Binned and un-binned?

[lil_coz]

Here's a question I've always been meaning to ask...

What is binned and un-binned when it comes to AP?

Thanks!

[knobby]

Very loosely... Binning uses 4 pixels, combines the info then uses it as one... Increases sensitivity.

[lukebl]

As knobby says, it increases sensitivity, but at the cost of lower resolution. Often used to capture RGB colour with shorter exposures, where resolution isn't so important.

[lil_coz]

Ah kool. Thanks for the clear up guys ) So it's not a case of "one's better than the other". I'm guessing depending on what you're after, depends on what method you'll use.

[Kropster]

Is it a software thing or does the camera have to be capable of doing it?

[lukebl]

Is it a software thing or does the camera have to be capable of doing it?

The answer is 'both'!

[ollypenrice]

Better or Not better depends on the rate at which you are sampling the sky in arcseconds per pixel. The atmosphere llimits the resolution possible. Some debate rages around this but going for very fine resolution costs speed and simply isn't worth it if the seeing blurs it out. You really have to decide on the number of arcseconds per pixel at which you want to work and then either use native small pixels or binned larger ones to work at the chosen level.

Binning colour under high resolution luminance gains a little speed but removing luminance from stars tends to improve them - but not if they were captured in binned mode when they can be coarse and blocky.

No free lunch!

One shot colour  cameras cannot normally be binned because the colour information is captured pixel by pixel. Rare exceptions seem to be appearing but not for AP.

Olly

[ChrisLX200]

The remaining issue is dynamic range. Sensors can have large or small pixels. A single large pixel may have (for e.g.,) an electron 'well depth' of 50,000, whereas for the smaller pixel size it may be 18,000. The larger the number then the larger the dynamic range before the pixel becomes saturated. The pixel can keep collecting photons until it reaches that maximum number, at which point it can collect no more - it's as bright as it's ever going to get. So the larger the well depth then the less likely you will burn out highlights, and image definition will still be there in the bright bits and post-processing will bring this out. For smaller pixels, where the well depth has been exceeded (and the pixels saturated) then image data in the highlights is lost and no amount of post-processing will bring it back.

Now, for 'binned' small pixels in an array of 2x2 (for example), although the total for the 4 pixels can be summed to make one large pixel (thus effectively increasing sensitivity at the expense of resolution) the well depth of each individual pixel is still only 18,000. That means the dynamic range is still only the same as for the small pixel sensor and it is still all too easy to burn out highlights by saturating pixels. 

My point is that 'binning' a small-pixel array does NOT give you the dynamic range advantages of having larger pixels at lower resolution, although you can replicate the sensitivity of the latter up to the (lower) saturation point.

ChrisH

[lil_coz]

The answer is 'both'!

AP.... never a simple solution lol.

Thanks for the info Olly!!

[lil_coz]

The remaining issue is dynamic range. Sensors can have large or small pixels. A single large pixel may have (for e.g.,) an electron 'well depth' of 50,000, whereas for the smaller pixel size it may be 18,000. The larger the number then the larger the dynamic range before the pixel becomes saturated. The pixel can keep collecting photons until it reaches that maximum number, at which point it can collect no more - it's as bright as it's ever going to get. So the larger the well depth then the less likely you will burn out highlights, and image definition will still be there in the bright bits and post-processing will bring this out. For smaller pixels, where the well depth has been exceeded (and the pixels saturated) then image data in the highlights is lost and no amount of post-processing will bring it back.

Now, for 'binned' small pixels in an array of 2x2 (for example), although the total for the 4 pixels can be summed to make one large pixel (thus effectively increasing sensitivity at the expense of resolution) the well depth of each individual pixel is still only 18,000. That means the dynamic range is still only the same as for the small pixel sensor and it is still all too easy to burn out highlights by saturating pixels. 

My point is that 'binning' a small-pixel array does NOT give you the dynamic range advantages of having larger pixels at lower resolution, although you can replicate the sensitivity of the latter up to the (lower) saturation point.

ChrisH

That's a great help when it comes to deciding which CCD cam to go for! Thanks for that input Chris!

[red dwalf]

wondering about this myself, finding it confusing which is`nt hard for me, 

just so i understand it correctly,

it`s better to capture luminance channel normally i.e. 1 x 1 binning and the RGB channels in 2 x 2 binning ?

or is it more involved than that

[dph1nm]

The only real scientific reason for binning is if your pixels are dominated by read-noise - then you will get a better S/N ratio from the binned pixel than from combining the four smaller ones. This binning has to be done in hardware NOT software. Otherwise it is really just cosmetic (less data to deal with etc).

NigelM

[ollypenrice]

Yes, assuming you have a reasonable sampling rate at native pixel resolution you'd only bin the colour in order to beat the clouds. I stopped doing it some time ago because the quality of the stellar image suffers. Some will argue that the increase in sensitivity at Bin 2 is 4x unbinned but measurement says otherwise. Expect a gain of around 1.6 times with bin 2.

Olly

Edit. One idea, though. We currently image in widefield with 2x Takahashi FSQ106/Atik 11000 systems in tandem. The 11000 has big 9 micron pixels. The dream scenario would be to retain one of these cameras for colour capture and shoot luminance in a forthcoming 29 megapixel CCD with the same size detector. This way you ought to obtain all the advantages of the large pixels and all the advantages of the small ones! This would be better than binning for the reasons Chris has pointed out. The disadvantage is the cost. T'was ever thus!

[ChrisLX200]

As Olly says, you have to give some thought about matching your pixel size to the f/l of your scope (measured as arc-sec per pixel) but IMHO don't get bogged down with the calculations. To figure out the ideal match you would need to know your average 'seeing' resolution at your usual observing site (I'm lucky to get ~2 arc sec seeing on a good night where I am and very rare for it to be much better than that). Personally I reckon it's better to 'over-sample' rather than 'under-sample' (that is, to sample a little more finely than the seeing allows) in order to avoid getting blocky stars in the final image and recovering the finest detail the seeing allows. There are so many variables here though that it's possible to scrutinise too closely just what is the ideal pixel size, which after all will only be valid for one focal length at one resolution of 'seeing', and likely you'll be using more than one OTA to image with under all conditions from good to bad (so this is an excuse to buy more cameras perhaps ). However, that's where binning can come in useful - always bearing in mind its limitations. My 490EX, a high-res camera with small (3.7um) pixels is such a compromise: for my short f/l refractors (TV NP127is is just over 650mm, TV Genisis is 540mm) and telephoto camera lenses (up to 300mm). With these I'll image at full 1x1 resolution. However, connected to my 10" Cat @ f/10 (2500mm) it needs binning (2x2 or 3x3) to avoid ridiculous over-sampling. If I were just using the 10" Cat I would certainly have gone for a camera with larger sized pixels.

The calculations (and assumptions involved which are critical to those calculations) can get quite involved, however when you're out in your back garden with your basic amateur astro telescope under the usual poor UK seeing then I'm not sure how valuable it is to finesse the pixel-per-arc-sec argument. Some people seem to ramble on about it forever and will berate you for choosing the 'wrong camera' backing up their argument with endless math. There aren't that many really _bad_ choices in amateur level cameras, just avoid specialist or extremes (e.g., using an MX916 11um square pixel sensor with DSLR camera lenses would be one...) and you'll be fine. If you want to read a brief, but more involved description then here is one: http://www.astropix.com/wp/2011/03/01/sampling-and-pixel-size/  and of course there are many more out there if you want to go deeper into the subject.

ChrisH

[Kropster]

Only sounds worthwhile ploughing through the calculations if you are about to take out a second mortgage for some new gear,

 or if you need to baffle the  spouse/partner with some mathematics to justify a new purchase! (Won't work if he/she is better at maths than you are!)

[Physicist13]

Basically think of binning as a noise reduction method. Eg if u bin 4 pixels you get a root(4)2 reduction in the random noise. So to get tht same reduction by taking more lights, you wld have to take 4x more subs! Hence the advantage in binning. Whether you lose out on resolution depends, as Olly says, on what is limiting the effective resolution... Quite often this is the seeing.

[Kropster]

Is this related in any way to the noise reduction built in to DSLRs?

I have it turned off on my Canon 1000D....is that right, or should I consider switching it ON?

[Physicist13]

No its not related. The "noise" reduction, as I understand it (its hard to get the exact details of what goes on!), take a single "dark" frame immediately after the light exp using same exp time, ISO etc, and most importantly, its at the same temp as the light. This "dark" is then subtracted off the light to produce a "noise reduced" final image. This is valid if you can only take one light frame of the image. For Astro Phtotog you can usually take many images (since the target isnt varying in time) and in this case better noise reduction is achieved by taking many "dark" frames after you've taken your lights, but trying to keep the camera at the same temp as the lights. Then producing an averaged "master dark" that is subtracted from each light frame.

so, you are correct to turn the camera's "noise reduction" off for AP, unless, in the v rare instance only a single frame is possible.

Patrick

[iksobarg]

Hello, I was just browsing and your responses have piqued my interest. It was such a simple question! And your answers have sent my mind running. I'll have to read your entries over a few times.

Anyway I'm reading an intro. to AP book and have not seen it covered yet and it's a little off topic maybe(?) but ChrisLX200 went into this ~x arcseconds viewing at the usual viewing site. I have a basic understanding of arcseconds but not so much in relation to his reference or maybe I do and just don't know that I do. Could someone expand on that a little?

In regards to binning , I understand enough that it has a place as a tool in AP and like most or all things has a proportional property.

[ollypenrice]

An arcsecond is a 60th of an arcminute which itself as a sixtieth of a degree, so an arcsecond is a 'size on the sky.' In amateur circles fine resolution means that a pixel records half an arcsecond of sky. This is expressed as 0.5 arcseconds per pixel. At the other extreme you might use a system such as our Tandem Tak which records 3.5 arcseconds per pixel.

Now imagine that both setups mentioned above are pointing at the same evenly bright part of an object. On of these setups a pixel is receiving light from a piece of object measuring 0.5 x 0.5 arcsecs. The other is receiving light from a much larger piece of the object, 3.5 x 3.5 arcseconds in area. Clearly the second pixel is getting a lot more light in a given time.

But, alas, if the bit of object in question has changes of brightness within the 3.5 x 3.5 patch of sky the second setup cannot record those changes but the first one can distinguish between them.

For those who don't like sums (me!) it is all done for you here; http://www.12dstring.me.uk/fov.htm  Just plug in your kit and discover your pixel scale.

Olly

[iksobarg]

The concept strikes me as particularly meaningful so I'm going to check it out in depth. Thanks so much for your response, Olly.

Coincidentally, I'm interested in any one book that covers light emissions (in depth) from the AP angle.

I'll check the web/forums as well.