dpreview iso comparison

The only way I can see this argument being settled is if somebody can do a back to back test of a stationery subject. Camera on a tripod. Focused only once. Fixed aperture. Fixed ISO. Single immobile variable intensity light source (a dimmer switch would). Exposure balanced so that histograms look consistent between shots. Anything else is always likely to be open to conjecture. I'd volunteer, but I'm at work :)
 
Phil Young said:
We were discussing underexposed shadows at high ISO will cause noise to be more noticable.
Click to expand...

Yes, it will, and anyone should be able to prove this to themselves in just a few minutes by taking several shots at different exposures and viewing the results.

I'm not a great believer in posting shots on line and arguing about the merits and differences when in these digital times we should all be able to test these little scenarios and theories out in just a few minutes and draw our own conclusions that give each of us a way of working that works for us.

Intersting read, though.
 
Last edited:
gad-westy said:
The only way I can see this argument being settled is if somebody can do a back to back test of a stationery subject. Camera on a tripod. Focused only once. Fixed aperture. Fixed ISO. Single immobile variable intensity light source (a dimmer switch would). Exposure balanced so that histograms look consistent between shots. Anything else is always likely to be open to conjecture. I'd volunteer, but I'm at work :)
Click to expand...

Exactly what I was thinking...
I've set up a test, just got to wait for the light levels to drop then we will have something.

I'm using a 5DIII at ISO 12800, 135mm f2 at f5.6, manual focus, tripod, live view, 2s timer, metering of the centre area should fingers crossed be consistent regardless of what happens outside.
It's currently bright and sunny, lots of indirect light - just got to wait for a rain shower or dusk.

But the important thing is that the ONLY thing I'm changing between shots is shutter speed.

I've picked ISO 12800 as my testing convinced me it's the highest I'm happy printing to A3+ with no noise reduction.
These tests are to satisfy my own curiosity as much as anyone elses!
One thing is for sure, the bright light shot looks remarkably clean compared to the low light shots I've pixel peeped in the past - this could be interesting!

Oh - and in case I fluff the test conditions please can someone else set this up too - I like the dimmer switch idea!
 
Last edited:
Phil Young said:
Have you been following the thread?

We were discussing underexposed shadows at high ISO will cause noise to be more noticable.

You done exactly that and proved only what we already concluded...
Click to expand...

Yes - I managed thanks :LOL:. Seeing as apparently I underexposed shadows you'd copy a histogram from above, and show the shadows areas on it. Theres not a whole lot of data there. Certainly not enough for you to be making that sort of somewhat silly claim :shake:.

But I know you dilike histograms, so shall I do it for you? :thinking:
 
DuncanDisorderly said:
Oh - and in case I fluff the test conditions please can someone else set this up too - I like the dimmer switch idea!
Click to expand...

Give me an hour and I'll try it. One of my rooms has blackout curtains. I'm only going to shoot the highest native iso of a d800 I can though.
 
gad-westy said:
The only way I can see this argument being settled is if somebody can do a back to back test of a stationery subject. Camera on a tripod. Focused only once. Fixed aperture. Fixed ISO. Single immobile variable intensity light source (a dimmer switch would). Exposure balanced so that histograms look consistent between shots. Anything else is always likely to be open to conjecture. I'd volunteer, but I'm at work :)
Click to expand...

Can we just be sure in what the objective is;

1) To see if there is any more noise (clearly visable) between sunlight and ambient light.

2) To see if slower shutter speeds (bar long exposures) induce more noise than faster speeds.

I'd do it too but don't finish work until 10pm.
 
gad-westy said:
The only way I can see this argument being settled is if somebody can do a back to back test of a stationery subject. Camera on a tripod. Focused only once. Fixed aperture. Fixed ISO. Single immobile variable intensity light source (a dimmer switch would). Exposure balanced so that histograms look consistent between shots. Anything else is always likely to be open to conjecture. I'd volunteer, but I'm at work :)
Click to expand...

just one suggestion, a neutral meter reading ie no comp. balanced histograms will cause rows
 
I've just done this at home with curtains open and curtains closed :D

What's it supposed to prove?

I thought the point was that underexposed images / underexposed areas make noise more visible? At the same ISO and aperture with just the shutter speed changing as long as the exposure isn't too long what would you expect to see?

I see very similar images.
 
Phil Young said:
Can we just be sure in what the objective is;

1) To see if there is any more noise (clearly visable) between sunlight and ambient light.

2) To see if slower shutter speeds (bar long exposures) induce more noise than faster speeds.

I'd do it too but don't finish work until 10pm.
Click to expand...

'Sunlight and ambient light' doesn't mean much, as sunlight IS ambient light...

If you mean between sunlight and non strobe artificial light, that's better.

Then you need to specify your artificial light (tungsten lamp, sodium vapour, halogen lamp, xenon lamp, LED lighting, flourescents (and then CCFL or mercury?) because every one of those has a different spectral output (some of them have a black body like output so are similar to sunlight in character if not peak wavelength, some are discrete wavelength so have huge gaps in output which can really mess with colour, some (sodium vapour) :bang: have a nearly single wavelength output and are horrible for colour and will answer the question differently.

And after all that, you'll get the same answer as boyfalldown's test - it's not inherently indoor or outdoor light that changes the output, it's the quality of the lighting :LOL:
 
Last edited:
ausemmao said:
'Sunlight and ambient light' doesn't mean much, as sunlight IS ambient light...

If you mean between sunlight and non strobe artificial light, that's better.

Then you need to specify your artificial light (tungsten lamp, sodium vapour, halogen lamp, xenon lamp, LED lighting, flourescents (and then CCFL or mercury?) because every one of those has a different spectral output (some of them have a black body like output so are similar to sunlight in character if not peak wavelength, some are discrete wavelength so have huge gaps in output which can really mess with colour, some (sodium vapour) :bang: have a nearly single wavelength output and are horrible for colour and will answer the question differently.

And after all that, you'll get the same answer as boyfalldown's test - it's not inherently indoor or outdoor light that changes the output, it's the quality of the lighting :LOL:
Click to expand...

I'm not the one who thinks the lighting makes a difference.

I'm with you.

And yes sunlight vs any artificial lighting.
 
Phil Young said:
Can we just be sure in what the objective is;

1) To see if there is any more noise (clearly visable) between sunlight and ambient light.

2) To see if slower shutter speeds (bar long exposures) induce more noise than faster speeds.

I'd do it too but don't finish work until 10pm.
Click to expand...

1) I don't really understand this. Sunlight and ambient light could be the same thing surely? Either way they're uncontrollable which is why I suggested a single light source that is controllable. A single light bulb with a dimmer switch in an otherwise completely unlit room sounds ideal to me because it will keep all of the highlights and shadows in the same place.

2) Erm, well it wouldn't show that as not only has shutter speed dropped but light level has dropped. Two variables. You can read into what you will as to what the cause and affect are. How you'd ever test those variables independently I wouldn't know, but seriously I don't mind what's concluded, it would just be interesting to see what happens when it is only light intensity that is varied.
 
boyfalldown said:
:D not at all. I just don't want to be told I've underexposed shadows, when there nearly no data in the shadows area. :LOL:
Click to expand...

What i was trying to say was that the exposures were different, as the shutter speeds and histograms show. This much is evident!
 
Any photos yet......the suspense is killing me..........
 
scottthehat said:
yes i do phil, maybe you should try it go out and take a photo at f4 and iso at 12800 shutter speed at 400/5000th sec what ever you need to get the correct exposure,
Now go in to a dim room and use iso 12800 and f4 and the correct shutter speed to get the correct exposure and you tell me the level of noise is the same(y)

now if you had a full fram camera you could do the same with both and see that the full frame does a better job,
And no im not saying that crop cameras cant do high iso the can and have come on shed loads, I used a pentax k5 the other day for a 2nd cam at a wedding as my d300 was away for a service, and the iso did beat the d300 but when it come to some big prints and high iso over 6400 my d700 beat it.
Click to expand...

Just to clarify people.

This was the original argument back on page one.

Please read and digest to see where I'm at.

Very basically: I'm saying you won't see a difference in noise if you expose correctly for the areas that are more predominent in noise (shadows).
 
tiler65 said:
Any photos yet......the suspense is killing me..........
Click to expand...

Full image followed by 100% crops.

_1090245c.jpg


C1. ISO 1600, 1/40 sec.
_1090246c1.jpg


C2. ISO 1600, 1/4 sec.
_1090245c2.jpg


I see a slight difference in exposure but no real difference in noise. Which I think is what I'd expect. I'd expect to see more noise if I underexposed one of the shots.
 
Last edited:
Rapscallion said:
What i was trying to say was that the exposures were different, as the shutter speeds and histograms show. This much is evident!
Click to expand...

I know - but there was someone else argueing they were underexposed (not you) when the vast majority of the data was still to the right of the charts, and I'd included underexposed shadows when there was next to no data in that part of the graph
 
Phil Young said:
Nooooo lol.

2 different exposures again.

Please make sure they look the same in exposure levels bearing in mind this is not a metering test!
Click to expand...

Agreed!
My test is just metering on the area I am going to 100% crop from.
Fingers crossed, the exposure in that area is going to be almost identical.

Progress so far.
Bright conditions - 1/3200s
Overcast - 1/1600s
Light is now fading so will be getting more normal shutter speeds soon.
Though I'm beginning to wonder if I should have stuck a 2 stop ND on there to start with more conventional shutter speeds.
 
If the point is to take two shots at the same ISO and aperture, leaving the camera to meter and set the shutter speed, that's easy but the histograms don't look exactly the same.

If the point is to use the same ISO and aperture and expect to get histograms that look exactly the same... I can't do it. The histogram always looks slightly different even with compensation.
 
woof woof said:
If the point is to take two shots at the same ISO and aperture, leaving the camera to meter and set the shutter speed, that's easy but the histograms don't look exactly the same.

If the point is to use the same ISO and aperture and expect to get histograms that look exactly the same... I can't do it. The histogram always looks slightly different even with compensation.
Click to expand...

Agreed - but using spot metering it is possible to minimise the problem.
Your book corner has fairly different amounts of light falling on it, they could be far closer to each other.
 
They're taken just seconds apart while I close the curtains. I think it's a metering issue. I'll do it again with spot.
 
OK. I tried to get the histograms as closely balanced as I could. Theres not as much difference between the slower shutter speed (poorer light) and the higher shutter speed as I expected, but it is there. I also noticed that the RAW files tend to be about 8-10% bigger with the poorer light, suggesting more random data (noise that can't be compressed)

OK ISO 6,400 f/5.6 .25 exposure

HM5_4293.jpg


and its histogram

Screen-Shot-2012-09-25-at-15_22.jpg


and the other. Excatly the same but 1/15 shutter. And lots less noisy

HM5_4294.jpg


and its histogram

Screen-Shot-2012-09-25-at-15_23.jpg


both were shoot varying the light levels only using a dimmer switch.
 
boyfalldown said:
OK. I tried to get the histograms as closely balanced as I could. Theres not as much difference between the slower shutter speed (poorer light) and the higher shutter speed as I expected, but it is there. I also noticed that the RAW files tend to be about 8-10% bigger with the poorer light, suggesting more random data (noise that can't be compressed)

OK ISO 6,400 f/5.6 .25 exposure

and its histogram

and the other. Excatly the same but 1/15 shutter. And lots less noisy

and its histogram

both were shoot varying the light levels only using a dimmer switch.
Click to expand...

There's a +.3 difference in the less noisy one.

To ensure a fair test both images should have the same brightness.

My suggestion;

Either compensate to look the same or stick it in manaul mode.

Much better test.
 
boyfalldown said:
OK. I tried to get the histograms as closely balanced as I could. Theres not as much difference between the slower shutter speed (poorer light) and the higher shutter speed as I expected, but it is there. I also noticed that the RAW files tend to be about 8-10% bigger with the poorer light, suggesting more random data (noise that can't be compressed)

OK ISO 6,400 f/5.6 .25 exposure

Screen-Shot-2012-09-25-at-15_22.jpg


and the other. Excatly the same but 1/15 shutter. And lots less noisy

Screen-Shot-2012-09-25-at-15_23.jpg


both were shoot varying the light levels only using a dimmer switch.
Click to expand...

to my eyes, these two exposures looks exactly the same. :thinking:

many thanks for the comparison shots.
 
Phil Young said:
There's a +.3 difference in the less noisy one.

To ensure a fair test both images should have the same brightness.

My suggestion;

Either compensate to look the same or stick it in manaul mode.

Much better test.
Click to expand...


It was in manual mode ;). Nobody suggested that there had to be x stops of difference between the two exposures and the histograms and exposures) to all extents and purposes are as indetical as possible. You can't keep moving things cause you don't like the results.

The only difference between the two images is white balance. Both are an auto, and the cameras compensated for the changing light level.
 
boyfalldown said:
OK. I tried to get the histograms as closely balanced as I could. Theres not as much difference between the slower shutter speed (poorer light) and the higher shutter speed as I expected, but it is there. I also noticed that the RAW files tend to be about 8-10% bigger with the poorer light, suggesting more random data (noise that can't be compressed)

both were shoot varying the light levels only using a dimmer switch.
Click to expand...

filament bulbs (tungsten, halogen) are all black body emitters, so lowering the brightness you still get a broad spectrum response though the spectrum is skewed. What you want to do (to see a real difference) is compare 'colour corrected' sodium street light exposure or flourescent exposure with halogen or daylight.

Dimmer switch means the temperature of your bulbs dropped, which will have reduced the blue content. Out of interest (and to demonstrate a point) if it's not too much hassle can you run just blue channel noise reduction on the longer exposure and post the result ;)?
 
Last edited:
wuyanxu said:
to my eyes, these two exposures looks exactly the same. :thinking:

many thanks for the comparison shots.
Click to expand...

the histograms?. They were as close as I could get them, the main difference is in the blue channel in the highlights,
 
boyfalldown said:
It was in manual mode ;). Nobody suggested that there had to be x stops of difference between the two exposures and the histograms and exposures) to all extents and purposes are as indetical as possible. You can't keep moving things cause you don't like the results.

The only difference between the two images is white balance. Both are an auto, and the cameras compensated for the changing light level.
Click to expand...

The results are in my favour?

Just saying that because the first is slightly less exposed it could affect the noise
 
ausemmao said:
Dimmer switch means the temperature of your bulbs dropped, which will have reduced the blue content. Out of interest (and to demonstrate a point) if it's not too much hassle can you run just blue channel noise reduction on the longer exposure and post the result ;)?
Click to expand...

am ashamed to admit I don't know how to do that. Give me a quick pointer I will
 
I've just tried spot and the histograms look different. I didn't try setting a different shutter speed to what the metering set to try and get the histograms closer together.

Is this all going too far? :wacky: :D
 
Phil Young said:
There's a +.3 difference in the less noisy one.

To ensure a fair test both images should have the same brightness.

My suggestion;

Either compensate to look the same or stick it in manaul mode.

Much better test.
Click to expand...
Really?

Really?

Have you ever seen that kind of difference from a third stop difference in exposure?

Really?
 
You must log in or register to reply here.
Back
Top