Are AO systems Obsolete?

[glend (Glen)]

Recent developments with so called "lucky imaging" software and cameras such as the ASI1600mm for DSO imaging, which are offering some initial impressive results have got me wondering if moving optics surfaces (ala AO) has any sort of future, at least in so far as imaging is concerned? Why mess about with relatively slow moving optical surfaces when you can simply discard an image that is out of bounds in a stack of video frames? What do you think?

[h0ughy (David)]

no each has its place

[Slawomir (Suavi)]

I agree with David. Some targets may be successfully imaged with short subs, but for example narrowband imaging of fainter objects will always require longer exposures.

[RickS (Rick)]

Originally Posted by Slawomir

narrowband imaging of fainter objects will always require longer exposures.

Low enough read noise would potentially allow the use of sub-second frames even for narrowband, but we're definitely not there yet, Suavi!

[gregbradley]

Gee, early days to be saying that. Let's see how these new cameras perform first but I do see the potential.

Greg.

[alpal]

Originally Posted by RickS

Low enough read noise would potentially allow the use of sub-second frames even for narrowband, but we're definitely not there yet, Suavi!

Dim targets will always require a lot of time.

[RickS (Rick)]

Originally Posted by alpal

Dim targets will always require a lot of time.

Yes, but that's total integration time, not sub length. If there is no read noise then 60,000 x 1 second subs gives you the same SNR as 100 x 600 second subs.

[codemonkey (Lee)]

Originally Posted by alpal

Dim targets will always require a lot of time.

Not as long as you might think, in terms of individual sub length...

Originally Posted by RickS

Yes, but that's total integration time, not sub length. If there is no read noise then 60,000 x 1 second subs gives you the same SNR as 100 x 600 second subs.

This ^

As for the OP, I'm not so sure. Isn't the whole sales pitch of AO that you can perform corrections multiple times per second? Even if you're down to 5 second subs, there might still be some benefit there.

[alpal]

Originally Posted by RickS

Yes, but that's total integration time, not sub length. If there is no read noise then 60,000 x 1 second subs gives you the same SNR as 100 x 600 second subs.

There is always going to be noise -

it's a physical property of semiconductors.

[Shiraz (Ray)]

Originally Posted by alpal

There is always going to be noise -

it's a physical property of semiconductors.

but the little bit of noise that is left is now close to being vanishingly small compared to detected signal - some lucky imaging systems have sensors with such low noise that they can count the arrival of individual photons and that process has no electronic noise at all (all that is left is the shot noise in the signal). The new CMOS chips are not in that league, but they aren't far off.

Glen, I think that tip/tilt AO will still have a place for some time to come, but the low read noise chips and lucky imaging have the potential to bypass it in some applications - will be interesting to see how this all pans out.

edit: Re narrowband, just ran my system model, but replaced the read noise (5e for the 694) with a read noise of 1.5 and bumped up the dark current to 0.05. Sky-limited narrowband with that assumption cuts in at a bit longer than 2 minute subs for a 5 nm filter and average sky - even at 30 second subs, the stacked SNR is only down to ~0.8 of maximum possible. there definitely seems to be some resolution benefit with subs this short.

[rally]

Ill go with Horses for Courses !.

If your target source is only providing you with 1 electron every minute or two per well/pixel, then a Lucky Imaging sub in the 1/100s range isnt really going to provide you with anything meaningful, whereas a few hours of exposure is.

That will be until we get true noiseless capture or noiseless photon multipliers ! - Can't wait !

[Shiraz (Ray)]

Originally Posted by rally

Ill go with Horses for Courses !.

If your target source is only providing you with 1 electron every minute or two per well/pixel, then a Lucky Imaging sub in the 1/100s range isnt really going to provide you with anything meaningful, whereas a few hours of exposure is.

That will be until we get true noiseless capture or noiseless photon multipliers ! - Can't wait !

if there is no read noise, a few hours of very short subs will yield exactly the same result as a few hours of long subs. The lack of any signal in a sub is not an insignificant result - it tells you just as much about the signal level as having a photon there. You just add up how many photons were detected in all of the subs and that's the signal - if there are subs with no signal, they do not contribute to the sum, but contribute to a calculation of the photon rate. mind you, 1/100 seconds is way beyond what is being discussed for DSOs

[gregbradley]

One weakness with the super short subs is the need to go 8 bit to keep file sizes manageable.
This will reduce dynamic range like that M51 shot with the badly blown core.
Or could that have been avoided?
Greg

[Shiraz (Ray)]

Originally Posted by gregbradley

One weakness with the super short subs is the need to go 8 bit to keep file sizes manageable.
This will reduce dynamic range like that M51 shot with the badly blown core.
Or could that have been avoided?
Greg

the dynamic range scales with the number of subs. If you have 2000 subs with 8 bits (256 levels) dynamic range and stack them, the final result will have 2000x256 possible levels, which is actually pretty good. Planetary imagers routinely stack huge numbers of 8 bit images and end up with much more than 8 bit dynamic range. On dim targets it is not unusual to have only 5-6 bit data - it still stacks out to much better than 8 bits dynamic range.

You would not necessarily use really short subs for DSOs - broadband optimum would be maybe 10-30 seconds with a fast scope.,

[rally]

Ray,

We dont have no noise cameras, my comment was about reality not fiction - but thats why I said "That will be until we get true noiseless capture or noiseless photon multipliers ! - Can't wait !"

But if shot noise still works the same, on a short 1 photon sub it would be 100% ! - a pretty horrible SNR !
Whereas with a long exposure its going to be orders of magnitude less.

I still think its horses for courses

Originally Posted by Shiraz

if there is no read noise, a few hours of very short subs will yield exactly the same result as a few hours of long subs. The lack of any signal in a sub is not an insignificant result - it tells you just as much about the signal level as having a photon there. You just add up how many photons were detected in all of the subs and that's the signal - if there are subs with no signal, they do not contribute to the sum, but contribute to a calculation of the photon rate. mind you, 1/100 seconds is way beyond what is being discussed for DSOs

[alpal]

Originally Posted by Shiraz

but the little bit of noise that is left is now close to being vanishingly small compared to detected signal - some lucky imaging systems have sensors with such low noise that they can count the arrival of individual photons and that process has no electronic noise at all (all that is left is the shot noise in the signal). The new CMOS chips are not in that league, but they aren't far off.

Sorry , - I don't see that at all.
Even a piece of wire has noise in it.
White noise generators use the noise amplified from across a resistor.

[Shiraz (Ray)]

agreed, there is some residual noise, in this case 1.3 electrons RMS per pixel for every read action - and there is also some dark current of about one electron every minute or so - there is no other noise at all (apart from shot noise in the signal). The resistor white noise that you mention is the dark current, and it is negligible due to cooling. Amplifier noise is incorporated in the read noise and, because the pixel-level charge amplifiers are part of the cooled chip, they are very quiet. Because the signal is amplified before it gets off the chip, it is well above any electronic noise in the output circuit, so you actually do end up with almost zero noise. The huge CMOS advantage of having pixel-level amplification was also it's main downfall, because the gains and offsets of the amps varied, introducing fixed pattern noise. The new CMOS chips harmonise the amps, giving very low-noise high-quality images and blistering download rates - well suited to lucky imaging.

The new chips are not sCMOS, but they are getting close to that regime http://www.andor.com/learning-academ...mos-read-noise