Focal Reducer vs Pixel Binning?

[The Admiral]

This question relates to using short sub-lengths of less than ~40s taken with an APS DSLR. I'm not referring to long exposures using an accurately aligned EQ here.

If I use a focal reducer, the FoV is increased and photons from an object which fall on the sensor are spread over fewer pixels compared to without a reducer. Consequently, those pixels are filled at a greater rate and the exposure can either be reduced, or one can benefit from increased SNR. Now, when I come to processing I have the option of binning pixels (Star Tools), which again has the effect of increasing the SNR, if I understand correctly.

So my question is, is there any advantage to using a focal reducer when I can get the same effect using pixel binning in processing (except of course when the object extends beyond the normal FoV)?

I can appreciate that with perfectly black skies using a reducer will help overcome the inherent system noise and allow one to see deeper, but in normal skies with the ever present light pollution, is that of particular relevance?

Ian

[Astroboffin]

Hi,

first of all I am confused by a few of your statements.

Using a focal reducer does increase your field of view, and the photons are spread over fewer pixels, but you will be getting more arc seconds of sky per pixel, so your pixel scale will go up.

so yes as you will be getting more light then you can lower the exposure.

as for binning, if,you bin 2x2 in the camera, and you have say 5 micron size square pixels, then after binning those pixels are now 10 micron size in effect and twice as sensitive as before so it makes the camera a lot more sensitive, but at the cost of halving the resolution, but you are still imaging the same amount of sky, so the image will be the same but lower resolution when binning.

binning in the software will not give you the same effect, as hardware binning in the camera.

so the reducer and binning are two completely different things, and binning in the software will not increase your FOV like the reducer will, so there is no way binning can take the place of a focal reducer, as it would have to alter the FOV of the image to do that, and it doesn't, it just lowers the resolution, but makes the camera more sensitive.

so the two things are completely different.

i hope that all makes sense, and is of some help.

AB

Edited January 31, 2016 by Astroboffin

[The Admiral]

Thanks for that AB. Yes it does make sense, but let me take an example.

For the sake of argument, suppose I am imaging, with a sensor of 2000 x 2000 pixels, an object whose dimensions are half that of the sensor (i.e. 1000 x 1000 px). Now, I introduce a 0.5x focal reducer. The image of the object becomes 500 x 500px in a 2000 x 2000px frame. In other words, all the photons from the object are now spread over 250,000 pixels rather than 10⁶ px, and so each pixel will receive 4x more photons. At the same time, the image will have 'shrunk', the resolution will have dropped, and so the image will need cropping if it is to appear the same relative size in the final image. The cropped image size would need to be 1000 x 1000px.

If instead, I bin 2x2 in the processing, the SNR will improve. My image proportions will remain the same (i.e. I won't need to crop), but the overall image size will halve to 1000 x 1000px.

So it looks as the two approaches achieve the same thing. I end up with a smaller, lower resolution, 1000 x 1000px image (assuming I want the object image to be the same relative size), in which the SNR is improved.

I can see that if you actually want to image a wider field, then a reducer is the only way to do it, but if I don't, then from my very naive perspective it seems that using a reducer or binning pixels are equivalent.

So I'm really looking for confirmation from you experienced guys whether or not my naive logic is correct, and if not, what have I got wrong.

As to whether binning the camera output or in post-processing is the same thing is a moot point really, as I'm not in a position to bin the output from the DSLR, though I can see that there would be a difference.

Ian

[ollypenrice]

This is a vexed question and is flirting with the dreaded F ratio myth.

But firstly the only form of binning possible here is software binning because we cannot bin a one shot colour camera.

Whether or not focal reducers 'speed up' capture depends on what part of the image interests you because what they clearly don't do is bring in any new photons from an object which fits on the chip without reducer. Viz;

In my opinion you might just as well crop and software bin in the scenario above as use a focal reducer. If you downsize the native FL image the information it contains is concentrated onto fewer screen pixels and the image will look less noisy. This is why, when you stretch an image, you must always keep checking it at 1 to 1 because when software binned (downsized to fit the screen) the noise won't show. If you decide to present an image at 50% of full size you need far less data than if you want it to hold up at full size. I see no advantage here in using a focal reducer.

Where a focal reducer really does bring in useful new photons is where it brings into the frame something you want to see;

In this scenario you have useful light from NGC1977. The light from the empty sky around M33 was not useful.  Of course the pixels 'fill' faster with a FR and the widefield M42/NGC1977 will reach an acceptable S/N ratio faster than the M42-only image on the left. But if you gave both images the same exposure time and then cropped out M42 from the right hand image it would be no better than the M42-only from the left reduced to the same screen size.

Such is may take on all this, anyway.

Olly

Edited January 31, 2016 by ollypenrice
Typo

[The Admiral]

Thank you Olly, it's comforting to know that I'm not going on a wild goose chase!

Ian

[dph1nm]

Hmm - with only 40s exposures I suspect a typical DSLR shot might be read-noise limited. In which case a focal reducer might improve the overall S/N per object (less read noise per square arcsec on the sky), whereas software binning will not.

NigelM

[The Admiral]

Thanks NigelM, I wondered about whether that might be an important factor, which is the reason for my last paragraph in the OP. But I've no idea of the scale of these factors however, and whether the signal from sky background (incl light pollution) dwarfs the read noise, or vice versa. If the former, then I presume a reducer would be of marginal help?

Ian

[dph1nm]

Yes, if you are truly sky noise limited then you won't get any benefit.  Whether you are depends on may things - your telescope aperture, pixel size, sky brightness, camera efficiency & read noise etc, so it is really something you have to determine for your particular situation.

NigelM

[The Admiral]

Thanks Nigel

Ian

[Uranium235]

4 hours ago, Astroboffin said:

binning in the software will not give you the same effect, as hardware binning in the camera.

 

Thats a question I've been asking myself for quite some time! And it confirms my hunch (ie: in relation to Maxim 2x2 software binning) - no, its not as good as hardware binning.

[dph1nm]

21 hours ago, Uranium235 said:

Thats a question I've been asking myself for quite some time! And it confirms my hunch (ie: in relation to Maxim 2x2 software binning) - no, its not as good as hardware binning.

If your pixels are sky noise limited then software or hardware binning should give the same effect. If you have a significant read noise contribution then hardware binning is best.

NigelM

p.s. if you are going to software bin a DSLR it needs to be done before debayering the data.

[The Admiral]

32 minutes ago, dph1nm said:

if you are going to software bin a DSLR it needs to be done before debayering the data.

Forgive me asking a stupid question, but how is binning done on 4-pixel arrays?

Ian

[ollypenrice]

1 hour ago, dph1nm said:

Sorry, false post. I misread a reply!

Edited February 1, 2016 by ollypenrice

[Astroboffin]

4 hours ago, The Admiral said:

Forgive me asking a stupid question, but how is binning done on 4-pixel arrays?

Ian

you can't hardware bin a colour camera, whether it be OSC CCD or DSLR, but I guess you can software bin, but have no clue how as never needed to do it.

AB

 

[vlaiv]

5 hours ago, dph1nm said:

If your pixels are sky noise limited then software or hardware binning should give the same effect. If you have a significant read noise contribution then hardware binning is best.

NigelM

p.s. if you are going to software bin a DSLR it needs to be done before debayering the data.

Can you elaborate on this one? Noise is noise, no matter which source it comes from (sky flux poisson process or read noise or thermal noise), it may have different statistical characteristics, but in general it will be random unwanted addition to signal.

My view on things including original question is as follows:

Prefer mosaic + software binning (of stacked and stitched mosaic) over focal reducer (at least over any reducer that has not been specially designed for particular scope and of highest quality).

Pros of mosaic + software binning:

Smaller FOV to work with - less off axis aberrations of any kind (coma, astigmatism, field curvature) which are usually amplified by focal reducer.

Making mosaic in general does not require much more time to achieve target SNR (not really true due to read noise, but close enough). For example: you consider to take 16 subs over target area. In case of mosaic (2x2) you would use 4 subs for each mosaic segment. 16 subs would give you x4 SNR after stacking, while for other case: 4 subs per segment give you x2 SNR reduction for segment, and then you bin 4 segments (mathematically it is equivalent of stacking 4 subs) - x2 SNR increase again - 2x2 = x4 SNR, same as in first case. SNR has quadratic dependence, but so does area.

Edited February 1, 2016 by vlaiv
typo and addition

[dph1nm]

The advantage of hardware binning is that you only get one dose of read noise per binned pixel, whereas e.g. with 2x2 binning you would have four doses of read noise in the unbinned image covering the same area of sky. So the read noise contribution per square arcsec goes down. With software binning you always have the four doses of read noise whatever you do. This is the only real reason professional observatories hardware bin their cameras - mostly for spectroscopy I think, although it is sometimes possible to do it for imaging.

NigelM

[dph1nm]

23 hours ago, The Admiral said:

Forgive me asking a stupid question, but how is binning done on 4-pixel arrays?

You simply treat the red green and blue pixels as separate images and bin up adjacent pixels of the same colour. You can then reform this back into a bayer array (if you want). Unfortunately I don't know whether any readily available software does this!

I think there were some Kodak CCDs which genuinely did hardware colour binning, by some fancy clocking of the readout so that the different coloured pixels ended up in different registers, but you don't hear of these much nowadays.

NigelM

[vlaiv]

On 2.2.2016. at 14:45, dph1nm said:

The advantage of hardware binning is that you only get one dose of read noise per binned pixel, whereas e.g. with 2x2 binning you would have four doses of read noise in the unbinned image covering the same area of sky. So the read noise contribution per square arcsec goes down. With software binning you always have the four doses of read noise whatever you do. This is the only real reason professional observatories hardware bin their cameras - mostly for spectroscopy I think, although it is sometimes possible to do it for imaging.

NigelM

This implies that even in sky noise dominated scenario there will be difference between hardware and software bin, maybe not significant one, but still, even in this case hardware bin will be better than software. Only when there is no read noise they are the same.

Vladimir

[Doversole]

Sorry to resurrect such an old post, but having just invested in a Stellalyra RC8, I was thinking about the benefit of a focal reducer.

It seems reading all of the replies above that:

- if your object is small with no interest to increase the FOV, binning is preferable (the F ratio myth)

- even if you want to increase the FOV, the preference would be mosaic and binning rather than a FR (that's something I need to look at!)

If I didn't misread you, you just saved me some £££ and headache to fit the whole image train with the smaller backfocus of a FR!

[ollypenrice]

3 hours ago, Doversole said:

Sorry to resurrect such an old post, but having just invested in a Stellalyra RC8, I was thinking about the benefit of a focal reducer.

It seems reading all of the replies above that:

- if your object is small with no interest to increase the FOV, binning is preferable (the F ratio myth)

- even if you want to increase the FOV, the preference would be mosaic and binning rather than a FR (that's something I need to look at!)

If I didn't misread you, you just saved me some £££ and headache to fit the whole image train with the smaller backfocus of a FR!

A mosaic will take longer but produce a larger image. Will this larger image contain more resolved details? Not if you're oversampled - as you are likely to be with a long focal length and small modern pixels. So will you make the most of the theoretically more detailed mosaic alternative? I think it unlikely if you're sampling below an arcsecond per pixel and even that might be beyond your local seeing. Do you have an idea of what it will support?

I recently worked on an OSC dataset which had been stacked in APP and resampled downwards at the stacking stage. I don't know the details of the process but the data, though captured in only 90 minutes, was incredibly clean.

I must say that I'd be wary of adding complexity to an already challenging RC - but that's just me!

Olly

Edited March 27, 2021 by ollypenrice
typo

[alacant]

7 hours ago, Doversole said:

the benefit of a focal reducer

Hi

Caution. Hands on experience with long focal length telescopes follows.

We consider ourselves lucky if we get 2 nights per month when the atmosphere is stable enough to allow us to image even at 1200mm. DSLR.

Neither does there seem any reliable way to determine what the seeing will be on any given night, but do tell us otherwise. What we normally do is plan a session with a shorter focal length telescope. Then, if we find the guiding up to it, swap to our long reflector. [1]

Your big advantage over our Newtonians is that you don't need to introduce glass into the imaging train. Even our -good quality- 4 element cc introduces annoying colour and fatter-than-need-be-because-of-it stars. Our €0.02 would be to wait for the good nights for your RC and image with something shorter on other occasions.

Cheers

[1] EDIT: Our current changeover record is around 7 minutes. But that's with two of us. One on the hardware, The other on the software. We're working toward sub 5 minute changes!

Edited March 27, 2021 by alacant
ortografía

[vlaiv]

3 hours ago, Doversole said:

Sorry to resurrect such an old post, but having just invested in a Stellalyra RC8, I was thinking about the benefit of a focal reducer.

It seems reading all of the replies above that:

- if your object is small with no interest to increase the FOV, binning is preferable (the F ratio myth)

- even if you want to increase the FOV, the preference would be mosaic and binning rather than a FR (that's something I need to look at!)

If I didn't misread you, you just saved me some £££ and headache to fit the whole image train with the smaller backfocus of a FR!

Depends on the sensor you are planing on using.

For smaller sensor - it actually makes sense to use focal reducer on RC8.

Most currently available cameras have small enough pixels that you'll have to bin in order to get to good sampling resolution with such long FL scope, so reducer will not help you there, but it will help you make the most use of telescope.

RC8 has about 30mm of usable field. I think it is a bit less - maybe APS-C size, so say about 28mm - but in either case it is both big and small - depending on how you look at it. It is much larger than 4/3" or smaller sensors but not big enough to cover full frame sensor.

Putting camera like ASI183 that has 16mm of diagonal is sort of waste of good field of telescope. Even ASI1600 / ASI294 that have diagonals of about 22-23mm are smaller than usable field of RC8.

In all of these cases - it makes sense to add focal reducer to exploit whole usable field, and not only that - camera with smaller sensor + focal reducer costs less than alternative larger sensor dedicated astro camera.

Take for example ASI294mc which has 23.1mm diagonal and costs £980 and expensive reducer like SharpStar 2.5" x0.8 reducer for RC scopes which costs £400 - you'll cover the same amount of sky 23.1 / 0.8 = 28.88mm as APS-C sized sensor of ASI2600MC that costs £1900 or say something older like ASI071 that costs £1440.

Not to mention that you can get CCDT67 or CCD47 to do the same job for half the price of SharpStar reducer.