum binned x1 and 3minute RGB binned x2 to get the balance right.

Or buy a camera with a huge well depth and bypass all this crap
Reply With Quote
  #20  
Old 10-08-2012, 06:51 AM
Peter.M's Avatar
Peter.M
Registered User

Peter.M is offline
 
Join Date: Sep 2011
Location: Adelaide
Posts: 947
The big reason that binned subs are not 4 times the signal is that we usually bin RGB frames, and they are filtered shots. Therefor the signal is lower just because you are only capturing one channel. I am sure it is not 100 % efficient also.
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 09:54 PM.

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