RC's and Newts and Theory

[Satchmo]

Originally Posted by bratislav

Stan's analysis is sound if, and only if, you use correspondingly larger pixel camera (with equivalent noise characteristics) on a longer f/ratio scope. But amateurs never do that - we simply use whatever CCD we've got.

I thought he was making an unspoken assumption somewhere and you have pinpointed it.

[bratislav (Bratislav)]

In fact, I've just read Stan's article properly now, and look what I have found :

"But if the object and sky are both low-level (dim or short exp) then camera noise may become significant and degrade the image. Because camera noise is pixel based, this potential degradation is sensitive to the number of pixels used to capture an object or sky-area, and thus it is sensitive to f-ratio."

Clear now ?

[Peter Ward]

Originally Posted by bratislav

... Because camera noise is pixel based, this potential degradation is sensitive to the number of pixels used to capture an object or sky-area, and thus it is sensitive to f-ratio."

Clear now ?

This is simply because you'd be operating in a regime not dis-similar to film.

CCD exposures that are so short as to be camera, rather than *sky* limited, are so full of noise you'd have no hope of getting a deep image hence your point is rather moot

[Peter Ward]

Originally Posted by AlexN

if increasing pixel size does not increase S/N why would you ever bother binning 2x2? I suppose though, that is not physically increasing your pixel size, rather, adding 4 pixels worth of signal into 1?

You get more gain...ie signal. I'm not disputing that. You also get more noise. (rarely is there a free lunch with CCD's)

With a really high QE sensor...eg Hamamatsu, SITe, etc indeeed there is a slight S/N improvement....my thoughts are this is because the shot noise has less of an effect on high QE devices.

I'd expect this benefit is far less with front illuminated sensors eg Kodak.

BTW hands up from all those are actually running a Hamamatsu back illuminated sensor

[AlexN]

I dont even know what a hamamatsu sensor is.. I'm happily running a Sony sensor now, and will be running a Kodak in 6 ~ 8 months time.. Im happy to concede that I know little about CCD science.. All I know is I point my CCD towards the sky with a focused optical system attached to the front and take pictures!

[sjastro]

Originally Posted by Satchmo

Steve

So as the raw image projected on the CCD chip is 4X as bright on the F4 and an F8 scope, won't the brightest parts of the object fill the CCD pixel to full well capacity much faster . So if you can get more exposures in doesn't more stacked images that are at full saturation mean better signal /noise ?

I'm on a learning curve here , but this is one hurdle I'm not over yet...

Hello Mark,

Before the pixel reaches saturation, the signal and noise increase by the same factor hence S/N ratio remains the same where the noise is purely confined to photon noise. That's because you're not actually increasing the total signal, instead the signal is distributed over a smaller number of pixels.

When a pixel reaches saturation it's impossible to measure the S/N ratio. The photon noise is the standard deviation of a series of measurements of a given pixel over a number of exposures. When a pixel reaches saturation it has reached its maximum value.

When the signal is increased by exposure time or by increasing aperture, the signal and noise increase at different rates.

By quadrupling the exposure (or doubling the aperture), the signal increases 4X but the noise increases by only 2X.

Hence the S/N ratio doubles.

Regards

Steven

[bratislav (Bratislav)]

If anything, Hamamatsu sensors are far less sensor/quantization/amplifier noise limited than any of the commercial devices we're dealing with. That is, if camera noise is a factor in a Hamamatsu device, SBIG/QSI/etc will be just as bad (in reality much worse).

But more importantly, binned graphs (in this sense equivalent faster scope, i.e. more photons per combined pixel) continue to be quite clearly above the unbinned curve even when exposure is well into 1000's of seconds.
Check again y axis on that graph - it reads 'Signal to noise ratio'.
Yes, you can reach the similar S/N ratio with slower scope (== unbinned CCD), but it will cost you time. BIG time. For faint objects you will not even be able to expose long enough if a scope is too slow.

[Bassnut (Fred)]

The reason for binning as I understand it, is that its done on chip, so the read out noise for example, will be 4 times less for 2*2 than for bin 1.

[gregbradley]

Thanks Bratislav I think you got that exactly right. I have actually observed this point with various telescopes and you have communicated what I have observed very well.

Faster scopes pick up the fainter nebula etc than slower scopes. Slower scopes will pick it up but at much greater exposure times.

Look at Thomas Davis's dust images with his 12 inch ASA F3.8 scope as a classic example.

You won't see all that dust background so clearly in a 12.5 inch F9 RCOS image without some serious hours of exposure time and even then probably not.

So the cost of long focal length imaging is patience and exposure time to get that lovely closeup image of things.

If you are in a hurry or you travel to image your F5 scope will give a brighter wider field with more of the dim/faint dust/nebula showing up.

This is exactly my experience on the subject.

On top of this of course is a low noise high QE camera as the next most important factor.

Hence in conclusion: ideally you want a high QE, really low noise camera (high cooling, excellent low noise electronics with minimal chip defects) with pixel size most ideally matched for sampling (.66 arc seconds/pixel) for your scope, with largest aperture OTA and the fastest F ratio you can for the image scale you want to image at. All on the best tracking and overmounted mount you can get. Then iamge from the darkest, best seeing location you can for the longest time you can for the resulting maximum best astrophotographic results.

Greg.

[AlexN]

Tom's images are actually a fantastic example now that you mention it.. Get any 12" F/8 scope with the same camera and try to extract the same amount of dusty data with equal sub exposure length, and equal total exposure.. I think you might be hard pressed...

[gregbradley]

Check again y axis on that graph - it reads 'Signal to noise ratio'.
Yes, you can reach the similar S/N ratio with slower scope (== unbinned CCD), but it will cost you time. BIG time. For faint objects you will not even be able to expose long enough if a scope is too slow.[/QUOTE]

Yes I experienced that last time I was imaging. It was 2/3rds moon and I imaged luminance on a galaxy and at 1x1 it was too faint with the light pollution but same exposure length 2x2 gave a quite nice luminance image wiht good contrast.

Greg.

[Peter Ward]

Originally Posted by Bassnut

The reason for binning as I understand it, is that its done on chip, so the read out noise for example, will be 4 times less for 2*2 than for bin 1.

Good point Fred.....but this only applies if your signal-to-noise per pixel is dominated by read noise....this is very rare with deep exposures.

So I’ll modify my earlier position, you can indeed increase the signal-to-noise by binning but the disadvantages of using on-chip binning follow

-it will reduce the resolution of your image
-the relative number of pixels affected by cosmic rays increases
-the relative number of `hot' pixels increases proportionally to the binning factor.

As I said, no free lunch....plus the contribution of the skyglow goes up as well.

[Satchmo]

Originally Posted by sjastro

When the signal is increased by exposure time or by increasing aperture, the signal and noise increase at different rates.

By quadrupling the exposure (or doubling the aperture), the signal increases 4X but the noise increases by only 2X.

Hence the S/N ratio doubles.

So in an F4 scope for a given number amount of time I have available I will be able to increase my signal to noise more efficiently than an F8 scope of same aperture as my raw image is already 4X brighter.

[bratislav (Bratislav)]

Originally Posted by Peter Ward

Good point Fred.....but this only applies if your signal-to-noise per pixel is dominated by read noise....this is very rare with deep exposures.

So I’ll modify my earlier position, you can indeed increase the signal-to-noise by binning but the disadvantages of using on-chip binning follow

-it will reduce the resolution of your image
-the relative number of pixels affected by cosmic rays increases
-the relative number of `hot' pixels increases proportionally to the binning factor.

As I said, no free lunch....plus the contribution of the skyglow goes up as well.

The point may be good, but it is factually incorrect. If you bin 2x2, signal from the extended object adds 4 times, but noise adds sqrt(4) = 2, so signal to noise ratio goes up only 2 times.
On the other hand, if you had twice as large pixel camera, the readout noise is the same, signal still adds up the same, and S/N ratio does go up 4 times.

As pointed out, for well exposed frames it will be the photon noise that dominates the noise floor, and S/N ratio will be relatively independent from f/ratio. (you pay the price with much longer exposures though)

But in astronomy we deal with faint, sometimes extremely faint stuff. There may not always be enough photons to reach photon-noise limited region.

And that was my point from the very first post.

[Terry B]

This is an intresting topic. F ratio seems to important sometimes and as stated by some it can have a bearing on SN ration for very dim extended objects.
I have a question though.
I want to image a small dim object ie a small PN that may only be 20 pixels wide at f8. If I reduce the f ratio to f5.6 (1/2) then I will have a brighter images but it will only be 10 pixels wide. Obviously resolution will suffer in this situation and the only way to improve this would be to increase the aperture to increase the S/N ratio rather than making the image more concentrated on each pixel.
I cant see how this is different to fine structure in large extended objects.
At the lower f ration the dim clouds will have a higher SN ration but any small structure in them will be reduced.

[sjastro]

Originally Posted by Satchmo

So in an F4 scope for a given number amount of time I have available I will be able to increase my signal to noise more efficiently than an F8 scope of same aperture as my raw image is already 4X brighter.

For an equivalent aperture and the same exposure time the answer is no.
It once again boils down to the fact your are still dealing with the same number of photons in both cases.

The best way of explaining this is with an example.
Suppose with your f/8 scope you take an image of 4 equally bright stars. Lets assume the photons from each star fall on four separate pixels. You therefore have 4 separate pixel S/N ratios.

Now you change from f/8 to f/4 by reducing the FL.
All the photons from the 4 stars are now superimposed on a single pixel and saturation hasn't occurred.

While the photon count on that pixel has increased by a factor of 4 as has the signal, it is due solely to the sum total of photons from 4 separate stars. It is not the same as increasing the signal from a single star by 4X which will increase the S/N ratio (N=Photon noise) by 2X.

You would probably find the S/N value will not be significantly different.

Stan Moore's comparison image supports the case.

Regards

Steven

[bratislav (Bratislav)]

Originally Posted by sjastro

For an equivalent aperture and the same exposure time the answer is no.
It once again boils down to the fact your are still dealing with the same number of photons in both cases.

But in one case you have all the photons in one pixel (== one readout+thermal noise figure) while in other case you have the same number of photons split over 4 pixels, with combined electronics' noise that is sqrt(4) or 2 times higher.
This may well be insignificant in cases where signal is so high that shot noise grossly dominates readout & thermal noise.

It is not insignificant for really faint stuff.

Let's be realistic - how many images do we see of Mandel-Wilson stuff with a f/16 Cassegrain equipped with a small pixel camera ? All I've seen are done with instruments f/5 or faster, with CCDs covering several arc seconds per pixel.

Any volunteers ?

[Satchmo]

Steve

The case where 4 stars at F8 get merged into one at F4 seems like an a quite esoteric case and just further confuses me..I think we are just getting bogged down about signal to noise.

Can you turn your explanation towards imaging diffuse objects..galaxies and nebulae.

The general consensus seems to be that the brighter the image or the longer the exposure the better the signal to noise ratio. F4 will record faster with better signal to noise on very faint diffuse objects. Stan Moores two pictures at F4 and F12 show this clearly..for a 10 minute exposure there is less detail and definition in the diffuse areas and the the F12 shot looks dimmer and grainy. There is a slight very impovement in the sharpness of the stars. The F12 instrum,ent would only show 1/9 the sky area of the F4, so it doesn't appear to be a very good trade off.

I'm not sure how this stellar s/n theory translates to diffuse objects that aren't point sources. Will I have no gain in image depth/contrast imaging faint nebulae and galaxies at F4 rather than F8?

[Terry B]

Originally Posted by Satchmo

Steve

The case where 4 stars at F8 get merged into one at F4 seems like an a quite esoteric case and just further confuses me..I think we are just getting bogged down about signal to noise.

Can you turn your explanation towards imaging diffuse objects..galaxies and nebulae.

The general consensus seems to be that the brighter the image or the longer the exposure the better the signal to noise ratio. F4 will record faster with better signal to noise on very faint diffuse objects. Stan Moores two pictures at F4 and F12 show this clearly..for a 10 minute exposure there is less detail and definition in the diffuse areas and the the F12 shot looks dimmer and grainy. There is a slight very impovement in the sharpness of the stars. The F12 instrum,ent would only show 1/9 the sky area of the F4, so it doesn't appear to be a very good trade off.

I'm not sure how this stellar s/n theory translates to diffuse objects that aren't point sources. Will I have no gain in image depth/contrast imaging faint nebulae and galaxies at F4 rather than F8?

At F4 the image will be smaller than at F8 (1/4size) for the same diameter scope.
I think that this is OK for big nebulae but very few galaxies are big and making the image 1/4 the size limits what you can image.

[coldspace]

Originally Posted by Satchmo

Steve

The case where 4 stars at F8 get merged into one at F4 seems like an a quite esoteric case and just further confuses me..I think we are just getting bogged down about signal to noise.

Can you turn your explanation towards imaging diffuse objects..galaxies and nebulae.

The general consensus seems to be that the brighter the image or the longer the exposure the better the signal to noise ratio. F4 will record faster with better signal to noise on very faint diffuse objects. Stan Moores two pictures at F4 and F12 show this clearly..for a 10 minute exposure there is less detail and definition in the diffuse areas and the the F12 shot looks dimmer and grainy. There is a slight very impovement in the sharpness of the stars. The F12 instrum,ent would only show 1/9 the sky area of the F4, so it doesn't appear to be a very good trade off.

I'm not sure how this stellar s/n theory translates to diffuse objects that aren't point sources. Will I have no gain in image depth/contrast imaging faint nebulae and galaxies at F4 rather than F8?

I do understand some things of whats been discussed here on this interesting thread, but although I do not fully understand I know the faster I make my 12 inch F10 Meade the better my Mallincam hypercolor video shows objects like faint nebulae and galaxies, but this also has its limmits to sky fog out of my images.

At f7 the Mallincam shows a fair amount of detail in say Trifid, lagoon etc,etc at 14 seconds exposure/intergrations.
When I drop the F ratio down to F5 using my optec reducers alot more detail/colour appears on my CRT/LCD screens.
Going down to F3.3 really shows alot of detail but on some bright objects such as Omega centaurus/laggon/trifid at F3.3 my Mallincam goes into saftey shutdown mode to protect itself from overload.
On the Orion nebular I can't use more than 7 seconds at 3.3 as te object just way over exposes and the camera goes into saftey shutdown as its a bright object so I either slow the scope to F5 or F7 or turn the gain right down and keep the exposure/intergration times down to 7 seconds or less.

At 28 and 56 second intergrations I have to lower the gain down a few notches on alot of the brighter objects at F3.3 but at F7 or F10 I can crank the gain right up and run the intergrations at 28 or 56 seconds with no problems but I only use these settings and F ratios if am chasing a really faint/ small planetary or a newly discovered Mag 18 supernova.
Alot of guys overseas attach the hyper-star F2 system to their Cats, and are getting fantastic results with the Mallincam in pristine dark sky but they soon found out that at F2 the images would washout or the camera would go into safty shutdown from sky glow in light polluted skies.
This is why F.3 will do me as I live in a light polluted environment

So for my camera at least in practise F3.3 will be the fastest I will run at the lower intergration times for wide field extended objects and F7 and longer intergration/exposure times with a little more gain for chasing small dim objects.
Not sure if this means anything to anyone but thats what I find using my camera/scope setup in practise.
I think you will get more detail at F4 than F8 as thats what happens to my setup but you will also get to the sky fog limmit alot faster so shorter exposures will be required.

Regards Matt.