How many hours is "long enough"

[Shiraz]

There is a clear trend around here to take longer exposure images and some that have been recently posted demonstrate how effective long exposures can be. But then the question is, how long is "long enough".

Well, it seems that it depends ....

To demonstrate some of the issues, take 2 hypothetical but quite practical broadband systems that may be used by amateurs:
system 1: 12 inch f3.8 with a KAF3200ME based camera used under dark sky
system 2: 4 inch f7 with a small pixel (4.5micron) DSLR used under suburban sky (assume sky is about 2 mags brighter)

- system 1 has at least 3x the broadband quantum efficiency of system 2 by virtue of the Bayer filters,
- system 1 has ~8x the aperture area of system 2, assuming some central obstruction in system 1,
- both have about the same pixel scale, so there is no major sensitivity or resolution difference due to sampling,
- system 1 sky produces <1/2 the noise of system 2 sky

Combining all of these factors and assuming that system 1 will require additional time to gather colour data, yields the conclusion that system 1 will have more than 10x the signal level and less than 1/2 the noise of system 2 for a given exposure time. If system 1 needed 2 hours to image NGC 253 in colour, system 2 would need almost a week spent under the stars to produce a result with the same SNR - and this is for two quite practical systems.

two messages:
1. design your system very carefully if you want to minimise imaging time - system choices can greatly affect sensitivity and imaging time
2. different systems may require vastly different imaging times to reach the same signal to noise ratio on a given object, so there is no universally applicable yardstick as to how long is "long enough".

thanks for reading - appreciate any discussion. Regards ray

[RickS]

Ray,

An interesting discussion...

Another important, but less technical, parameter is the quality level desired by the imager. Three years ago I was happy with an hour of data from an OSC camera producing a recognizable image. Now I'm chasing features that need 30 minute subs with a 12" scope just to become faintly visible.

Cheers,
Rick.

[gregbradley]

Also a key factor is patience. Its easy to suffer from object-itis where you feel the need to image several different objects in a small amount of time!

Also sometimes you can invest time in mega hours on an object and the resulting image is really not worth the effort. Not all objects image well for your system, location, time of year etc etc.

But generally speaking the top images are always long exposure times.
The main difficulty is noise reduces by a much larger amount of exposure time.

To reduce noise by half and therefore double signal/noise requries something like 4X exposure time so its diminishing returns at some point.

Some objects are worth it though as Rolf's recent and Mike Sids earlier Cent A shows. Some of those shell structures and jets simply do not show up in under 15 -20 hours of imaging.

Same with Helix with its 2nd outer shell.

Greg.

[jjjnettie]

If you only get to use your gear one weekend a month, one has to take what one can get. It's a smorgasbord of delights up there and if all you can get is a nibble on each one, so be it.
There are only a few of us able to image the same object night after night to achieve the hours necessary to get a deep image.

[multiweb]

+1 . What she said :) You observatory yuppies forgot the smell of grass. :P

[naskies]

... and the automated observatory yuppies forget the bone chilling fear induced by hearing the pitter-patter of rain on the roof when you're warm in bed at 2:30 am thinking, "hmmm, is my telescope covered up?" :lol:

[multiweb]

Got plastic bags for that mate. ASCOM compliant. Better than max dome.:P

[cometcatcher]

System 3... Any telescope with unmodified DSLR trying to image in summer tropics between clouds with night temps @ 30C.

I've discovered it doesn't matter how long I image for under these circumstances. The DSLR just turns to a seething pile of noise at these temps. Pretty much limited to the Sun and Moon in summer.

[renormalised]

I don't even bother trying....if it's not raining, the sky is washed out and hazy.

[RobF]

After years gone by hunting things visually, I still get a real kick out a seeing single 5min sub showing more detail than I could EVER hope to see with a remotely affordable visual rig.

There's definitely no harm in running the sums on focal ratio, spectrum sensitivity, FOV etc. I suspect an ED80 and DSLR will be about 10 times cheaper than the 12" F3.8 CCD rig too though!? :)

[strongmanmike]

It all comes down to "is it worth it?"

If there is something faint you think you can reveal with a longer exposure, or as Fred shows us :rockband: you want to squeak the most out of your processing/sharpening, then long exposures is certainly the go, no question, sure it is harder the less automated and/or smaller your equipment (:whistle:)...but thinking that long exposures should be the norm and if you aren't doing them then somehow you aren't producing something good...is quite frankly, missing the point of our hobby and as Ray has shown, plucking arbitrary exposure times of 20 or 30hrs out of thin air is kinda meaningless really in the scheme of things...that's my take anyway :)

Mike

[Shiraz]

thanks for the thoughtful posts and interesting discussion.

Looking back it seems that the original post was bit naive to ignore user requirements and expectations, as well as weather limitations and cost. However, the general thrust still seems to be backed up by the later posts - there really can be no prescriptive requirement on the length of exposures, for a whole gaggle of reasons. I wasn't trying to knock any systems, just to point out how radically different they could be in sensitivity.

Must say that I personally enjoy watching subs as they come in - nice to be part of the acquisition process. I can see how automation would allow many more/dimmer targets to be imaged though - I guess it all depends on what each of us wants from the hobby :).

[strongmanmike]

Yeah I enjoy that too, although I do leave the scope for extended periods throughout the night...but even more, when the seeing is behaving, in Astroart I looove watching the guide star centroid history plot in real time make a beautiful tight round cluster of dots for hours on end :D...something about that really turns me on :evil2:

Mike

[avandonk]

Starizona have a quick guide on optimum exposure for any CCD system.

Here

http://starizona.com/acb/ccd/advtheoryexp.aspx

They have an ideal exposure calculator here

http://starizona.com/acb/ccd/calc_ideal.aspx

All you need to input is your camera and sky background for a given exposure.

It is also worthwhile if you want to get into the mathematics to have a read of the info on Stan Moore's and John Smith's sites.

Stan Moore here
http://www.stanmooreastro.com/

John Smith here

http://www.hiddenloft.com/notes.htm

and here

http://www.hiddenloft.com/notes/acq.htm


The basic take home message is to take 2N+1 exposures that are half or a third of the exposure needed for sky background to overwhelm read noise.

With 3nm narrow band this of course allows for very long exposures. Then the inherent camera thermal and random noise become a problem. That is why a faster optic is preferable.

Bert

[Shiraz]

these are some of the best references out there Bert - thanks.

the ideal exposure calculators provide "optimum" sub exposure time (read noise = 5% total), but not the overall exposure needed for a given object. Do you have any ideas on what overall SNR is required to get a "good" image?

[lazjen]

That's interesting information. Is there a way to work out the details for DSLR cameras as well (it seemed CCD centric for the Starizona articles)?

[RickS]

An alternative view on subframe lengths, optimal number of subframes, etc. (Warning: lots of mathematics):
http://www.cloudynights.com/item.php?item_id=1622

[avandonk]

It is almost impossible to have a guide for DSLR's as the thermal noise varies so much with ambient temperature.

I used to have a Canon 5DH and a 300mm lens at F3.6 which I cooled with a Peltier fridge to -12C. This reduced the thermal noise considerably but not quite enough to image dim objects such as the Vela Super Nova remnant or the dim nebulosity in the LMC and SMC in NB as clearly as I wanted in a reasonable time. The Bayer filter array is the major reason for this.

Bert

[avandonk]

It purely depends on how 'bright' the target is. A low signal to noise will record bright stars in seconds. As objects get dimmer the signal to noise needs to be far higher. This is so that the dim signal is larger than the noise. If it is not, it will never be recorded!

If the signal is just above the noise then the integrated sampling time needs to be very long to improve the ratio of S/N.

As in all of science it is the signal to noise ratio that is the only dogma for evaluating data.

We must never forget though that one human's noise can be another human's signal. Even then the measuring system must differentiate between the two. I changed man to human for Jocelyn Bell who really was the 'man' to discover pulsars and was ignored in the Nobel Prize!

Bert

[Manav]

nice read and some very useful info. I need to point out though, a while back I read somewhere if one can't get long exposures then aim for multiple short ones. Not sure how effective this would be....

[lazjen]

Ah, thanks for that - I thought it might be something like this. Ok, I guess I'll have to keep experimenting for my system.

[swannies1983]

This is my understanding but there is still quite a bit of debate on this. Scott Rosen (www.astronomersdoitinthedark.com/) has produced some of the best DSLR images. He tends to do short subs, sometimes only 3 mins, but collects hours of data (10+).

[Shiraz]

provided the chip temperature does not wander around, the only SNR difference between lots of short subs and a single equivalent long one is that each read of a short sub introduces a burst of read noise. If you have a low read noise camera, short subs are entirely practical, since the overall read noise from lots of them will still end up fairly low. DSLRs will generally have fairly low read noise, so, even though they have low QE, short subs could be applicable for some targets and sky conditions. The trick is to keep the total read noise well below the shot noise and thermal noise. Short subs are not applicable to cameras with high read noise, or if doing narrow band imaging.

In the planetary imaging area, subs as short as 1/100 sec are routinely used on bright targets and an image may be assembled from many thousands of them. There is nothing inherently wrong with stacking huge numbers of short subs - the random nature of noise ensures that the signal will build up out of the noise in exactly the same way as it does with a long single exposure, provided the read noise does not intrude. It is fascinating to stack a hundred plus subs that show nothing but stars and a fairly flat noise background and find that you have a nice nebula standing out clearly in the resulting image.

[Paul Haese]

18 months ago I had a conversation with Martin Pugh about how much integration was needed for any particular image. His opinion has proven right as far as I am concerned. He said nothing under 20 hours. 30 is better. Also image subs longer to go deeper.

I sort of knew this from taking long subs several years ago for many of my deep sky images. For some reason I started following what many do here with 10 minute subs. Now with a narrow aperture this just does not work. It resulted in images with lots of integration but not smooth. I had to stretch the data too much to get the brightness that I wanted and that resulted in lots of uncontrollable noise. Longer subs gives me a better signal to noise and a background ADU over 1000. If you have a wider aperture with a fast f ratio you can use shorter subs to achieve the same result. The point being you need to experiment a little to find what length sub produces low noise with high signal and that will mean it will be different for many objects as each has varying dimness.

However, back to the central issue, is that longer integration time means smoother images. I have found colour data has to have a minimum of 3 hours in each filter, so I don't have to use smoothing. 4 hours works really well. Add the lum or narrow band of 10 hours and you are in the 20 hour mark.

Bottom line for me is the quality I am trying to achieve. Going automated has helped to gather lots of data. So I will stick to what Martin said and aim for that each time and try to be patient.

[gregbradley]

I'm largely with you Paul. Definitely longer is better with both total exposure and subs. But of course practicality kicks in and clouds, bad weather, moon, work all conspire to reduce total available time. So that is the goal to take long exposures but sometimes you've got to take what you can get.

Also some bright objects don't seem to gain too much from longer exposure times or with small well cameras a 15/20 minute sub even 30 minutes may blow out highlights.

So its not a rote rule. Some judgement is involved in my opinion.

Greg.

[avandonk]

Both Paul and Greg are quite correct. But first you must evaluate whether collecting mega data is worth it. The only real criteria is that your signal to noise of the dimmest bits of the image you wish to show or record is above at least one or far better, two sigma.

When I was working out what to get for my 'new' system I did not compromise anywhere. I wanted to do very deep NB of dim objects at a moderately wide field. I could most probably spend the rest of my life trying to get the detail in the Vela SNR and the two Magellanic Galaxies and still not get it anywhere near perfect.

If you take as a starting point a very fast optic then the best quality largest 3nm NB filters are a no brainer. All else also follows e.g. a very low noise camera and best focuser available.

With 16 minute exposures with any 3nm NB filter the signal to noise is still improving even after thirty exposures. I would say that double this (60 exposures or 16 hours) is still worth doing.

In a target rich place such as the LMC it is surprising what pops up.

This is about twenty four hours of total exposure. 9MB


http://d1355990.i49.quadrahosting.com.au/2013_10/NGC2070_NB_HP_N.jpg

Bert

[Shiraz]

no argument that longer is better (see original post).

I was trying to point out that there is no clear correspondence in imaging time between systems. For example, I imagine that Bert's system will quite comfortably produce an image of the outer parts of Helix in 30 minutes - mine takes a few hours and a DSLR based system could take days - there is no correspondence in imaging time between these systems.

You may decide on exposure times for a given system and for your expectations of quality, but there is no point in someone else with a different system trying to use that as a guide. Even within a given system, you can change the required imaging time by huge amounts - eg, if you decide to 2x2 bin your data, you can reduce your imaging time to 1/4 that required at maximum resolution.

[rustigsmed]

how about a situation with a OSC vs mono filtered /RGB imaging?

Would you say that the OSC, would need say a third of the time, 2/3s? or doesn't it work that way and to get a similar quality image it would take about the same amount of time?

cheers

[Shiraz]

Hi Russell
Not knocking OSCs, they certainly have significant advantages. However, they do start out behind the 8 ball in sensitivity, since fewer than 1/3 of the photons falling on the chip actually make it through to the pixels - the rest are absorbed by the Bayer filters. For this reason, you will probably need to image a bit more than 3 times as long with an OSC to get the same luminance result as with an equivalent mono camera (luminance is synthesised in the case of the OSC). To add lower resolution colour data to the mono camera image, you will need to image maybe as long again (in RGB) as for luminance, so overall, the OSC will take about 1.5 times as long as an equivalent mono camera with LRGB filters for roughly the same result.

The situation is exacerbated if the OSC is uncooled, since thermal noise becomes a big issue if the signal levels are inherently low.

[rustigsmed]

thanks for the info Ray :thumbsup: thats very interesting, and something for people to consider in whether they are going OSC or mono w filters.

[gregbradley]

Ray I don't think the difference between mono and osc is 3X the time. That's because of interpolation. Where the firmware calculates likely values for a pixel based on its surrounding neighbours. In a sense this is similar to binning.

I would estimate the difference between a mono and a osc CCD to be more like 50% or even less. I have had an STL11 colour and it was quite sensitive. Where it had trouble was in the dim areas or dusty areas. They often were noisy.

Same with DSLRs. Some DSLRs have very high QE compared to CCDs. Nikon D800E Soony Exmor 36.3mp QE is supposed to be 59%. That is the same as my FLI Microline 8300. Having used both that seems about right.

Greg.

[Shiraz]

Hi greg

OSC makers normally quote peak QE at the maximum of the green Bayer filter - ie if the chip is illuminated with narrow band green light of just the right wavelength, 59% of the photons will be detected by those pixels that are under green filters. What they do not say is that 0% of that green light will be detected by either the red or blue pixels - it won't even get to them. So the QE claims for OSCs are not remotely apples to apples when comparing with figures from mono chips - QE data will only be comparable at one chosen wavelength and then for only half of the pixels in the OSC.

The Bayer filters do not redirect photons, they absorb them, so if you illuminate the chip with white light, the Bayer filters will stop the red and blue light from getting to the green pixels, the red and green light from the blue pixels and the blue and green light from the red pixels - ie, about 2/3 of the light is absorbed before it gets to any one pixel. If you remove the filters, each pixel will now see the full red+green+blue of the white light, so it will have ~3x as many photons pouring into it - this is a mono chip. The mono chip will be much more sensitive to broadband light (~3x), simply because it sees so many more photons.

The Bayer matrix is limited in fundamental resolution to that of the 2x2 pixel group that make up a Bayer cell - ie Bayer chips have lower resolution than mono chips. The interpolation process attempts to estimate what the missing data might have been to try to recover some of the lost resolution and it is generally very successful in doing so. However, it does not increase sensitivity - that was irrevocably lost when the Bayer filters removed 2/3 of the photons.

OSC chips seem to have higher internal gain than mono chips so that you see the same sorts of signal levels when you switch between mono and OSC. However, the OSC is giving you the same signal from a lot fewer photons by running at higher gain and that will show up as noise in darker regions of the image - which is what you have observed.

rest assured, your 8300 will eat any OSC for breakfast when it comes to broadband SNR. That doesn't mean that OSCs are no good - far from it, they have many advantages, but they are definitely less sensitive than mono chips. regards ray


EDIT: just realised that I have both colour and mono versions of the original QHY5. Its hardly a top line chip, but I can see vastly more guide stars with the mono version - 3x more sensitivity would be about right.

[gregbradley]

Good writeup Ray.

I think you are right about the extra gain making up for a lack.

Don't worry I am not planning on OSC. Been there done that. They are great for brighter targets but struggle with dim objects and narrowband.

The Kodak True Sense LRGB colour filter array claims 2X increase in light sensitivity so perhaps 3X is true between mono and colour. I always thought it was less but the gain factor may be what I was confusing with QE. Like high ISO on a DSLR.

Greg.

[renormalised]

Most people would go with an OSC simply because it's a lot more convenient and simpler to run than having to contend with filter wheels, filters etc. Less to setup and still takes good piccies. A mono camera, filter wheel, filters etc etc, is a big investment. Great for a permanent setup, but I feel can be somewhat of an overkill if you want to just haul a light but good setup around, to use visually and also take piccies with. In any case, taking pics through filters reduces the sensitivity and resolution of your mono camera as well.

[RickS]

Taking pics with a mono camera through filters does reduce the sensitivity, but not the resolution. The big advantage is the ability to do luminance frames and make use of all the photons in the visible range.

I am a nomadic imager who started with an OSC and moved to a mono camera very quickly. The better quality is worth the extra time and effort to me. I understand that others may make different choices...

[nebulosity.]

Thanks Ray for starting this thread and for everyones informative comments, I've learnt a heap :thumbsup:

Cheers
Jo

[naskies]

I agree with Ray's original point that total integration times do not directly correlate with the same SnR across different systems and locations. A 12" scope has over twice the aperture (area) than my 8"... so all else being equal, 20 hours for someone else may be similar to 45 hours for me.

However, if you look at the subset of people who actively aim to go really deeeep, produce images that elicits "wow" reactions, gets APODs, and so on, things start to look very similar: mag 21-22 dark skies, 100 - 130 mm refractors for nebulae, 12" reflectors with appropriately matched pixel sizes for galaxies, Astrodon/Astronomik/Baader/etc filters, 11000 / 16803 / 694 chips are all pretty common.

In this case, comparing 1 vs 10 vs 50 vs 100 hours does give a rough idea of what to expect.

[multiweb]

I can relate to this. I always take my OSC with me when I drive out to dark skies once in a blue moon. At home I'm stuck doing NB with a mono. I would never contemplate a filter wheel and doing color on the field. Not enough time for this.

[Shiraz]

Hi Jo - I agree, there have been some very thoughtful and valuable contributions.



Interesting observation Dave, maybe my original thesis is not applicable within subsets of the astro community - Darwin rules, even in Astro imaging? Even so, there is a 2:1 range of imaging times inherent in the QEs of the chips you mentioned.





and with an OSC, you never get stuck with 3.5 hours of luminance, 21 minutes of red and clouds for the next week - followed by the moon.

[avandonk]

Yep there is nothing wrong with DSLR's or OSC cameras. I used to use hypered film in the 1970's and a DSLR would have been far better, heaven in fact!

This is a twelve panel mosaic of a blend of NB and RGB done with my Canon 5DH and 300mm F2.8L lens at f/3.6. The camera was cooled to -12C. 20 MB

http://d1355990.i49.quadrahosting.com.au/2013_05/VELASNRMOS.jpg

I did this some years ago. I was pleased at the time but longer or more subs would not improve the image much at all. I had hit the inherent signal to noise wall of the total system.

When Vela rises I will attempt to do the same mosaic with the new system. Here is a thread on just one panel.

http://www.iceinspace.com.au/forum/showthread.php?t=106569&highlight=vela

Bert

[renormalised]

Not only that, but if you've got a hyperstar, like you have Marc, and you want colour piccies, the OSC is the only way to do it. You could take your mono along too and shoot lum or Ha shots, if you wanted to.

But like I said, lugging along filter wheels and all the other extraneous stuff that can go south on you if it decides to just have a night off doesn't make for a pleasant imaging experience or a good night of astronomy.

[renormalised]

Here's an interesting point...we've been talking about the advantages of various camera types and scope combination, integration times for optimal photos etc etc, but what would you think the optimal scope/camera combination would be for each area of astropics? You could do a Ken Crawford, Rob Gendler etc and have scopes in the 24-32" range with 16803 chipped cameras and that would be great. But they're not the best for nebs and other large targets since you need a wide field to take in most, like the NA neb or Rosette. Not only that, going large isn't always the best, since you have to contend with other problems that become magnified due to the larger aperture, such as scintillation, air currents etc. If you're looking for a system that will give you maximum details in your piccies for aperture and bang for buck, where do you think such a scope would sit? It most likely means you'll need two scopes, one for general shots and one for wide field, but there should be a "sweet spot" for each type. Where do you think that lies?

[multiweb]

I haven't used that many camera/scopes combinations but I have used OSC and mono on short to very short FL. My OSC with large pixels will fall apart with 100/200mm lenses. Even with an image scale of 2-3 seconds of arc per pixel. The debayering process just hacks the data to pieces. The only reasonable channel is the green channel because it has two pixels instead of one. The mono shines in the same situation.

But I was pleasantly surprised when I used the same OSC on my 8" F/4. The image scale was closer to 1-1.2 seconds of arc per pixel, better sampling and the RGB channels were balanced. I couldn't really tell visually of any significant loss of details. Of course the mono in NB was sharper but not using straight LUM due to seeing conditions which then become an important factor.

[gregbradley]

If you're looking for a system that will give you maximum details in your piccies for aperture and bang for buck, where do you think such a scope would sit? It most likely means you'll need two scopes, one for general shots and one for wide field, but there should be a "sweet spot" for each type. Where do you think that lies?[/QUOTE]


I agree Carl. One scope won't fit all. A widefield high quality scope around 4 inches is a proven formula like FSQ106ED, I like my TEC110 fluorite for this. There are others. Pentax 125 was one. Televue NP101 and 127i.

Then refractor sweet spot is around 6 inches. My TEC180 is also close to the sweet spot in that its the largest refractor (perhaps a 200mm is still fine) that is still portable. Above that and you probably start needing a PME to handle or an AP1200. So 6 to 8 inches of APO.

Then reflectors. There are lots of great images from the 8 to 12 inch range. 12 inch compound of any type is starting to get large. A 12 inch RCOS is too large to handle easily although my closed tube 12 inch RCOS was still "portable".

Above 12 inches and I think you are looking at a permanent installation unless you are really strong and have a large cargo space in your vehicle and a large powerful mount to handle it all.

So 4 inch refractor for widefield, 6-8 inch APO for more detailed clear round star no diffraction spikes work and 8-12 inch compound scope for bright enough aperture to capture many galaxies and dim objects with fine detail and still be portable.

Then add to that the question of F ratio and flexibility of the imaging system.

For example Marcus Davies used a very nice Tak TOA150 APO. Tak makes an extender to take it up to F12+ for some magnification on small galaxies and a reducer to bring it down to F5. something. So its very flexible.

Faster scopes need more careful handling so best for more experienced scope handlers who don't mind the careful adjustments, fussy collimation, flexure issues. So perhaps not in the sweet spot.

I always consider F5 to be a sweet spot for F ratio. You can still get a rigid system that is fast at a nice dark site with a good camera that is still easy to handle and super fussy to flexure and miscollimation etc etc.

So if I can get F5 in my scopes or close to it I am happy and how well F5 works in the images.

For example my TEC110 is F5.6 -it could be a tad faster but then a Proline camera is very heavy so no thanks. My AstroPhysics 140 has a reducer that makes it around F5.6 as well and that is very nice for that scope.

My TEC180 though at F5.25 was a bit ho-hum. But an adapter to use the AP reducer on the TEC could be spectacular as Yuri does not make reducers for TEC scopes. He doesn't believe in them.

Greg.

[leinad]

Interesting discussion, one that seems to pop up again now and then.

The other discussion might be diminishing returns as well as your cameras maximum dynamic range limits (ADU limits)that determines which method suits and 'how many hours is long enough'...

I might have my understanding wrong(learning continues), but I couldn't see how multiple short exposures amounted to the same results as a single long exposure method, it just didn't make sense to me unless the light gathering conditions could be met with both instances where the same sum of signal brightness is gathered.

But if you're averaging you can only have so much signal to weed out of the noise in the end?

I see it that The count total of short exposures you'd want to take could work two ways, one would be using a light bucket with a low end camera, your ADU might blow out so you take say a 5min exposure, then you take multiples, as many as you want but your only improving the SNR of limited range of signal over the noise.

Then you use a high end camera with a huge well depth, as long as the tracking mechanics of your equipment allows you to with good accuracy, you could image for an hour. Here your SNR also is improved and you're seeing an increase of the signal brightness over time, and this can also be improved more by improving the SNR again using multiple images and averaging them.

But from both scenarios depending on how long you expose for with the cameras well depth and the telescopes ability to gather photons slow or fast, every instance has a point of disminishing return that is limited by your camera and telescope.

So whether it's 20hours of short 5min exposure stacks, or 20hours of 60min stacks, they can never be equaled.
And considering that, what you decide on in the end has a diminishing return value of how much signal you capture, it's just the method you decide on to get there and how dark a site you have that can capture light fast than noisy light polluted sites.

Edit: I see you guys have touched on this already at this first posts :)
My bad?

[Shiraz]

Following up on Dave's idea, I did some modelling of a few typical hi res galaxy imaging systems with the attached results. The starting point was a Mag23 galaxy against a Mag21 sky (per arc sec2) and 0.8 atmosphere transparency. The output is time taken to reach 20dB SNR, the ISO standard for "adequate" image quality - better quality will require longer. The optimum sub is based on the Starizona 5% read noise criterion - it is the shortest sub where shot noise will still overwhelm read noise and you could use longer if you wish. Of course, this only applies to visible broadband and the optimum sub length for any system in narrowband is "as long as possible".

there is lots of variability even in this smallish group - one system needs 6.5 hours , another needs 1.5 hours to get the same SNR. Sub length varies from 3 minutes to 40 minutes - and this is for the same environment and task. Actual times could be a bit different due to initial assumptions, but nothing like enough to account for such a large spread in imaging time. Seems that, even within this subset of imagers, there is no such thing as a universal best imaging time or an optimum sub length - it strongly depends on the system design. The most significant factor is the pixel scale (imaging time will depend on 1/pixelscale^2), but QE is also very important and read noise is a major factor in determining the sub length. regards Ray

[RobF]

This evidence-based discussion is like a breath of fresh air. My perception of medicine and pathology for the last 30 odd years is that (often younger) specialists are now almost embarassed to express an opinion without being able to reference evidence in peer-reviewed literature. Very different to many years ago when senior specialists and their empirical knowledge were unquestioned.

Opinions from experienced astrophotographers will always be important, but evidence and sold mathematics are great to see, especially as the number of options in range of amateurs continues to expand. And the statement that there's never one perfect 'scope for everyone has never been more true.

I'm just embarassed I've never looked at a decent spreadsheet for exposure times, although I have read up a bit. Must get on with it!

[naskies]

Thanks for that, Ray; very interesting. I think it shows that an appropriate image scale is critical for decent sensitivity (as we all know from painful experience or learning from others).

Ignoring the clear case of poorly matched scope/camera (KAF-8300M on a 12-inch RC...), there's not *that* much difference in total integration time amongst the other contenders - a constant factor, rather than orders of magnitude.

Object selection, such as Orion Nebula versus a mag 10 galaxy with low surface brightness, will make more of a difference. However, I'm guessing that some of those scope combinations will have a much deeper limiting magnitude for the same integration/sub time.

[Shiraz]

My understanding of why summed short subs and one long exposure can be the same is that the primary broadband noise is due to shot noise - that comes as part of the signal-plus-background and you get the same number of photons (target, background and noise) in a given period however you choose to collect them. This argument falls over if you have a camera with high read noise - then the read noise can add to the overall noise and make the SNR worse for multiple short subs. However, it is often still possible to use relatively short subs and still keep the total read noise contribution below the level where it makes any real difference.



Thanks Rob. I was surprised to find that there was no readily available design methodology for putting together an imaging system. Suck it and see will eventually give a result, but you can waste a lot of time and money getting there in this hobby. A decent model is a lot less costly and can provide insight that is just not available from empirical considerations. There is a place for the wisdom of the elders, but I am personally too old to wait around for understanding to come via that route.



Yes you are right Dave (crikey that sounds like HAL from 2001....) - there is not such a large difference between the reasonably well sampled systems - image scale sure is critical.

I am not sure where the concept of limiting magnitude fits in, since it seems to generally be applied to unresolved objects and all of the modelling to date has been on extended objects - will have to do a bit of reading.

Regards ray

[avandonk]

With my new system I took a leap into the 'unknown' only because it had not been done.

My background is in Physics and optics and over forty years in scientific research.

I did calculate all aspects of every part of my new system. My solution was a unique solution because it was correct. For me!

This is not a recipe for anyone else's system.

What is the strength of us amateurs, is that no matter where you look at the sky the chances are that the pro's are not looking at your bit of sky!

Evidence based decision making rules!

Bert

[Shiraz]

thanks Bert. Sounds like we have a very similar background, even down to the how long bit.
When you designed your quite unique system, what tools did you use? - seemed to me that there was nothing much out there.

[gregbradley]

[QUOTE=Shiraz;1028469]My understanding of why summed short subs and one long exposure can be the same is that the primary broadband noise is due to shot noise - that comes as part of the signal-plus-background and you get the same number of photons (target, background and noise) in a given period however you choose to collect them. This argument falls over if you have a camera with high read noise - then the read noise can add to the overall noise and make the SNR worse for multiple short subs. However, it is often still possible to use relatively short subs and still keep the total read noise contribution below the level where it makes any real difference.


Something is not right here. The top imagers are going for longer exposures and getting better results. If your tracking can handle it I have noticed for example you start getting dimmer, fainter details when using 15 minute subs versus 10 minutes.

Don Goldman R Jay Gabany both mentioned they are now using 30 minute subs for their imaging.

John Gleason of Ha fame uses 40 minute subs on his FSQ and Ha imaging.

The usual barrier is tracking. Can your setup handle 30 minute subs and still get tight round stars or do they become elongated or even just bloated from the bouncing around from corrections and periodic error?

I plan on going longer (I currently image at 10 or 15 minutes for LRGB and 20 minutes for narrowband) as it seems to me the really faint parts have to get above the noise floor of the system. It has the best chance of being imaged if you go longer if your skies permit it (otherwise you will get an excellent bright image of the light pollution!). I intend tweaking PEC and Polar Alignment and getting large T-point models to make long exposure with tight round stars possible every time.

Greg.

[Merlin66]

All this discussion on SNR.....
Actually how DO you guys actually measure the SNR in the images you take????
(I have to use some processing software to determine the SNR in our spectra..not easy.)

[Shiraz]

[QUOTE=gregbradley;1029014]

thanks Greg - you have proved my point. I have no desire to second guess DG or RJG, but put their systems into the system model with very dark sky (21.6). The suggestion for Don's system is for 27min subs and for RJay, 21 minute subs. Since I made a number of unknowable assumptions, I think that this is pretty good validation of the model. All along, the suggestion for NB has been "as long as possible", so JG's use of 40 minutes is also some sort of validation.

The point of the original post was that just because someone like DG uses 30 minute subs, there is no valid for anyone else to immediately change over to 30 minute subs (unless they have a similar system) - sub length should be individually chosen for the characteristics of the scope, camera and sky. For example, Mike S's system under the same conditions does not need subs longer than 3 minutes, even though Mike uses longer ones - he would be throwing away a lot of dynamic range if he used 30 minute subs. Your CDK with the 16803 needs subs of at least 15 minutes under very dark sky. There is no single sub length that is required if you want quality images and there may even be some very good reasons for using shorter subs than recommended by the maths to overcome seeing variability over a scale of a few to tens of minutes.



Trying to measure SNR for a single complex image is even less easy - it is not possible to separate out structure and noise. In assessing equipment, have used a flat, removed bias and then selected by eye a region where there is no obvious gradient. then used Nebulosity to get the stats for a 21x21 region in the selected part of the image.

ISO has standards/measurement techniques for image quality/SNR. http://en.wikipedia.org/wiki/Signal-to-noise_ratio_(imaging) might be worth a read.

[Merlin66]

Ray,
WIKI repeats the theoretical methodology....
Do programs like Nebulosity/ Maxim etc actually give a SNR for a nominated area??
I'm just interested as AP's seems to take this into acount in their analysis but I haven't seen the data.....

[Shiraz]

Neb gives mean and SD, so yes it provides SNR for a flat image. The only object I know of that can be used to get an estimate of SNR in a real image is Helix - parts of that are very flat. Or you could use a bit of featureless sky background if you can find one. Otherwise, it is only reasonable to measure system SNR with controlled images.

[Merlin66]

Ray,
""it is only reasonable to measure system SNR with controlled images""
Is this what AP's actually do???
I'm tending to think SNR is very subjective and not rigorously applied....

[Shiraz]

I have done so when validating imaging system models, can't speak for others.

SNR is of no use in assessing real images and is never used that way as far as I know - it is used for doing comparative assesment of equipment and imaging technique options - where it can be applied rigorously.

[avandonk]

There is only one criteria, maximise the signal to noise without sacrificing too much spatial information! For the last thirty years of my working life I collected x-ray diffraction data from protein crystals with very large unit cells. Most of the experimental innovative effort in our lab was to maximise the signal to noise of our data. Our home laboratory system had a better signal to noise than many/all synchrotron based systems. The protein beam line at Melbourne's synchrotron uses many of our concepts and is one of the best on the planet.


Bert

[swannies1983]

Further to my earlier post, here are some posts made by Scott Rosen in a Yahoo group I'm part of....

and....

but....

[gregbradley]

Dan, I think that discussion was about DSLRs not astro CCDs. The main difference being 9 micron pixelled CCDs have 100,000 electron full well capacity (how many electrons the pixel can hold before spilling over).

DSLRs well capacity would be less. Also they are quite a bit less sensitive than the CCD's which are usually 50% at the low end and up to 77% or more at the high end for CCDs. Also having a colour filter array this makes them even less sensitive.

ISO simply is an amplifier of the signal. Its like turning the sound up on your stereo. It does not add anything merely magnify what is there.

The whole point there is getting the faint details above the noise of the camera. At what point that occurs is the question. From my experience of imaging faint objects hundreds of times, that will not occur in 45 second-4 minute subs with a basically not very sensitive DSLR.

The M101 example he posted proves the point. There is a lot more faint detail missing from that image. Its impressive from the standpoint of a DSLR captured image but not impressive compared to a reasonable CCD image of similar duration from an 8 inch scope. It looks black clipped as he mentions it probably was.

DSLRs are great for bright objects but the dim ones they are unsuitable because of the above.

I definitely see a gain in 15 minutes over 10 minute subs. Its subtle but its there. Tracking has to be good though as elongated stars are worse than not getting some tiny faint detail.

Greg.

[swannies1983]

All good points but the OP didn't specify this as a "CCD only" discussion ;). In fact, there's another thread (http://www.iceinspace.com.au/forum/showthread.php?t=113727) on the first page purely discussing CCD and exposure length.

Obviously, if you compare CCD vs DSLR, CCD will win hands down. But the advantages of long subs over short subs is still up for debate.



I disagree. You only have to look at Scott's work to appreciate what can be achieved with a modded dslr under dark skies. Just check out some of his work

M81, M82 and the Integrated Flux Nebula (http://www.astronomersdoitinthedark.com/index.php?c=151&p=480)(ok, ten minute subs for this one).

Witch head (http://www.astronomersdoitinthedark.com/index.php?c=151&p=519)(5min subs @ ISO800)

Leo Triplet, including tidal stream of NGC 3628 (http://www.astronomersdoitinthedark.com/index.php?c=113&p=501)(only 3 min subs for this galaxy)

Whirlpool Galaxy (http://www.astronomersdoitinthedark.com/index.php?c=113&p=396)(5 minute subs)

He has many more examples of what can be achieved using short subs with a dslr. Granted, the amount of hours he puts in to get these images is much more than one would do with a CCD camera, but some of his images easily surpass CCD images.

[gregbradley]

Yes that guy is hardcore. 49 hours exposure - wow.

I saw the name Roger Clark and I know he does a lot of DSLR testing and reports on their performance. I also read some of it and they were talking about the histogram which is more of a DSLR imaging technique.

I agree you can take some great images with a DSLR. Its just harder work on the dim objects and the difficult part with a DSLR is to maintain star colours. They often wash out to white in DSLR images. Even those you linked the star colours are off.

The basic theory is the signal has to get above the noise floor of the camera system to see anything. My point really is with a CCD camera with deep wells and low read noise that is rarely less than 15 minutes.

If the top guys are seeing a gain with 30 minute subs using these types of cameras given stable weather and good tracking its worth a try to see for yourself.

I think also it depends on the target object. Some are quite bright and require a different strategy. I wouldn't do 30 minute subs on the Orion Nebula for example, it would be a mess but on a faint spiral galaxy at a dark site with excellent tracking and a deep well camera I would.

This is all highly theoretical anyway because I imagine very few would be able to achieve round stars in 30 minute subs anyway. That's a whole long runway up to achieve that. Its gonna require a very good mount that works reliably, perfect polar alignment like only achieved with T-point modelling, a very good PEC (a whole world in itself), really well balanced, a good autoguiding system with no flexure so that means self guiding or an off axis guider or an AO unit ( it also means guide scopes are unlikely to be able to achieve it), no cable drag, a solid pier or tripod that is well supported and a reliable power supply. If you setup every night it would be a no go. If you have a permanent mount but not all the bells and whistles needed to set it up its a no go. Unless you have a very expensive mount it is probably unrealistic.

Accurate tracking is a large enough barrier to make greater than 10 minutes unreachable for most. It really is the big obstacle in astroimaging. 10 minutes is hard enough most of the time!


Greg.

[rmuhlack]

I'll chime in here. All of my imaging is with a DSLR (i have a 1000D and a 400D). I don't have a permanent setup and have to setup prior to each session. I have two setups: a VC200L at f6.4 (FL 1278mm) on a EQ6, and a 130mm f5 newt (FL 650mm) on a HEQ5. All of my finished images to date have been with a guidescope, although I have recently purchased a Lacerta OAG to be used with my 1000D on the VC200L (although I don't have any complete projects with the OAG yet).

As my DSLRs do not have cooling, thermal noise is significant and will vary with sub length as well as ambient temperature. I have done a number of tests as per Roger Clark's methods here (http://www.clarkvision.com/astro/canon-10d-signal-to-noise/) to determine the required sub length to optimise SNR. For example, see my image of NGC6744 here: http://www.iceinspace.com.au/forum/showthread.php?t=108354

So that takes care of sub length. I terms of total exposure, given that I am already behind the QE 8-ball by using a DSLR my strategy is basically to take as many subs as I can, however beyond about 200 and my PC is working pretty hard to integrate all the calibrated subs. So given that I typically use 3-4min subs during winter, that gives a total exposure of up to 10-14ish hours.

[rmuhlack]

ps an added benefit of taking so many (dithered) subs is that minimal (if any) noise reduction is required during processing.

[Shiraz]

hi Richard. So your sub length is essentially limited by the thermal noise as the chip heats up during imaging - have i got that right? Dark calibration must be a pain if that is the case. Does the optimum sub length vary with ambient temp?

[gregbradley]

So that takes care of sub length. I terms of total exposure, given that I am already behind the QE 8-ball by using a DSLR my strategy is basically to take as many subs as I can, however beyond about 200 and my PC is working pretty hard to integrate all the calibrated subs. So given that I typically use 3-4min subs during winter, that gives a total exposure of up to 10-14ish hours.[/QUOTE]

You can go longer. You stack as many as your computer can handle and label it a submaster. Then stack the next load and so on. Then stack the resulting submasters. I do that all the time as my computer bogs down with 10 x 1x1 binned 16803 images at 32mb each. You may lose a bit of your statistical rejection of outliers but it you still get the gain in SNR as far as I know.

Greg.

[rmuhlack]

Based on my interpretation of the test results I have seen you are correct - thermal noise is the limiting factor. I have dark masters taken at various temperatures, and do my best to match sub frames with the appropriate master (within the limits of practicality). My dark masters are typically 50 subs or more. My bias master has 100 subs.



Yes it does. As a rough approximation, optimal sub settings for me would be 4 mins ISO1600 in winter, 3-4mins ISO1600 in autumn/spring and probably 1-2mins in summer. Admittedly the first time I used this method of sub calculation was in May of this year, so I haven't had the opportunity to test it at ambient nighttime temps >20C

[rmuhlack]

Thanks Greg - that is good to know :thumbsup: