Noisy at 16 minutes

[arith]

Well! I thought Collins work was exceptional; but thought what would my camera perform like for 16 minutes. It is a miserable failure and you wouldn't believe the noise. It is all a learning curve. For what it is worth. just a bit of colour here it is

[Alis]

Well! I thought Collins work was exceptional; but thought what would my camera perform like for 16 minutes. It is a miserable failure and you wouldn't believe the noise. It is all a learning curve. For what it is worth. just a bit of colour here it is

What camera did you use and what setting?

[Dave Humphries]

Hi Alis,

EXIF says Canon 10D, 974 seconds at f8 at iso200

I wouldn't have said noise was that bad on what I can see here, but it may have been tweaked and would certainly have been reduced in the resizing down to 800 px width.

Cheers,

[arith]

Yes, indeed an experiment and for the software as well as the camera. I couldn't take out all the noise on normal settings and even made a new sensor fixed noise map. This is the best I can manage on a normal shot, but HDR might be a little better.

Just shows how hard it is.

[Colin Southern]

Keep in mind though that it's not just the "16 minute" exposure; the primary thing is still to have a good exposure or what levels you DO have will be buried down amongst the noise floor.

This a common issue with digital photography -- Sensors are (for the most part) linear whereas our eyes aren't and that's a BIG problem. If you stop a lens down 3 stops from wide open and then repeatedly press and release the DoF prefiew button you can get an idea what a 3 stop change looks like to our eyes ... it's noticeable, but not by a particularly significant amount. From the cameras perspective though you've just chucked away around 7/8ths of all of the info that you could have captured. So the highlights that you've captured as mid-tones get moved up (stretched) to become highlights again, the midtones that you've captured as shadows get shifted up to become midtones again (with a bit of noise) and the shadows that you've captured as a psudo-sludge of dark shadows and noise get upshifted to brighter levels of dark shadows and noise.

The reason we get away with it is that most reflective scenes only require 4 stops - so if you under-expose by 3 stops you're "only" 7 stops into your dynamic range - and since most sensors are good for around 12, you're still 5 stops above the noise floor. If you under expose too much though (like the shot above appears to be) then large portions of your image ARE the noise floor.

So first rule is ALWAYS to nail the exposure, which can be tricky with sunsets and sunrises because (a) the camera can't meter it and (b) it's changing considerably over that period of time (which is where a lightmeter comes in handy) (allowing for around 10 stops of filtering - 2 stops to -up-shift the metered highlight to a captured highlight). To be honest, it's not always easy; the shot of mine that you saw was pretty well exposed (a little under, but I like it that way so that I retain better saturation) - the shot that followed in the NEXT 16 minutes - even though I doubled the ISO - was hopelessly under-exposed.

[wjh31]

like Dave, i cant see any noise, what i do see however is no high frequency component, its all smoothed and rounded, like it has had noise reduction added to it to a completely excessive degree

[arith]

like Dave, i cant see any noise, what i do see however is no high frequency component, its all smoothed and rounded, like it has had noise reduction added to it to a completely excessive degree

Yes, the highest most aggressive settings, and all is left is a blurry blob; but Colin is right about exposure, the histogram only went up to mid tones with tiny high tone peaks.

But it is a silly test since it was really dark and I don't expect it to be that dark normally, the room is only lit by a street lamp a few hundred yards away.

[john w revie]

Well! I thought Collins work was exceptional; but thought what would my camera perform like for 16 minutes. It is a miserable failure and you wouldn't believe the noise. It is all a learning curve. For what it is worth. just a bit of colour here it is:

Arith

The way I get round the problem of high noise levels with long exposures from my 10D
is to take multiple frames, the photo below is a blend of 26 frames @ 10 seconds
total time 4.3 minutes.



John

[Colin Southern]

Excellent point John - assuming you can get a good exposure for each frame it serves as a great "psudo" ND filter.

[arith]

Is this HDR. Is it 26 frames starting at 10 secs ending 260 secs? Or something else.

[KentDub]

I still don't think I completely understand this blending of multiple exposures. Is each shot correctly exposed (i.e. higher ISO), and then the shots averaged together? It would seem that you are just doing a noise reduction technique.

[john w revie]

I still don't think I completely understand this blending of multiple exposures. Is each shot correctly exposed (i.e. higher ISO), and then the shots averaged together? It would seem that you are just doing a noise reduction technique.

The explanation of this would take hours to go through.
In 2005 a British photographer Ed Murphy wrote a paper on this subject
A Complete Introduction to Stacking Images.
it can be downloaded from here
http://www.pdfxp.com/example-stack-pdf.html
I hope you find this useful

john

[Colin Southern]

I still don't think I completely understand this blending of multiple exposures. Is each shot correctly exposed (i.e. higher ISO), and then the shots averaged together? It would seem that you are just doing a noise reduction technique.

Noise reduction is one part - but don't forget that in the case of John's shot, there's also star movement between the shots.

Blending is easy - I just pop them into Photoshop as layers and change the layer opacity; 100% for the bottom then (working up) 50%, 33%, 25%, 20% Etc - works like a charm
.

[arith]

I had a look at the site and can't get anything, but I am interested.

[john w revie]

Arith

This is the direct to the PDF file

http://users.zoominternet.net/~ed.murphy/stacking.pdf


John

[eNo]

Even though long exposures give cleaner results at low ISO than one would get from the equivalent exposure at a higher ISO, there are physical limitations as to how a sensor will perform when it's "on" (in the on-and-capturing sense) for long periods of time. Thermal noise starts creeping in as the sensor heats up during the long capture. I know earlier Nikon D80's had problems with amp noise during long exposures which were resolved in later releases of the camera, so there are clearly design features manufacturers can put in place to minimize this thermal effect. I bet the Canon 40D is a better performer in this respect than your earlier 10D.

[arith]

Ta for that, but it will take time to read.

[arith]

I've found some time to read about stacking; Ed Murphy doesn't say if all the exposure values are the same or mention the blending mode.

Assuming they are all the same to produce random noise; I found a website suggesting the use of averaging, I've got every blend mode in Gimp except averaging.

So I thought if I put a mark to represent noise on a base image and stack a good one one on top I could try them all to see what happens.

Apart from a change in opacity nothing.

If it is only an opacity thing surely stacking increases noise.