Have I understood Binning correctly?

[Fordos Moon]

Mono CCD camera on the way!

I have been trying to get my head around "binning".

From what i understand when I am doing a LRGB image, I can "bin" the RGB captures to reduce capture times.

Questions:

1. With "binning" am I increasing sensitivity at a cost of resolution, the latter being less important with RGB?

2. Does this mean if I do for example 10x1200 seconds luminance I do 10x300 seconds for each of RGB?

3. I will need to resize the RGB DSS results before combining them with the Luminance in Photoshop?

4. I presume Atik's Artemis software makes option to "bin" straightforward as from what I read it is very good software?

Many thanks as always.

Bob

[michael.h.f.wilkinson]

Cannot help with all points, but the first (theoretical) point is correct. The human eye is less sensitive to changes in colour than changes in brightness. This is why JPEG stores chrominance data at half the resolution of luminance data.

[PhotoGav]

Hi Bob, exciting times ahead!

1. Exactly. However, is resolution less important for RGB? Discuss!!!

2. Yes, if you are after the 'equivalent' of 1200s subs in RGB, though you might well want much shorter subs for RGB to avoid over exposing the stars.

3. Yes, you will need to upscale the 2x2 binned images by 2x to match the 1x1 images.

4. No idea, but I'm sure it will be a simple setting change. I use ATP and it is a drop down menu with all the possible binning options for my CCD.

Regarding binning RGB data in general, I have read differing opinions on this. Yes, it is faster, but if you stick to 1x1, maximum resolution, surely you are going to produce a better image? I have experimented with both, but cannot draw any firm conclusions, yet. Olly has mentioned combining L and RGB subs when all shot at 1x1 to increase your total exposure time, but I haven't worked out how to actually do that yet!

Good luck.

[teleskopjo]

I still have not understood what actually is binning however ;....!

[thanders]

From the QSI583 manual:

Binning allows you to combine multiple pixels into what are effectively larger pixels with up to sixteen times the surface area and sensitivity as an individual pixel. The standard binning mode is 1x1 meaning that each logical pixel is equal to one physical pixel. The QSI 500 Series supports symmetrical and asymmetrical binning up to at least 3 pixels in either axis. Some models offer additional binning options. Larger binning modes like these provide increased sensitivity and dynamic range at the expense of resolution. Asymmetrical binning (e.g. 1x3 or 3x2) is not commonly used but can be very useful for specialized applications such as spectroscopy.

The following image illustrates how rows and columns of pixels are combined when binning an image. Notice how the effective dimensions of the CCD changes with the degree of binning.

1x1 Binning 2x2 Binning 3x3 Binning
3326x2504 1663x1252 1108x834 QSI 583 image size

Binning other than 1x1 is typically used in two different situations. First, the human eye is much more sensitive to subtle changes in image brightness than it is to small changes in color. This can be taken advantage when producing color images by taking luminance images through the “L” filter binned 1x1 and red, green and blue images binned 2x2. When combined with the luminance image the 2x2 binned color image is almost indistinguishable from using 1x1 binning for the color frames.

The second situation where 2x2 or higher binning is used is when seeing conditions don’t support high resolution imaging. If each pixel on your CCD “sees” 2 arc seconds of the sky through your telescope and the local seeing conditions are only 4 arc seconds, you won’t benefit from the increased resolution offered by 1x1 binning. In this case, binning 2x2 will yield optimal results in a shorter amount of time. 

/T

[dph1nm]

The only real advantage to on-chip binning is that you get 1 dose of read noise instead of 4 (for 2x2 binning). If read noise is a significant contribution to the noise in each unbinned pixel then this will improve the overall s/n of your image and is a good thing to do. If your pixel noise is dominated by e.g. shot noise in the sky background then there is no advantage to on chip binning, as you could just take an unbinned shot for the same length of time and bin it up in software afterwards and get exactly the same result. So in this case binning does not mean you are able to exposure for a shorter time. You do have smaller images to store and process, however, so there might still be a practical advantage!

NigelM

[NickK]

Binning is one of those buzz words.

Cameras can do this in two ways with varied pros and cons:

* Binning using the timing pulses on the sensor read - this means the charge from pixels is summed in the read register. This means your charge going into the ADC is larger and therefore has the ability to provide a larger range than a single pixel' worth of charge alone. The problem here is that a charge read register usually has about the same capacity as a single pixel.. so it's easier to saturate.

If you have a bayer matrix then it does not make sense to bin in this way otherwise the colours are getting mixed up. Normally the silicon CCD sensor itself dictates this so any camera manufacturer that is claiming non-CCD based binning modes is either misleading or has additional expensive timing electronics capable of sorting CCD charge without adding noise.. which is normally out of the pockets of the amateur.

* Binning using in-camera software - this is where the camera reads each pixel and converts it in 1x1 binning but then simulates the binning by adding the pixel values in the image buffer. This can cope with colour bayer matrix but really isn't proper binning in my view as you don't gain any additional ADC charge range on faint objects.

[ollypenrice]

If you measure the gain in signal from binning it comes out far below the x4 that you might expect. It's nearer to x1.5 to x2.

I don't do it. For one thing I try to make my starfeilds from RGB only, not LRGB, so I want nice tight unbinned stars.

Olly

[carastro]

Binning is good if you are short on time for the RGB which we frequently are here in the UK.  It means your chances of completing an image without having to wait weeks to get the other filters are greater since the time required is far less.  

Re-sizing the images so they match the luminance or Ha is less easy.  Can't remember if I tried that in Photoshop, but I use Registar and it works a treat.  Registar is definitely worth having if you use different telescopes and cameras to combine images as well.  

This is an image I did over the space of 2 years.  The colour was done with a DSLR in 2011 and the Ha was done in 2013 with an Atik383L, no idea what scopes I was using, but by using registar I was able to combine them.

Sorry to go slightly off topic, but I thought the use of Registar to line up images example might be useful, as this will help for re-sizing binned images:

[dph1nm]

It means your chances of completing an image without having to wait weeks to get the other filters are greater since the time required is far less.

Only if you are read-noise limited. Otherwise the exposure time required is not less.

NigelM

[lensman57]

Mono CCD camera on the way!

I have been trying to get my head around "binning".

From what i understand when I am doing a LRGB image, I can "bin" the RGB captures to reduce capture times.

Questions:

1. With "binning" am I increasing sensitivity at a cost of resolution, the latter being less important with RGB?

2. Does this mean if I do for example 10x1200 seconds luminance I do 10x300 seconds for each of RGB?

3. I will need to resize the RGB DSS results before combining them with the Luminance in Photoshop?

4. I presume Atik's Artemis software makes option to "bin" straightforward as from what I read it is very good software?

Many thanks as always.

Bob

Hi Bob,

Olly is correct, someone had done a proper study of 2X2 binning and found that on average the gain in signal is about 1.6X max and not 2 at the expense of 1/4 resolution ( sorry I cannot remember whom and where but I could have read about it in CN forums ). That is all of course if you are imaging at a scale that allows you binning ie: well oversampling but as an exercise it maybe worth a try. There is also another line of thought that suggests that an RGB image made up of long enough RGB subs with something called a Synthetic Lum can actually be as good if not better than LRGB but due to the lovely Manchester weather, last proper imaging session was in last Nov, I have not had the chance to try this. As for resizing and all that you can just enable 2X drizzle  for RGB I would imagine and that should rescale the RGB to Lum but I maybe wrong there ( willing to try anything but pay for Registar, this has become an obsession with me ) .

A.G

[ollypenrice]

Yes, the other argument in favour of not binning is that you can extract a synthetic luminance from your RGB. I have never found such a a syn lum to be as good as a real lum even allowing for the 1/3 eficiency in shooting through RGB filters. I generally find that a syn lum is worth about 25% of a real one for equivalent time but, even so, that is worth having. So maybe the gain in shooting RGB in bin 2 is not quite what it seems since there is no point in extracting a syn L from a binned RGB.

By the way, the tests to which I referred were carried out by Dennis, now on PAIG.

Olly

[dph1nm]

Olly is correct, someone had done a proper study of 2X2 binning and found that on average the gain in signal is about 1.6X max and not 2

Sorry - I have to object slightly to this! With on-chip 2x2 binning of a uniform light source (e.g. the sky), the signal in the binned pixel should be 4x that in each unbinned pixel. The Shot noise will be the sqrt of this i.e. 2x higher, so the s/n in the binned pixel  should be 2x better that than in an unbinned pixel. There is also less read noise, so the actual improvement may be slightly better than this.  If this is not the case, then you have a problem with your camera, as it implies the unbinned pixels are correlated (or, I suppose, the read noise for the binned pixel is much higher than it should be).

NigelM

[kirkster501]

Great thread,  I have abandoned binning and am doing 1x1 for everything.  That way i have go the data in a "perfect" form, thats the way I view it for right or for wrong....

Despite clear nights recently I have been too tired to do much imaging and have just been observing.  I seem to be permanently exhausted lately, it's been a long winter!

[ollypenrice]

Sorry - I have to object slightly to this! With on-chip 2x2 binning of a uniform light source (e.g. the sky), the signal in the binned pixel should be 4x that in each unbinned pixel. The Shot noise will be the sqrt of this i.e. 2x higher, so the s/n in the binned pixel  should be 2x better that than in an unbinned pixel. There is also less read noise, so the actual improvement may be slightly better than this.  If this is not the case, then you have a problem with your camera, as it implies the unbinned pixels are correlated (or, I suppose, the read noise for the binned pixel is much higher than it should be).

NigelM

This was based on measured ADU values using an SBIG1100, I think. I found that I agreed with it using my Atik 4000s. I haven't tested my 11000s. My own tests were based on real images which, after all, is what matters. 

Olly

[dph1nm]

This was based on measured ADU values using an SBIG1100, I think. I found that I agreed with it using my Atik 4000s. I haven't tested my 11000s. My own tests were based on real images which, after all, is what matters.

Interesting - it could be there are problems with amateur chips, but certainly for professional CCDs you get what the theory says you should. We have various cameras (SBIG, QSI) around here which we bin 3x3 for the students to use on 10-14 inch Meades, so maybe I will try and persuade someone to do some unbinned shots as well ...

NigelM