Hi Fred,
If the camera performs as per specifications, the dark current will amount to only 20adu for a 40 minute sub.
There is a spreadsheet you can access on the web page I posted a link to (above) where you can plug all your numbers into. It will calculate your optimum sub exposure times.
My interest in your system is that I am curious to see how the 6303e performs using 3nm filters. In theory, this is one application where you want an ultra low noise camera (like an fli 694) Other ways to minimise the ratio of camera noise to shot noise would be to bin 2x2 and use a faster focal ratio.
Incidentally, I'm writing this on my phone so I can't use the spreadsheet... (lightning from the recent storms has fried my modem)
Hi Fred, my guestimate (from the manufacturers specs) is as follows. :
Read noise= 11e
Dark current @ 0C = 0.3e/s
Dark current doubling temperature = 6C
Therefore, @ -35C dark current= 0.00526 e/s
And for a 40m exposure, dark current= 12.6e
Dark noise is therefore 12.6^(0.5) = 3.55e (this is very good)
If the 40 min subs have a background level of 1000adu, and the camera gain is 1.47, the charge depth per pixel is 680e,
Therefore the shot noise is (680)^0.5= 26.1e
Total noise for the background per sub is therefore: (26.1^2 + 3.55^2 + 11^2)^0.5 = 28.5e
If you are using maxim, you will need to subtract 100adu from your baseline, in which case the shot noise value 24.1e and the total background noise will be 26.7e RMS per pixel.
Ergo, the sky background brightness is by far the dominant noise source in your system. That is mildly surprising considering you are using a 3nm bandpass filter @ f9.
Interesting. I used his calculator to determine that I can do subs in Lum on the STXL at 24.01 minutes for the RC. I am doing those at 30 minutes now so it is close, but perhaps an extra 15 minutes per hour would be better served getting another sub.
Do you think an 11002 sensor has better characteristics than the 11000? From what I have seen it is far noisier. My subs from the STL while cooled less were always cleaner than on the 11002. And; this has been a common comment from other users.
In any event it looks like long subs do work in a dark sky environment and negate the need to go shorter. Am I reading this right? Since his figure of 0.9 for F is arbitrary but looks reasonable. Thoughts?
Interesting. I used his calculator to determine that I can do subs in Lum on the STXL at 24.01 minutes for the RC. I am doing those at 30 minutes now so it is close, but perhaps an extra 15 minutes per hour would be better served getting another sub.
I think the spreadsheet works on the premise that for a given total exposure time the signal to noise minima is achieved with sub exposure lengths which reach the point of diminishing returns.
I take that to mean that it is better to have lots of good frames, rather than a few (slightly) better ones. The spreadsheet is an attempt to derive the best compromise, statistically. Obviously, anything that causes a variation in the ratio of signal electrons to readout noise electrons will have an influence on the optimum sub exposure length.
Implicitly, this includes filters and moon phase.
I'm not sure what happens in that spreadsheet calculation when you specify an infinite total integration time. (My computer is offline) Intuitively, I would assume it would be best to set sub exposure length to limit saturation.
Last edited by clive milne; 04-02-2015 at 04:02 PM.
Do you think an 11002 sensor has better characteristics than the 11000? From what I have seen it is far noisier. My subs from the STL while cooled less were always cleaner than on the 11002. And; this has been a common comment from other users.
I'm not sure....
In my discussion on the subject with Peter, iirc) he indicated that the STXL has much quiter electronics so the sins of the chip become obvious, where previously they were buried. I have never held one in my hands so can only defer to his judgement, or your judgement for that matter, being as you have a long history with both cameras.
Last edited by clive milne; 04-02-2015 at 05:02 PM.
In any event it looks like long subs do work in a dark sky environment and negate the need to go shorter. Am I reading this right? Since his figure of 0.9 for F is arbitrary but looks reasonable. Thoughts?
Yes, that is consistent with my understanding of the maths at least. Basically, you just need to expose long enough for the shot noise to be dominant.
Fred: when you say you have a 1000 ADU reading in the dark background areas are you talking about a sub which is raw, direct from the camera or one that has been calibrated? If you are talking about an uncalibrated sub what's the rough average ADU value in a bias frame from your camera?
I'd be really surprised if your 40 minute narrowband subs are shot noise limited. Mine certainly aren't with a KAF-16803 and 12" of aperture at f/9 under much brighter skies than yours.
Rick. As per post 20.
I calibrated a sub with 1000ADU back ground and now the background ADUs are 20-40. A bias sub has 800ADUs, roughly, it bounches round a lot over the view.
The moon was up, but far from the object.
Sorry Fred, I missed that second time through the thread
Quote:
Originally Posted by Bassnut
Thanks Clive, I was actually collecting all what you said above when I noticed your reply. Thanks for those calculations.
I calibrated a sub with 1000ADU back ground and now the background ADUs are 20-40. A bias sub has 800ADUs.
That's a little odd. I would have expected background ADUs after calibration to be around 180 ADU (1000 ADU less 800 ADU bias and 18.5 ADU dark current.) I'm not sure exactly how the Maxim pedestal works. Maybe that explains part of the discrepancy?
Quote:
Originally Posted by clive milne
Hi Fred, my guestimate (from the manufacturers specs) is as follows. :
Read noise= 11e
Dark current @ 0C = 0.3e/s
Dark current doubling temperature = 6C
Therefore, @ -35C dark current= 0.00526 e/s
And for a 40m exposure, dark current= 12.6e
Dark noise is therefore 12.6^(0.5) = 3.55e (this is very good)
If the 40 min subs have a background level of 1000adu, and the camera gain is 1.47, the charge depth per pixel is 680e,
Therefore the shot noise is (680)^0.5= 26.1e
Total noise for the background per sub is therefore: (26.1^2 + 3.55^2 + 11^2)^0.5 = 28.5e
If you are using maxim, you will need to subtract 100adu from your baseline, in which case the shot noise value 24.1e and the total background noise will be 26.7e RMS per pixel.
Ergo, the sky background brightness is by far the dominant noise source in your system. That is mildly surprising considering you are using a 3nm bandpass filter @ f9.
Clive: I think the charge depth per pixel above is way too big given Fred's additional info. If it is actually around 20-40 ADU (as per the calibrated sub) the the background shot noise is much smaller and read noise is dominant. Do you agree?
Sorry Fred, I missed that second time through the thread
That's a little odd. I would have expected background ADUs after calibration to be around 180 ADU (1000 ADU less 800 ADU bias and 18.5 ADU dark current.) I'm not sure exactly how the Maxim pedestal works. Maybe that explains part of the discrepancy?
Clive: I think the charge depth per pixel above is way too big given Fred's additional info. If it is actually around 20-40 ADU (as per the calibrated sub) the the background shot noise is much smaller and read noise is dominant. Do you agree?
Cheers,
Rick.
Hi Rick,
Yes i agree, my calculation is in error. I'm a little confused though by the calibrated frame background level being 20 - 40. I would have thought (as you have done) that with a bias of 800, and subs of 1000, the background should be 200 minus dark current = 180 (or 80 if the 100 adu maxim pedestal is present)
Anyway, irrespective of which value we take as being representative of the sky background (20, 40, 80 or 180) the camera noise is dominant by a large margin.
In some respects, this is good news.
I'm in exactly the same situation with even one hour narrowband subs with my 12" f/9 scope and KAF-16803 camera. An ICX-694 with much lower read noise would give me better SNR but I'd find it hard to give up the large FOV.
Hi Rick,
Yes i agree, my calculation is in error. I'm a little confused though by the calibrated frame background level being 20 - 40. I would have thought (as you have done) that with a bias of 800, and subs of 1000, the background should be 200 minus dark current = 180 (or 80 if the 100 adu maxim pedestal is present)
Anyway, irrespective of which value we take as being representative of the sky background (20, 40, 80 or 180) the camera noise is dominant by a large margin.
In some respects, this is good news.
regards
c
Sorry guys, I measured a bin 2 bias. Your right, a bin 1 bias is 1050 odd.
BTW, here is a more stretched (not processed much) version I didnt bother with, looked a bit messy.
Fred, that looks impressive, actually better than the original. .. the caveat being my internet access is reduced to my galaxy s4.
I'm a little confused, however, with your bin1 bias value. Perhaps it is a result of a typo or I am interpreting it incorrectly, but it seems inconceivable that a bias frame could have a mean background value greater than a 40min sub?
Fred, that looks impressive, actually better than the original. .. the caveat being my internet access is reduced to my galaxy s4.
I'm a little confused, however, with your bin1 bias value. Perhaps it is a result of a typo or I am interpreting it incorrectly, but it seems inconceivable that a bias frame could have a mean background value greater than a 40min sub?
best
C
Yeah, I thought that was odd. I measured the master bias now, same, 1050 odd. I looked again at a sub again and its background 1000 to 1030 odd and 20-60 calibrated. I dont use flats. The dark master is 1060 odd. Im not very bright, im just telling you what I measure. CCD stack scales cal with temp, I dont know if that makes a difference. Temp is always very close to -35deg though.
Ive wondered, CCD sensors "deteriorate" over time apparently. Im guessing cycling -35 to ambient say 20 deg average every day would be stressfull, 55deg delta at least. My cam for the last 2 years has always been below 0, 24/7 . -35deg imaging and -10 daytime idle. I wonder if that affects the CCD spec, long term?.
I have a significant amount of variation in my bias values over time, maybe 30 or 40 ADU. That's what lead me to do overscan calibration which makes everything wonderful
I'm in exactly the same situation with even one hour narrowband subs with my 12" f/9 scope and KAF-16803 camera. An ICX-694 with much lower read noise would give me better SNR but I'd find it hard to give up the large FOV.
With Fred's indulgence, I would like to submit two prescriptions optimised for narrow band imaging for your consideration.
If we stipulated a 46 arc minute fov (this is what you get with a 16803 and a 12" f9)
you could achieve this using an icx694 chip on a 12" F4.28 primary mirror with an ASA 0.73 reducer corrector.
Because each pixel in the Sony chip covers 1.5x1.5x the area of the sky (compared to the 16803 scenario) and has a 40% qe advantage, the shot level will be a little more than 3x greater (all be it at 0.667x the resolution) If you were to house the 694 in a microline body, the camera noise would be reduced to 1/3 that of a typical 16803.
The signal to noise ratio for narrow band imaging would be a factor of 10 better in the system using the Sony chip. It then follows that to achieve the same signal to noise ratio as a 100 hour exposure using a 16803, would require just 1 hour with a microline 694. The caveat being the resolution is reduced and this is only valid for 3nm bandpass imaging.
Prescription two is probably dream-ware for all but a handfull of people but is still do-able.
A 32" primary mirror is imho) the point of diminishing returns for a number of reasons that I don't need to go in to here.
If we specified an f4.5 primary, the Keller corrector would bring the focal length down to 105 inches (within a few percent of a 12"f9) so the native resolution would be equivalent If we binned the icx694 2x2.
The 32" /694 combination would be accumulating signal more than 10x faster than the 12" / 16803.
Now if you used a Raptor Photonics Kingfisher V icx694 camera which has cooling capacity of 110 degrees below ambient (0.000,001e- s) and a readout noise below 1.5,
the signal to noise ratio is 60 times better than the 16803.
What you could do in one night with this rig would take you 20,000 hours with your current equipment. Basically a lifetime.
This assumes that the background level for a 1 hour exposure at f9 through a 3nm bandpass filter is around 20 adu. If it is higher, then the shot noise will limit the s/r advantage of the second system.
You can perhaps see why I was interested in the sky background levels you are getting.
With Fred's indulgence, I would like to submit two prescriptions optimised for narrow band imaging for your consideration.
If we stipulated a 46 arc minute fov (this is what you get with a 16803 and a 12" f9)
you could achieve this using an icx694 chip on a 12" F4.28 primary mirror with an ASA 0.73 reducer corrector.
Because each pixel in the Sony chip covers 1.5x1.5x the area of the sky (compared to the 16803 scenario) and has a 40% qe advantage, the shot level will be a little more than 3x greater (all be it at 0.667x the resolution) If you were to house the 694 in a microline body, the camera noise would be reduced to 1/3 that of a typical 16803.
The signal to noise ratio for narrow band imaging would be a factor of 10 better in the system using the Sony chip. It then follows that to achieve the same signal to noise ratio as a 100 hour exposure using a 16803, would require just 1 hour with a microline 694. The caveat being the resolution is reduced and this is only valid for 3nm bandpass imaging.
Prescription two is probably dream-ware for all but a handfull of people but is still do-able.
A 32" primary mirror is imho) the point of diminishing returns for a number of reasons that I don't need to go in to here.
If we specified an f4.5 primary, the Keller corrector would bring the focal length down to 105 inches (within a few percent of a 12"f9) so the native resolution would be equivalent If we binned the icx694 2x2.
The 32" /694 combination would be accumulating signal more than 10x faster than the 12" / 16803.
Now if you used a Raptor Photonics Kingfisher V icx694 camera which has cooling capacity of 110 degrees below ambient (0.000,001e- s) and a readout noise below 1.5,
the signal to noise ratio is 60 times better than the 16803.
What you could do in one night with this rig would take you 20,000 hours with your current equipment. Basically a lifetime.
This assumes that the background level for a 1 hour exposure at f9 through a 3nm bandpass filter is around 20 adu. If it is higher, then the shot noise will limit the s/r advantage of the second system.
You can perhaps see why I was interested in the sky background levels you are getting.
best
c
Interesting thoughts, Clive. I wanted to run the numbers before I replied so it took me a little while.
If I was only interested in narrowband imaging it would certainly be worth investing in a camera with lower noise. I can run my scope at f/4.9 as well as f/9, so with a ICX694 that would be a very fast system.
The 32" system sounds very cool and would be fantastic for digging out faint planetary halos. That might be a project for my retirement. Hopefully EMCCDs will be affordable by then too
Interesting thoughts, Clive. I wanted to run the numbers before I replied so it took me a little while.
If I was only interested in narrowband imaging it would certainly be worth investing in a camera with lower noise. I can run my scope at f/4.9 as well as f/9, so with a ICX694 that would be a very fast system.
The 32" system sounds very cool and would be fantastic for digging out faint planetary halos. That might be a project for my retirement. Hopefully EMCCDs will be affordable by then too
Cheers,
Rick.
Hey mate,
I have been giving this a bit more thought over the last week or so and have come up with an extreme astrograph for the 694 chip which you could easily put together with off the shelf components.
Let's start with something like 25" to 32" F4.5 newtonian primary.
You would employ a field corrector that gives you a 40mm flat field with minimal vignetting (3" 1x Wynne would do)
Immediately after the Wynne, you would have your off axis guider + filter carousel. All pretty straightforward so far.
Rather than place your camera at the focal plane, let the light cone go through focus and start to diverge again.
The next element in line is a high quality 85mm camera lens. You would orient the lens essentially back to front, with the nominal focal plane of the lens coincident with that of the Wynne (does this need further clarification?)
Each light bundle exiting the 85mm lens will now be focused to infinity and be parallel. (This is an important enabler... but more on this later)
The parallel light bundles are then brought back to focus by the next element which is a 28mm f1.4 high quality lens.
Obviously the 85mm & 28mm lenses need to be oriented with their field lenses facing each other.
If the primary mirror chosen is a 32" f4.5, the resulting astrograph using these relay lenses will have a 1200mm focal length and an f ratio of f1.5
This system has a number of important advantages over conventional astrographs...
* Collimation is no more difficult than a typical large newtonian.
* If the camera is fitted with the appropriate bayonet fitting for the 28mm lens it should be orthoganal, so you avoid focal plane tilt issues typical of fast systems.
* Tilt and de-centre of the second relay lens with respect to the first does not degrade the psf at the final focal plane (within sensible limits)
* Because the filter carousel is at a location where the native focal ratio is f4.5, you don't suffer band pass shift with 3nm filters.
* This system gives you the opportunity of placing an ultra-narrowband filter (between the relay lenses) ... maybe an etalon? so you can bring the filter bandwidth down to 0.5nm This (theoretically) results in a reduction in the snr be a factor of 6x .... ie) it is 36x faster than a shot noise limited 3nm system (which doesn't actually exist)
.... There is more I could write but I am out of time so I'll finish with this comment;
I cannot even imagine what could be done with such an instrument. However, I do know that you would be breaking new ground with every exposure.
best
c
Last edited by clive milne; 19-02-2015 at 01:56 AM.
I just did a rough estimate of the beam angles in the afocal zone (between the relay lenses) and realised the system needed tweaking in order to make it suitable for an etalon in this location.
Relay lens 1 needs to have a 300mm focal length and a speed greater than f4.5 This brings the angle of incidence of the beams at the field edge down to 2 degrees and the transmission of the emission line should stay above 90%
The focal length of relay lens 2 is negotiable, with 85mm being the sensible bottom limit. This lens would work well: https://www.canon.com.au/en-AU/Perso...2L-II-USM-Lens
Final focal length for the 32" f4.5 system employing the 300mm/85mm relay lens arrangement would be 1020mm @ f1.275
The diameter of the etalon would be defined by the beam width at the afocal position, which is around 70mm.
The consistency of illumination accross the 694 chip would be excellent and unless you chose a really poor quality 300mm lens, the factor limiting the final resolution would be the sharpness of the 85mm lens. If you chose the Cannon 85/1.2, it would be pinpoint edge to edge with negligible lateral colour and vignetting.
One side benefit is that you could detune the etalon away from the Ha emission line and take 0.7nm continuum sub frames for later subtraction, or even pick up the N1 & maybe N2 line if you're lucky. (your pre-filter would need to be chosen to suit though)
Last edited by clive milne; 19-02-2015 at 04:09 PM.
