Determining LP limit

I'm looking for comments on determining the best exposure times based on sky darkness.
I'm sure I saw somewhere that I should aim for the majority of the (in camera) histogram just moving off the left axis to get get the best s/n. Does that make sense ?

Steve.

I dont think you can judge it from the Histo. As you approach the Sky Glow Limit the wanted signal merges with the sky glow so they will share the same space. If you want more information than you need the http://www.hiddenloft.​com/notes/SubExposures​.pdf has to give it (there are others on the same site).

Check my articles under the basics section of my website in my signature.

More than just LP to content with.

I generally do judge by the histogram (in camera or in DPP). If I'm shooting a faint object, I will push the hump in the histogram, which represents the sky glow, to the middle or even past the middle.

The concept of "sky fog limit" is often raised with respect to DSLR photography, but in truth it is completely inapplicable. Sky Fog Limit relates to film astrophotography and the phenomenon of reciprocity failure. As film is exposed to light, it loses its sensitivity in proportion to the amount of light it receives, so eventually the sky glow will catch up with the signal from the subject, and you must stop exposing before that occurs. With digital sensors, there is no reciprocity failure and thus no sky fog limit. The sky glow is always the faintest part of the image no matter how long you expose, up until it saturates the detector.

The histogram displayed on camera and in DPP has a logarithmic X-axis, and the mid point represents only a few hundred counts of signal - I analyzed it with my 350D once but I don't remember the actual numbers. The mid point of the histogram represented somewhere around 600 counts out of 4096. The bottom line is when the sky glow hump is in the middle of the histogram, you still have a lot of room left for signal. Living in a fairly light polluted area (orange on Clear Sky Charts), I consider learning this to have been a turning point in my DSLR astrophotography efforts.

Catanonia's article makes some good points, though. In addition to sky glow, you need to allocate your exposures according to the time you have available for imaging a subject, and the capabilities of your mount.

Don

As I understand it, exposing long enough to get the histogram away from the left hand edge is to make the captured signal stronger than the random read out noise that comes from the sensor when the image is captured. If the captured signal was mixed in with this it would be much harder to extract the signal.

The captured signal in this case still includes the light pollution and it is the stacking process that reduces the variation in the distributions of signals from sky glow and target objects. The reduced variation makes it easier to separate the sky and target objects in later processing.

All this is in addition to the processing to remove fixed noise...

That's good information about the scale of the in camera histogram... I always assumed it was linear.

http://starizona.com/a​cb/ccd/advtheoryexp.as​px has some good and reasonably uncomplex explanations. The key point is that a lot of the fainter objects have a flux that is only just above the sky flux of a reasonably dark sky i.e. as I understand it, when the photons arriving from the object we are trying to image are arriving no more frequently than those from the sky background you are wasting your time. LP filters work by filtering out common light pollution source wavelengths and will to some degree redress that balance by saying no entry to some of the arriving errant photons.

For me imaging towards the north is a no no without an LP filter, I am obscured by trees to the south west that gives me a pretty small window from east round to south. Between North and South the sky brightness measured fairly roughly in Maxim varies by a factor of nearly 100!

If you take a typical subframe and look at the histo in a program that shows it linearly 95% of the data is in a very narrow spike that is hard up against the LHS. In amongst that is the sky glow and the various noise contributors.

chris.bailey wrote in post #9424564
http://starizona.com/a​cb/ccd/advtheoryexp.as​px has some good and reasonably uncomplex explanations. The key point is that a lot of the fainter objects have a flux that is only just above the sky flux of a reasonably dark sky i.e. as I understand it, when the photons arriving from the object we are trying to image are arriving no more frequently than those from the sky background you are wasting your time. LP filters work by filtering out common light pollution source wavelengths and will to some degree redress that balance by saying no entry to some of the arriving errant photons.

For me imaging towards the north is a no no without an LP filter, I am obscured by trees to the south west that gives me a pretty small window from east round to south. Between North and South the sky brightness measured fairly roughly in Maxim varies by a factor of nearly 100!

If you take a typical subframe and look at the histo in a program that shows it linearly 95% of the data is in a very narrow spike that is hard up against the LHS. In amongst that is the sky glow and the various noise contributors.

Hi Chris,

I've seen that article before and it is certainly interesting. One of the problems you will encounter if you try to apply those formulas for optimal exposure length to data from your Canon DSLR is that Canon doesn't publish the gain and readout noise figures for their DSLR's (at least I haven't been able to find them). There are numerous online articles that quote values for these figures, but those values are all over the place. It seems everyone who tries to quantify them has a different idea of how to do it, and using published values found on the internet in the optimal exposure length formulas will result in values between about 15 seconds and 10 minutes, using typical data from my Bortle 5 night sky.

Using some published values for the 350D that seem to be in the middle of the range, 1.5 for gain at ISO 800 and 7 for readout noise, and using data I collected last night for sky brightness, I get 2.9 minutes for the optimal exposure time using the formulas in that article, which just happens to be close to the length of subexposures (3 minutes) I was shooting last night at ISO 800. The average ADU value for the sky background was 580, and the Canon 350D applies a "pedestal" value equal to 255 ADU, resulting in a measured sky flux value of 325 ADU times 1.5 divided by 3 minutes, or 162.5. Dividing that figure into 9.76 times 7 squared gives an optimal exposure time of 2.9 minutes.

DPP is one of those applications that can display a linear histogram, though it displays log by default. The adjustment tool in DPP has a checkbox for switching to linear mode. I have attached a screen capture from DPP showing the histograms of two images taken back to back. The upper histogram is shown with the normal logarithmic x-axis and the lower histogram is shown with a linear x-axis.

Though M42 isn't exactly a faint object, there is considerable detail there that you can capture with sufficiently long exposure times that would be missed with shorter exposure times. Based on your description of the linear histograms you have observed, it would appear that I expose longer than you relative to the sky glow level, as the hump in my linear histogram is clearly separated from the left side.

Now the article on Starizona's site confuses the issue (in my mind at least) by stating near the end of the article that after calculating the optimal exposure time, you should divide it by two and take twice as many subexposures! That seems pretty arbitrary to me. I was able to get 35 subs at three minutes each last night, and I don't believe that the benefit of taking 70 subs instead of 35 would have outweighed the advantage of the longer exposure time in this case. When I finish processing this image I can post it in this thread.

The bottom line IMHO is that we should forego being overly analytical in choosing exposure times in favor of experimenting and determining what yields the best results. As I stated in a previous post, for faint objects my experience indicates that pushing the sky glow hump towards the middle of the histogram works well, given my sky conditions and assuming I'm after the finest detail I can separate from the sky glow.

HOSTED PHOTO
please log in to view hosted photos in full size.

Don, really interesting stuff. I have been doing the same and got an optimal sub frame length of 3.3 minutes. I have previously tended to stick to 4 or 5 minutes as being optimal in terms of S/N ratio but it was gut feel rather than backed by any sort of science. My background ADU is 1550 in the best bit of sky and nearly 4000 in the worst. To counter that my CCD has high QE and a gain of 0.4. the higher ADU of my sky background perhaps explain the footroom (if that is the opposite of headrooom) of your histogram compared to mine.

http://starizona.com/a​cb/ccd/calc_ideal.aspx works out CCD sub lengths for you but does not include DSLR's.

The Starizona article does agree with the one from hidden loft if you read the two together but I would agree that the difference in terminology makes it more difficult. In essence the hidden loft articles suggests little penalty for 2n exposures of t/2 compared with n exposures of t. I assume the 2n+1 is assumed to make up the shortfall.

I totally agree about not being over analytical but it is interesting and perhaps sways one away from seeking longer and longer sub-frames with the view that it will make things better especially given the high risk and cost of an "event" occuring during one of those long exposures. Convinces me to stick in the 3-6 minutes range.

I need to type faster.

chris.bailey wrote in post #9424564
If you take a typical subframe and look at the histo in a program that shows it linearly 95% of the data is in a very narrow spike that is hard up against the LHS. In amongst that is the sky glow and the various noise contributors.

And faint fuzzies - The stuff we want.
That is what got me thinking about it. Surely if it's hard up against the left axis it means that there are photosites that never saw a single photon. So how much longer could I have exposed for to get something in those darkest photosites. Any more is time wasting and any less is missing out on potential data.

The other thing I noticed (and Don's images show it well) is that the peak becomes wider as it moves towards the right. Assuming the sky glow to be comparatively constant, I would interpret that as being more good stuff.

But if you were to carry that to the extreme and drive the histogram to the right even further, brighter objects would washout dimmer ones, you'd loose data through quantization and so on. Hence the question - When is enough enough and when is it too much ?

I don't mind the math. I find it interesting and it can be useful in explaining or proving a concept. However, I'm a long way from trying to get everything down to the nth degree. I'm at the "should I drive to a dark sky on a weeknight" level.

Steve.

chris.bailey wrote in post #9426561
My background ADU is 1550 in the best bit of sky and nearly 4000 in the worst.
.

What is your A/D bit depth? The sky glow ADU being a digital value, it's level depends on the A/D bit depth. My Canon 350D has a 12-bit A/D, so my sky glow of 325 ADU (measured using the RAW image file) corresponds to about 5200 ADU on a 16-bit scale. That was just last night, though, and last night the transparency was pretty poor here. On a good night I could go four minutes, and on an excellent night (which I haven't seen in a long time here in good ol' Georgia) I could go five minutes at ISO 800 with the LP filter, before the sky glow hump reached the middle of the histogram.

16 bit so pretty comparable.

DonR wrote in post #9426392
I was shooting last night at ISO 800. The average ADU value for the sky background was 580, and the Canon 350D applies a "pedestal" value equal to 255 ADU, resulting in a measured sky flux value of 325 ADU times 1.5 divided by 3 minutes, or 162.5. Dividing that figure into 9.76 times 7 squared gives an optimal exposure time of 2.9 minutes.

This makes my head hurt. I lost you at ISO 800. Interesting stuff that I wish I understood better.

SteveInNZ wrote in post #9427024
I need to type faster.

But if you were to carry that to the extreme and drive the histogram to the right even further, brighter objects would washout dimmer ones, you'd loose data through quantization and so on. Hence the question - When is enough enough and when is it too much ?

I don't mind the math. I find it interesting and it can be useful in explaining or proving a concept. However, I'm a long way from trying to get everything down to the nth degree. I'm at the "should I drive to a dark sky on a weeknight" level.

Steve.

I think the answer to "when is it enough?" depends on what you are trying to achieve. If you want to capture the delicate arms of a faint galaxy, then go as long as you can. If you're doing a constellation portrait and want to represent the visible star colours then you'd better hold back, because they'll all burn out to white.

As for me... at the moment I'm having alignment problems so I dream of over two minutes Next stop drift aligning to find out what my problem is.

MintMark wrote in post #9432193
Next stop drift aligning to find out what my problem is.

Good luck. When I pick up my pier on Saturday I'll be hoping for good weather so I can perhaps get it concreted in over the weekend. Then two weekends later I'll be doing some aligning myself.

Adrena1in wrote in post #9432653
Good luck. When I pick up my pier on Saturday I'll be hoping for good weather so I can perhaps get it concreted in over the weekend. Then two weekends later I'll be doing some aligning myself.

Some form of permanent set-up is getting far more attractive. To be effective though I would need to do some major deforestation whilst my neigbours were away.