372 replies to this topic

[Alex Parker]

 

Also I used to have a KAF8300 based camera. I think the ASI1600 will exceed it in performance for imaging. They key reason being lower read noise. You need much (30s vs. 300s) shorter exposures (although total exposure times will be similar but more subs) to generate the same SNR which equals much easier calibration.

 

30 seconds vs 300 seconds?  I'd probably check the math on that to make sure you have reasonable expectations.  There is no chance that is going to happen.  

 

 

Here's an example of a pretty nifty LRGB image shot with one of these cameras using 30sec subs..http://www.cloudynig...23#entry7191548

[groz]

There's not a lot of reason to have more bits than it takes to encode the well depth at 1e/ADU.

 

This is true, but, there is a significant penalty for having less bits than it takes to enode the well depth.  20K well depth needs 20K points to describe, and 12 bits only gets you 4K points.  14 bits comes close to properly describing a 20K well depth.

[freestar8n]

If people don't like math - please avert your eyes. There is no real need to have gain set so that it is less than 1 e/adu because the actual digitization error is gain/sqrt(12) - and you always have read noise blurring all the values. The total noise from digitization and read noise will be sqrt(gain^2/12 + r^2).

The gain needed to use the full well is about 5 e/adu (20,000/4096=4.9) so the digitization noise is 3.46e. But as you increase the gain value (make it less sensitive) the read noise also goes up - about to that same level.

So with perfect digitization the noise would be just the read noise at about 3.5e I think. And with digitization error it will be about 4.9e. That is much less than read noise alone in an 8300 sensor.

But if you are stacking many frames anyway, you don't need to fill the well and you can use a more sensitive gain setting - and have lower read noise and no issue with digitization.

So that's the math view - and it's different from just saying 12-bits is a fundamental problem.

But I also agree it will be good to see images to know just how well they behave in the field. If they do behave well and don't have unexpected noise terms that can't be corrected - or inconsistent behavior that doesn't calibrate well - I don't see why they shouldn't work very well for both planets and deep sky - including faint deep sky.

Frank

[FirstC8]

Most of the CMOS sensors also have imaging modes, including a HDR (High Dynamic Range) setting. This mode will typically takes 3 different exposures and combines the for a final image... all on sensor. Not that useful for Astronomy, but it's there. My Flea 3 CCD camera can do the same, but it's not done on the sensor, I don't use it there either :-).

It would be nice if the HDR mode can work in long exposures for astroimaging though. Are there any technical reasons to prevent this feature from happening at the chip/circuit level?

[FirstC8]

Regarding stacking to fill the void 12 bit well creates, of course stacking so far has been the holly grail of astroimaging all along, so that is good.

But any forward looking view would suggest at some point maybe stacking can be a thing of the past?

When I used my cheap 1/3 chip trained on M57 for an hour exposure, the single sub did not need any stacking or calibration to show a fine m57 core and the surrounding bright stars. The good signals had more than swamped any noises, whether sky noises or chip noises.

Only when I wanted to stretch the subs to reveal the faint outer emission rings, the noises began to show their ugly presence.

So theoretically if the chip becomes good enough, and seeing/tracking allow it too, at some point we can do very long exposures, as long as it is necessary to not have to stack or calibrate?

In other words, astroimaging one day can be like taking terrestrial images under decent ambient light and return some stunning images, rich colors with decent resolution, in just one click.

Edited by FirstC8, 08 May 2016 - 05:59 PM.

[Jon Rista]

Frank, just to make sure I understand...it is SQRT(12) because the bit depth is 12? If it was a 14-bit ADC, it would be gain/SQRT(14) and for 16-bit ADC it would be gain/SQRT(16)?

[FirstC8]

And this is that raw one hour long sub I was talking about, no stretching, no calibration, no darks no nothing, just one good old shot at M57 for an hour in H alpha, by a cheap 1/3 chip full of amp glows, vertical line patterns and other fixed noises, not to mention the light pollution and still bright moonlight in the background at the time.

 

It is not some far fetched dream that given the advances one day we can do this with most DSOs?

Attached Thumbnails

• 

Edited by FirstC8, 08 May 2016 - 06:11 PM.

[Jon Rista]

Regarding stacking to fill the void 12 bit well creates, of course stacking so far has been the holly grail of astroimaging all along, so that is good.

But any forward looking view would suggest at some point maybe stacking can be a thing of the past?

When I used my cheap 1/3 chip trained on M57 for an hour exposure, the single sub did not need any stacking or calibration to show a fine m57 core and the surrounding bright stars. The good signals had more than swamped any noises, whether sky noises or chip noises.

Only when I wanted to stretch the subs to reveal the faint outer emission rings, the noises began to show their ugly presence.

So theoretically if the chip becomes good enough, and seeing/tracking allow it too, at some point we can do very long exposures, as long as it is necessary to not have to stack or calibrate?

In other words, astroimaging one day can be like taking terrestrial images under decent ambient light and return some stunning images, rich colors with decent resolution, in just one click.

 

I don't think that would ever be the case. There will always be limits. Full well capacity for one is a key limiting factor...with a single hour-long expsoure, a lot of stars are going to clip, especially for non-galaxy objects nearer to our own galaxies arms and core. Assuming a perfect sensor with 100% Q.E. and 0 read noise, you would still need to stack to make effective use of the dynamic range of the sensor. It wouldn't really matter how long the subs were at that point...you would be photon counting, and would have no read noise...so you could stack short or long subs. The only thing that would matter in the end is the total integration time. For truly faint objects, you are still going to need several hours of exposure on those, just to reveal something (i.e. OU4, CTB-1). To reveal the faintest details with strong signal, you would still likely need several more hours. No matter how you slice it, you will always have shot noise.

[freestar8n]

Sqrt(12) is the sigma of a uniform distribution with width 1.0. So 1 e/adu has much less error than 1.0.

Frank

[freestar8n]

I mean 1/sqrt(12)

[FirstC8]

Regarding stacking to fill the void 12 bit well creates, of course stacking so far has been the holly grail of astroimaging all along, so that is good.

But any forward looking view would suggest at some point maybe stacking can be a thing of the past?

When I used my cheap 1/3 chip trained on M57 for an hour exposure, the single sub did not need any stacking or calibration to show a fine m57 core and the surrounding bright stars. The good signals had more than swamped any noises, whether sky noises or chip noises.

Only when I wanted to stretch the subs to reveal the faint outer emission rings, the noises began to show their ugly presence.

So theoretically if the chip becomes good enough, and seeing/tracking allow it too, at some point we can do very long exposures, as long as it is necessary to not have to stack or calibrate?

In other words, astroimaging one day can be like taking terrestrial images under decent ambient light and return some stunning images, rich colors with decent resolution, in just one click.

I don't think that would ever be the case. There will always be limits. Full well capacity for one is a key limiting factor...with a single hour-long expsoure, a lot of stars are going to clip, especially for non-galaxy objects nearer to our own galaxies arms and core. Assuming a perfect sensor with 100% Q.E. and 0 read noise, you would still need to stack to make effective use of the dynamic range of the sensor. It wouldn't really matter how long the subs were at that point...you would be photon counting, and would have no read noise...so you could stack short or long subs. The only thing that would matter in the end is the total integration time. For truly faint objects, you are still going to need several hours of exposure on those, just to reveal something (i.e. OU4, CTB-1). To reveal the faintest details with strong signal, you would still likely need several more hours. No matter how you slice it, you will always have shot noise.

Of course I have no illusion this can be done for all faint objects or features. What I was thinking is, with the advances of chip technology, combined with tracking capability, and excellent dark site conditions, one shot astroimaging may be possible for more and more DSOs over time.

[Jon Rista]

I think it would depend on how faint of details you want to reveal, and how clean you want them to be. With 100% Q.E. and 0e- read noise (and let's also say zero dark current), you have a perfect sensor. However, even with a perfect sensor, you still have to deal with shot noise in the signal itself, as well as shot noise from airglow (if you are at an exceptional dark site). You would need an extremely long exposure, hours, to image IFN with clean results, for example...because it will still have shot noise, and it is extremely faint.

 

Full well capacity is a function of pixel size (area of the photodiode). So, maybe if you were using 24 micron pixels, and if you had perfect polar alignment without any drift, no wind, no other environmental factors that might ruin the exposure, no frame intrusions like airplanes, meteors, satellites, etc. Then, when you have a perfect sensor and perfect imaging conditions and perfect tracking at a perfect dark site...you might be able to simply expose for hours on end until just before twilight, then end your exposure. 

 

That would really indeed have to be one heck of a PERFECT night, though. I've never experienced such a night myself. Tracking issues and frame intrusions alone would ruin even a single 1-hour sub, assuming I just wanted to keep that sub as my final image. There are key benefits to stacking, such as sigma rejection, and dithering, which allow us to get more ideal results than a single exposure allows. It is more reliable as well...because there is always going to be something that screws up long subs, and if you are going for just one big long sub over an entire night...then if anything messes it up, you just lost your entire night, rather than just "one sub". 

[josh smith]

And this is that raw one hour long sub I was talking about, no stretching, no calibration, no darks no nothing, just one good old shot at M57 for an hour in H alpha, by a cheap 1/3 chip full of amp glows, vertical line patterns and other fixed noises, not to mention the light pollution and still bright moonlight in the background at the time.

It is not some far fetched dream that given the advances one day we can do this with most DSOs?

That looks great but is also incredibly bright and parts of it are blown out. Thats why you always have to consider the object when people talk about making things work. People like to say an alt az mount is fine and it is depending on your expectations. I can guarantee that most images shown in that argument are going to be of M51, M42, the Ring, M31, and the Dumbbell.

That M101 above looks nice. I haven't seen a reasonable size image of it yet to look at and nearly any camera is going to do well with 300mm of aperture at a fast speed from dark skies.

It's clearly going to be a good deal and a nice camera. Whether or not it replaces 8300's and 694's completely, I don't know.

One of the main things that continues to not be discussed much is the package. If I have to buy a separate oag, fw, and camera, than that is 3 separate things that can go wrong from 3 separate companies assuming I got all of the spacing and hardware right.

I'm looking forward to more results!

[bobzeq25]

 

And this is that raw one hour long sub I was talking about, no stretching, no calibration, no darks no nothing, just one good old shot at M57 for an hour in H alpha, by a cheap 1/3 chip full of amp glows, vertical line patterns and other fixed noises, not to mention the light pollution and still bright moonlight in the background at the time.

It is not some far fetched dream that given the advances one day we can do this with most DSOs?

I'm looking forward to more results!

 

OK.  Granted, it's not your usual DSO setup.  A 16" Dob.

 

Pretty darn nice M51 here, 2000X1".  The outer reaches and background galaxies qualify as "dim" by my lax standards.  <grin> 

 

http://www.astrokraa...l_Kraaikamp.jpg

Edited by bobzeq25, 08 May 2016 - 07:35 PM.

[FirstC8]

That looks great but is also incredibly bright and parts of it are blown out. Thats why you always have to consider the object when people talk about making things work.

Of course, which is why I started this conversation by mentioning the HDR functions already imbedded in today's CMOS circuits. They are clearly not designed for astroimaging, but the technologies are already here and quite mature.

As the points Jon raised, they are well taken. I suppose there is a limit how long an exposure can realistically be before it becomes high risk. I'd put a limit at around two hours.

In a good night, there could be good 6 to 8 hours of exposure time, having a good two hour sub out of the night is not all unreasonable. And when very long exposure time is needed to bring out very faint features, occasional interference such as wind or tracking errors can be tolerated more so than one may think.

Also my so called "one shot astroimaging" does not rigidly exclude any calibration. Darks can still be used, stacking can still happen to remove outliers.

This is the opposite of lucky imaging. With conventional imaging in between, I am sure more options are better than less options.

Edited by FirstC8, 08 May 2016 - 07:45 PM.

[Jon Rista]

When I image at my dark site with 10-12 minute subs, just about every sub has an airplane trail, a meteor or a satellite trail in it. Many have more than one. My dark site is east of the Denver International Airport. The airport is quite a distance away, however there is still enough traffic overhead that I get a good deal of airplane trails in my subs. If I use 5 minute subs, I would say at least 50% of them have an intrusion.

 

If you were in a region devoid of air traffic, you might be able to get away with a 2 hour sub...but you are likely to still have some sat or meteor trails in it (they can be faint, but still visible.) 

 

I think one of the things that excites me about the ASI1600 and future CMOS cameras is you could get away with discarding more lesser quality subs, because you should be able to use shorter subs, possibly MUCH shorter. These days, I try to get over 50 subs regardless of what I am doing, because it takes about that much to fully take care of a large airplane trail. I am thinking in the future, I may simply discard those subs affected by a large intrusion. 

[Ken Sturrock]

Right. I'm excited that we might take shorter subs and fewer things can go wrong in that shorter time. I'm not concerned about the mount or tracking, but (being pretty ignorant about the subject) I wonder if seeing will molest the image noticably less with shorter exposures. Not exactly lucky imaging but maybe 150 seconds vs. 300 seconds.

 

It sure can't hurt.

[schmeah]

Yes, me thinks I would like to have one of these. However, I think I will wait until after the near hysterical excitement dies down, more real world images are produced, including those of faint narrowband emission nebulae, and a company like QSI decides to incorporate it into one of their packages so I don't have to deal with a new guider/OAG, filters/wheel, QA issues, etc etc. But perhaps at this price, the risk even now is not unfounded.

 

Derek

[FirstC8]

When I image at my dark site with 10-12 minute subs, just about every sub has an airplane trail, a meteor or a satellite trail in it. Many have more than one. My dark site is east of the Denver International Airport. The airport is quite a distance away, however there is still enough traffic overhead that I get a good deal of airplane trails in my subs. If I use 5 minute subs, I would say at least 50% of them have an intrusion.

If you were in a region devoid of air traffic, you might be able to get away with a 2 hour sub...but you are likely to still have some sat or meteor trails in it (they can be faint, but still visible.)

Or if the FOV is kept at a minimum. I live near two airports and regularly have flights overhead, yet never seen plane trails in my subs, nor meteor/satellite trails.

[Peter in Reno]

How about this one?

 

That's a single 10 minute sub taken with SXVR-M25C OSC camera and C-8 many years ago.

 

 

Peter

Edited by Peter in Reno, 08 May 2016 - 11:42 PM.

[FirstC8]

And this is that raw one hour long sub I was talking about, no stretching, no calibration, no darks no nothing, just one good old shot at M57 for an hour in H alpha, by a cheap 1/3 chip full of amp glows, vertical line patterns and other fixed noises, not to mention the light pollution and still bright moonlight in the background at the time.

It is not some far fetched dream that given the advances one day we can do this with most DSOs?

I'm looking forward to more results!

OK. Granted, it's not your usual DSO setup. A 16" Dob.

Pretty darn nice M51 here, 2000X1". The outer reaches and background galaxies qualify as "dim" by my lax standards. <grin>

http://www.astrokraa...l_Kraaikamp.jpg

Very nice!

Is it safe to say a 16" dob will not be a frequent visitor at a remote dark site?

[josh smith]

 

 

 

And this is that raw one hour long sub I was talking about, no stretching, no calibration, no darks no nothing, just one good old shot at M57 for an hour in H alpha, by a cheap 1/3 chip full of amp glows, vertical line patterns and other fixed noises, not to mention the light pollution and still bright moonlight in the background at the time.

It is not some far fetched dream that given the advances one day we can do this with most DSOs?

I'm looking forward to more results!

OK. Granted, it's not your usual DSO setup. A 16" Dob.

Pretty darn nice M51 here, 2000X1". The outer reaches and background galaxies qualify as "dim" by my lax standards. <grin>

http://www.astrokraa...l_Kraaikamp.jpg

Very nice!

Is it safe to say a 16" dob will not be a frequent visitor at a remote dark site?

 

 

Everything shown in that picture is still quite bright.  Have you imaged M51?  Again, it is an awesome picture, but doesn't seem any better than 2000 seconds worth of normal imaging at a dark site with a much smaller scope.  Also, ask Emil what he did to process this.  

 

Those cmos are the future.  I don't think the current one (asi 1600) is a risk.  It will perform well.  I don't see anything that says it makes regular imaging obsolete.  A good 16" dob and a good eq platform are a lot of money.  Then try to automate the focus and platesolving with that setup.  Deal with resetting the eq platform.  Ideally, you'll be able to reallocate money that didn't go towards a camera to more aperture and need less money in a mount.  Once and if these images can be calibrated properly, it will be a solid alternative, but this is venturing into lucky (high speed imaging) again.  I love the idea and the path, but it isn't replacing anything right now or making traditional imaging any less relevant.  It's another alternative with its own set of complications that would put you on the bleeding edge of figuring stuff out.

 

If you use this 1600 to image traditionally, it's going to be a good camera.  I'm not sure why everyone wants to convince you it's going to be something other than a good camera?   If it lives up to the specs, it will be an awesome bang for the buck.  I don't expect anything else.  I still don't want it instead of my camera.  It may be fun to get one to supplement it because it has more versatility and can do some things I can't.

[gregj888]

Groz,  Used to spook me too.  Turn the gain up to unity or more and set exposures for <75% fill, then stack.

 

On really bright objects the ~5e/ADU provides some smoothing...  May not be used a lot for AP, but could be otherwise.

 

FirstC6, HDR is kind of free with lots of subs, but could be used.  I think FC allows it with some cameras.  Not somthing I can think of a use for in AP, but others might.

 

cameras.

There's not a lot of reason to have more bits than it takes to encode the well depth at 1e/ADU.

 

This is true, but, there is a significant penalty for having less bits than it takes to enode the well depth.  20K well depth needs 20K points to describe, and 12 bits only gets you 4K points.  14 bits comes close to properly describing a 20K well depth.

 

[vdb]

 

 

 

Everything shown in that picture is still quite bright.  Have you imaged M51?  Again, it is an awesome picture, but doesn't seem any better than 2000 seconds worth of normal imaging at a dark site with a much smaller scope.  Also, ask Emil what he did to process this.  

 

Those cmos are the future.  I don't think the current one (asi 1600) is a risk.  It will perform well.  I don't see anything that says it makes regular imaging obsolete.  A good 16" dob and a good eq platform are a lot of money.  Then try to automate the focus and platesolving with that setup.  Deal with resetting the eq platform.  Ideally, you'll be able to reallocate money that didn't go towards a camera to more aperture and need less money in a mount.  Once and if these images can be calibrated properly, it will be a solid alternative, but this is venturing into lucky (high speed imaging) again.  I love the idea and the path, but it isn't replacing anything right now or making traditional imaging any less relevant.  It's another alternative with its own set of complications that would put you on the bleeding edge of figuring stuff out.

 

If you use this 1600 to image traditionally, it's going to be a good camera.  I'm not sure why everyone wants to convince you it's going to be something other than a good camera?   If it lives up to the specs, it will be an awesome bang for the buck.  I don't expect anything else.  I still don't want it instead of my camera.  It may be fun to get one to supplement it because it has more versatility and can do some things I can't.

 

 

http://www.astrokraa...l_Kraaikamp.jpg

 

For those who don't speak dutch, the image was taken in poor seeing conditions and the author is convinced that on better conditions there should be more detail ... Also it is just a little over 30 minutes of imaging time. Last night I did the same object, with RC at f5.4 and SXH814, I will compare ...

 

The first images I've seen is that it will perform as well as a traditional ccd when used as a traditional ccd, but for bright parts in PN or Galaxies it offers something extra ... at a cheaper price point.

Good to see the camera is available and Images are reaching the net. So far I'm convinced.

[blueman]

2000 subs, that is a lot of subs to have to process, even for a big computer. If the fits size is 32 megs like any other 16 meg chip, then you are talking 64,000,000,000 bytes total, that is 64 gigs.

Am I missing something here?

Blueman