Confusion over high ISO, noise, and light levels

Right...... Taken some test shots at 400, 800, 1600, 3200, 6400 and 128000 in RAW with exposure adjusted automatically in LR

You'll be able to see from the EXIF data that I shot in P mode with flash firing.

I can't see a massive amount of difference when viewing full size but at 100% crop there's a massive difference to the point that the picture actually looks OOF, even the white sheet of paper.

Not a good example for the church and bride scenario but it does illustrate how the same shot with the same exposure (give or take a little) at different ISO affects different tones.

_MG_0518Edit.jpg


_MG_0519Edit.jpg


_MG_0520Edit.jpg


_MG_0521Edit.jpg


_MG_0522Edit.jpg


_MG_0523Edit.jpg
 
Right...... Taken some test shots at 400, 800, 1600, 3200, 6400 and 128000 in RAW with exposure adjusted automatically in LR
Russ, there's no point trying to prove that high ISO settings produce more noise than low ISO settings. We all agree on that.
 
It seems to me that what Stewart is saying is this:

  • Subject = A
  • Shot 1 of A is in good light and settings are ISO 3200, f8, 1/90s
  • Shot 2 of A is in dim light and settings are ISO 3200, f1.4, 1/90s
A higher ISO in either shot would increase noise.
A slower shutter speed in either shot would also increase noise.

However, shutter speed and ISO remain the same for both shots - only the aperture has changed. Therefore noise levels should remain the same in both shots.
That's EXACTLY what I was trying to say. I think Tim/tdodd has produced some very good arguments as to why it might not be true, but that is what I was trying to say.
Possibly right - but noise will be far more visible in the dimly lit shot.
But WHY? You keep saying this, and you might be right, but all along I've been asking WHY. Argument by assertion is a logical fallacy.
 
Russ, there's no point trying to prove that high ISO settings produce more noise than low ISO settings. We all agree on that.

Yes but it seems to me what's being disputed is how the noise manifests itself so I thought having a shot with colour, hightlights, midtones and blacks in a uniform manor with a consistent light source (i.e. a flash) would demonstrate this.

If the majority of the picture is light, you'll get away with a higher ISO, if it's dark you wont.

So in the situation with the church and using ISO3200 and different f-values to get the same exposure.....

If the church has enough light to allow you to use f8 then the dark areas in the church wont be as dark than if you had to use f1.4, therefore they wont come out with as much noise as they're lighter to start with (However I'd rather shoot at ISO1600 1/90 F5.6 instead of 3200, f8 :lol:).

If you were in a room with no light at all you wouldn't be able to see anything (obviously), if you gradually increased the light, the first things that would become visible are the lightest parts of the room, the dark parts of room would require more light to become visible to you.
 
That's EXACTLY what I was trying to say. I think Tim/tdodd has produced some very good arguments as to why it might not be true, but that is what I was trying to say.

Yes, Tim makes some valid points, especially his point 2 (post 24) - we can't try it out though without some serious measuring gear and a laboratory.

Assuming we could accurately control the light levels to produce shots 1 & 2 (post 38) without changing the colour of the light at all, the noise in both shots should be exactly the same (ignoring 'perceived noise' due to more parts of the f1.4 shot being OOF).
 
Are sensors absolutely linear in their required exposure? I wonder if the noise perceived is because they aren't so that at higher isos you have to compensate more as the in camera exposure is often underexposed hence noise and lack of shadow detail. Some manufacturers seem to be using something akin to pushing film to get the higher settings which is why colours are more washed out and there's more visible noise. Would adding exposure compensation fix the issues altogether?

If the sensor is actually sensitive enough should there be any noise at all?

Maybe the only answer is to use HDR for every shot in low light :D
 
I wouldn't have a clue how to quantify the effect in terms of perceived noise, but you can pull out the WB multipliers from most raw files for different settings. These will show that R and B channels are almost always weak compared to G and need some degree of boosting. Just how much depends on how far off balance the light source is.

...snip

So switching between tungsten lighting and shaded conditions there is a big swing between the multipliers needed for the R and B channels. If you compound those natural (almost unavoidable) underexposures at the sensor with additional underexposure through poor camera operation then the potential for excessive noise is increased further.
.

Thanks Tim, yeah bad question, thanks for trying. I can see there are huge flutuations for both B and R gain as the light type changes. Ive not seen that angle of view before, very informative thanks for the details.

I took the OP's line - 'So ... why is {ISO 3200, f/1.4, 1/90th} noisier than {ISO 3200, f/8, 1/90th}, if the two photos are both exposed correctly?' _ to mean exactly or literarily perfect light wise, or WB wise if you like. ( I tailored it because I thought Stewart probably knew about histograms pushing to the right and whatnot)

So two shots both using the same quantity of light, and the same quality of light, to achieve correct exposure.
Would you agree that in that scenario RGGB Bayer gain adjustments will be the same for both shots.

Which does only leave our first thoughts, the oof CoC will cause errors naturally, and emphasised even more by a high ISO ...Unlike good sharp CoC which wont trigger such gain/multiplier side effects like noise at the same high iso.

That's EXACTLY what I was trying to say. I think Tim/tdodd has produced some very good arguments as to why it might not be true, but that is what I was trying to say.

:Cough: I see you not included me in that good argument line. So does that mean you really didn't understand me? ... I though you'd know I was referring to Circles of Confusion and DoF.
Or more precisely the angles of the light rays within the focus clusters of light at the focal plane, which all work together to produce the DoF effect .

Yes, Tim makes some valid points, especially his point 2 (post 24) - we can't try it out though without some serious measuring gear and a laboratory.

Assuming we could accurately control the light levels to produce shots 1 & 2 (post 38) without changing the colour of the light at all, the noise in both shots should be exactly the same (ignoring 'perceived noise' due to more parts of the f1.4 shot being OOF).

Yes I agree. the only difference between the shots is one contains some oof CoC and the other doesn’t.

My thoughs are,after reading Tims stuff and refreshing myself on the interweb this morning...

Some of the light making that exposure of the f1.4 image will be poorly defined onto the Bayer array, this poor definition is then attenuated into noise when processed through at the Raw level ...leavening perhaps half the information available to produce the good noise free clarity of the f8 shot.
 
Last edited:
So two shots both using the same quantity of light, and the same quality of light, to achieve correct exposure.
Would you agree that in that scenario RGGB Bayer gain adjustments will be the same for both shots.
The RGGB gain remains the same for the same white balance setting. Levels of light, ISO used, or any other aspect of the exposure do not alter RGGB gain values. Of course, for the WB setting to remain valid between shots you need to ensure that your light temperature and CRI does not alter.

I'm not sure about your use of the term "quality of light". Foremost in my mind when people use that phrase is hardness/softness of light moreso than colour or quantity. The truth is that in real world lighting scenarios it is likely that the brighter the light the harder and more directional it will be and the dimmer the light the softer and more diffused it will be. This may make it more difficult to compare overall exposure and shadow vs highlight detail and noise.

This was made abundantly clear when I tried to perform a little experiment to test out Stewart's query. I tried a shot in my living room with the curtains closed and then opened. It was pretty obvious that even though the exposures were nominally equal (autoexposure took care of that) it would be completely impossible to make meaningful comparisons of noise. Here are the images....

20110122_100153_.JPG


The first one has very soft, even lighting, thanks to the drawn curtains, while the second one has extremes of contrast due to the directional nature of sunlight streaming through the window. This was with nets still in place. Without nets the light would be even more directional. Of course there was also the issue of colour temperature too. The curtains had their own effect on the scene and with the streaming sunlight there was the choice between sunlit areas and those in shade - a minefield for RGGB channel consistency.

This is why I think that to perform the test "scientifically" you need to exercise tight control over all aspects of the lighting - colour, "quality" and quantity. Of course, that's great in an academic sense, but of limited value in the real world where it is exactly these things which do alter and thus affect the outcome.
 
Last edited:
:Cough: I see you not included me in that good argument line. So does that mean you really didn't understand me? ... I though you'd know I was referring to Circles of Confusion and DoF. Or more precisely the angles of the light rays within the focus clusters of light at the focal plane, which all work together to produce the DoF effect.
I'm sorry, but I really didn't understand you!

But I do now, and I agree the in focus / out of focus argument is persuasive.

All I need to do now is get my hands on a decent ND filter and I'll do some experiments...
 
This is why I think that to perform the test "scientifically" you need to exercise tight control over all aspects of the lighting - colour, "quality" and quantity. Of course, that's great in an academic sense, but of limited value in the real world where it is exactly these things which do alter and thus affect the outcome.

I pondered the same tests myself ..and decied against it for the same reason.

Yeah I agree, tricky test and not that useful. … I am persuaded by your arguments over ‘quality’ of light, as you suggest, it seems non-viable that the colour temp would not change as one dimmed the light, even theoretically. As you say in the real word a fixed temp is probably impossible to maintain.

I'm sorry, but I really didn't understand you!

But I do now, and I agree the in focus / out of focus argument is persuasive.

All I need to do now is get my hands on a decent ND filter and I'll do some experiments...

:thumbs: Get on board, Its a cool visualisation, well it works for me anyway. ;)


BOT
Isn’t there a factor in using a ND filter that you going to add artefacts/density to the Raw data and those could also trigger processing errors and introduce noise? ..so making the test less reliable.
Grads filter light by blocking light don’t they, that blocking will diffuse and diffract the light, hence making them a bad test for noise surely, especially since we think oof diffused light is the main cause..

I was thinking for the rggb noise idea, you could carry out independent tests using a magenta WB card as a raw WB pre-set (as Tim suggested) to provide a 1.0x multiplier/gain which in turn will provide a green cast image with poor colour saturation ..No CRI applied etc.
..
If you did this on both f1.4 and f8 then one could compare this green cast image for noise against the same exposure using a normal WB for each shot this would provide a clear answer to the RGGB noise ‘and all the other processing stuff that goes on’ idea.

I’m stuck on testing the oof noise idea though …would using a Light meter to note the actual temp at both light levels be more than useful if one could do the maths? …also does a flash maintain its colour balance. and can it go five stops apart. :thinking:
 
Last edited:
Longer exposures will heat up the sensor more and thus induce a small amount of noise, so unless absolutely required using a ND filter seems a bit pointless if it's going to impact the IQ
 
I can't contribute anything further to this discussion, but just wanted to say that I've been following with interest. :thumbs:
Definitely the most interesting and enlightening (excuse the pun) thread that I've seen on here for a long time. If anyone can figure out a way to introduce reasonable consistency across a range of test conditions I can't wait to see the results.
 
Really excellent discussion on noise, particularly the effect with WB. I do enjoy the technical side. I'm gonna bookmark this ;-)

Does this really answer Stewart's question, or more to point, is it overall relevant?

The question related to weddings and its unlikely white balance filters will be used here. Likewise, larger aperture lenses are always going to be chosen in this field over smaller aperture lenses. DoF is going to be most evident now, showing noise predominently in these OoF areas.

Larger apertures and the subsequent 'more noticable noise in darker OoF areas' is preferred to the smaller aperture alternatives of;
1.using an overall higher ISO
2.slowing the shutter potentially inducing motion blur/camera shake.

It seems that there isn't more noise at f1.4, only that its more noticeable due to smoother out of focus areas, WB being equal.

Have I got this correct guys are am I way off?



Andy
 
Isn’t there a factor in using a ND filter that you going to add artefacts/density to the Raw data and those could also trigger processing errors and introduce noise? ..so making the test less reliable.
Grads filter light by blocking light don’t they, that blocking will diffuse and diffract the light, hence making them a bad test for noise surely, especially since we think oof diffused light is the main cause.

I don't think you mean grads (graduated) but plain ol' ND filters. But if they were to do as you suggest then I think it might be a little troublesome to get a sharp and contrasty image from them. Think of putting on sunglasses. The image doesn't get destroyed; it simply gets darker. So I think that to keep things as consistent as possible, and to allow testing of exposure adjustments viq shutter speed as an alternative to aperture then NDs are probably a good way to go. A good quality brand would be advised, and one which is properly colour neutral.

I've also had another thought, which might be a complete red herring, but I'll throw it out there to see if anyone has any ideas. Due to the photonic nature of light it actually arrives at the sensor as individual random pulses, and the sensor counts these pulses in order to assign a numerical value to each RGGB sub-pixel as light flows through the open shutter. The random nature of these pulses mean that light itself is noisy. You do not get exactly equal numbers of photons arriving at adjacent pixels, even from a completely smooth and evenly lit surface. The variations in photon numbers means that from one pixel to the next you will get different photon counts and that is effectively noise. It is called "Shot Noise". At high rates of sensor fill, such as when shooting bright areas at low ISO the Shot Noise is insignificant. However, in the shadow regions, especially when shooting at high ISO, the fill rates for some pixels are very, very low. The Shot Noise becomes far more apparent in these areas.

What I can't figure out is whether in low light conditoins there is any increase in Shot Noise even if you equalise the exposure by opening up the aperture or slowing the shutter speed. My head/gut tells me that it all balances out to make no difference, but I don't know for certain. I thought I'd throw the idea out there in case anyone else has some inspiration. Is low level light noisier than bright light in and of itself? That's the question.
 
Great thread, if a bit of a brain melter.:)

Is there any mileage in thinking this could be in some way linked to a digital form of reciprocity failure when pushing the sensor to it's limit, and if so, could this be part of the whole noise 'problem'? Probably a useless addition to the debate.....
 
i would think that the harder you make the sensor work to define detail, the more noisy it will be, as someone said, if you do a long exposure you make the sensor warm, which introduces noise..show it a little bit of light and tell it that its got to render an image and it will perform differently to showing it a lot of light, iyswim? so putting the same settings but just changing the light going to the sensor will result in more or less noise, depending on how hard you make the sensor work in the given light
 
Great thread, if a bit of a brain melter.:)

Is there any mileage in thinking this could be in some way linked to a digital form of reciprocity failure when pushing the sensor to it's limit, and if so, could this be part of the whole noise 'problem'? Probably a useless addition to the debate.....

It's highly possible as a degree of linearity in exposure may have been assumed which may not necessarily exist in reality. If sensors are being used at iso settings which they are really capable of and are effectively being 'pushed' then it flatter colour and grain is what you used to see with film and is exactly what you get with higher iso images. But how do we know the iso 6400 in one camera system relates to exactly the same light sensitivity as another camera's iso 6400?

The exposure latitude of sensors may also be different. If this latitude reduces at higher isos then in camera accurate metering becomes essential to reduce noise.
 
If you take the same subject, at the same shutter speed and ISO, and adjust the light level by either, say, using an ND filter or adjusting the power of a flash gun and also change the f/number accordingly, then noise will be the same. As Tim said earlier on.

A normal scene might look different because of the change in depth of field, and also perhaps vignetting and edge sharpness, but if it was a flat 2D surface like a brick wall and a really good lens, they should look identical.

Going back to the earlier points about noise being more visible in low-light scenes, that is more to do with the likely subject and camera settings. Darker areas will show more noise, as we know, and generally speaking the shutter speeds will be longer, creating more long-exposure noise, also as we know. But it's not any inherant chacteristic of lower light, if you can compensate for that with a lower f/number and in that way collect photons at the same rate.

Also, digital sensors do not suffer from reciprocity failure as film does.
 
Last edited:
Did have a read through, but as I'm still half asleep, not sure if this was said already (or indeed if correct), but:

When shooting in a dimly lit area and using a larger aperture noise is visible in the oof areas because the noisy pixels in the image have a higher spatial frequency compared to the rest of the oof areas and are thus more obvious. This is compounded by the fact that the noisy pixels will be of a colour and/or brightness which renders them even more obvious to the eye as they are being viewed against a darker background.

If the same area was photographed in brighter light at the same settings (using ND etc) then the same noise would be present, by less visible due to the reduced contrast between the noise and the background (but still visible because of the high spatial frequency nature of the noisy pixels etc).

If the same are was photographed in brighter light using a smaller aperture, the dof increases, and so the difference in the spatial frequency between noise and scene decreases making the noise less obvious.

:shrug:
 
It seems that there isn't more noise at f1.4, only that its more noticeable due to smoother out of focus areas, WB being equal.

Have I got this correct guys are am I way off?

Yep thats the theory. ..although to me if you can notice more noise, there is more noise...chucking all noise into the one pot kinda thing.

I don't think you mean grads (graduated) but plain ol' ND filters.

......

......

The Shot Noise becomes far more apparent in these areas.

.....


Is low level light noisier than bright light in and of itself? That's the question.

Yes your right, ND not grads. thanks. I'll go with your experience on this one Tim, just musing the idea...My implication aside, they are a fab tool.

I like your 'shot noise' angle, and I agree in this case both shots receive the same amount of light within the same time so it doesn't apply.
However, thinking about the oof light rays which are effectively hitting those pixel counting pots at slightly more lose/open angles than at f8. I'm imagining that could increase the chances of your shot noise' simply because the averaging variables have been complicated slightly by the loser in-focus clusters of the f1.4 shot.

"Is low level light noisier than bright light in and of itself? That's the question"

That really is the question isnt it. ...

I say no, its all just light, (colour temp aside) its how we collect it that makes the difference.
 
Last edited:
A normal scene might look different because of the change in depth of field, and also perhaps vignetting and edge sharpness, but if it was a flat 2D surface like a brick wall and a really good lens, they should look identical.

.

I agree ..with digital sensors resolutions anyway. I’m imaging fractional discrepancies might creep in when large res film stuff is blown up huge?
 
Yep thats the theory. ..although to me if you can notice more noise, there is more noise...chucking all noise into the one pot kinda thing.

OK, so I was getting my head around it:lol:

I guess the issue then is that the only way around these more noticeable noise areas caused by OoF areas is to close the aperture down just enough and increase depth of field.

The only problem here is that there will be more overall noise, exposure will have to be compensated by ISO increments.

Could it then by argued that within the cameras native ISO range, a typical prime at f1.4 may show more noticeable noise than a lens a couple of stops slower at f2.8 due to depth of field that has been compensated with ISO increments?:thinking:


Andy
 
Last edited:
You could argue that I guess. ..which is worse, oof CoC or high iso …probably the high iso as the physical shape of oof CoC isn’t viewed as noisy by us viewers, in fact its deemed quite pretty. But the effects of raw processing errors on that oof Coc is seen as noisy.

Its like the chicken or egg which came first …The chicken in this case I think.

Its important to remember that we’re just pondering here, in real life photography NONE of this matters very much. … Noise or no noise Its far more important to get the photograph exposed correctly. A good photo with a little irrelevant noise is way better than no photo at all…or some other poor exposure.
 
Last edited:
I totally agree about the exposure thing. I aim to ETTR much of the time, even in good light.

As for DOF and OOF areas, here's a shot taken with my 7D at 6400 ISO. Viewed a a whole picture I have no issue with noise or IQ in general.

20100115_200707_2092_LR.jpg


But zoom in to pixel level and the story is very different in OOF vs sharp areas. Look at how the sharp transition between black and white on the player's shoe is relatively noise free, but the soft and OOF transition between black and white in the background exhibits huge amounts of very visible noise....

20110125_095042_.JPG


I think it is also interesting to note that the really dark areas - the blacks and almost blacks - don't actually look that noisy, and neither do the bright areas. It's the mid tones that suffer the most, and soft (OOF) edges can create mid tones where actually there should be none. The shadows under the table are a good example of this. The floor is fairly clean, as are the darker portions of the larger shadow areas, but where the edges between dark and light are blurred the noise is very noticeable.
 
Interesting article bringing in more questions. Dslrs have more noise in summer than winter.

That's down to sensor temperature. Put crudely, at higher temperatures there is more energy available allowing random electron movement within the sensor and that produces random noise artefacts. This is why imaging sensors on telescopes are cooled.
 
Off topic a bit Tim, no doubt the extra noise is there, but isn't that perhaps a post processing/sharpening artifact? It seems to be picking up specifically on very high contrast transitions, and has pulled some colour from out of focus areas (residual CA?).

I wonder how different software might handle that?
 
But when they are OOF areas you no longer get sharp transitions. You get soft transitions, and the only way to softly transition from black to white is via mid tones. That's where the noise shows up most of all.

Anyway, here are two 100% crops from DPP and LR with sharpening and NR set to 0 in both.....

20100115_200707_2092_DPP.JPG
20100115_200707_2092_LR-2.jpg


FWIW, here is the same crop from Lightroom but passed through Neatimage before outputting to JPEG....

20100115_200707_2092-Edit_LR.jpg
 
Last edited:
That's interesting Tim. Thanks :)

I guess the point I was getting at was how different programmes handle these things - when they decide that a hard transition has become a soft transition, and then what they do with it. The kind of stuff that only shows up when you push things a bit, and look closely.

Not much between DPP and Lightroom really, but NeatImage looks good. Is that your preferred noise reduction programme? How does it integrate with LightRoom?
 
I very rarely resort to NR outside of Lightroom, but when I need to it's Neatimage that I use. In the example above I put in very little effort to tweak NR settings. It was just a quick and dirty.

Neatimage can be integrated with Lightroom by specifying it as an external editor. Lightroom will generate a TIFF with all the LR adjustments thus far (e.g. white balance). It then automatically opens Neatimage with the selected TIFF file and re-imports the revised file back into the LR catalog. You can then carry on with sharpening etc. on the cleaned image. You can have LR automatically stack the new TIFF file with the original raw file so that if/when you wish to revisit the file the TIFF is on top of the stack and ready to play with further.

20110125_222831_.JPG
 
Thanks Tim. V useful info :thumbs:
 
I totally agree about the exposure thing. I aim to ETTR much of the time, even in good light.



20110125_095042_.JPG


I think it is also interesting to note that the really dark areas - the blacks and almost blacks - don't actually look that noisy, and neither do the bright areas. It's the mid tones that suffer the most, and soft (OOF) edges can create mid tones where actually there should be none. The shadows under the table are a good example of this. The floor is fairly clean, as are the darker portions of the larger shadow areas, but where the edges between dark and light are blurred the noise is very noticeable.


Nicely done, it been very intersting mate, great examples …the evidence is clear! ... Soz I mean blurred but clear. :D

Just to clarify for the members a bit then? ...At these soft resolutions the Bayer array and associated algorithm just cant distinguish the oof transitions precisely This means that the raw processing and structuring of the whole image also needs to generate 'fill-in' tones to fill these indistinguishable transitions, effectively calculating, or kinda guessing, what tones would be best added to make it look smooth.

So amongst a few other raw processing processes. (such as those that have been mentioned) Its mainly the added fill in tones/hues and the errors associated and initialised by their generation that causes the noise of the f1.4 shot.


.
 
Last edited:
Sorry for the thread revival, but yesterday I came across a discussion which raises a whole new explanation for increased noise at (very) wide apertures.....

http://photography-on-the.net/forum/showthread.php?t=950999&highlight=t-stop

I tried some experimental shots of my own. Now these were in overcast daylight, so not really a low light situation, and probably with a fair spectrum of light wavelengths, so not requiring a big white balance shift to correct for a spectral imbalance. The exposures here should be equivalent in terms of photons passing through the lens, but if the assertion in the article is accurate then quite possibly the sensor is not recording as many of them at f/1.4 as at f/4, requiring a secret ISO boost to compensate. One thing is certain - the shot at f/1.4 looks a lot noisier than the one at f/4. While DOF effects may well contribute to the appearance, I'm prepared to accept that there could be more going on here than our earlier discussions considered.

I've deliberately selected an area of the frame away from the centre, where vignetting may also have an impact, and I also chose to include areas which were OOF in order to further allow any noise issues to be revealed. The biggest tell-tale is in the fence at the rear, which is almost in focus, and upon which the noise is far more noticeable at f/1.4. The staked plant at the rear also shows a marked difference, even though they aren't very OOF in the f/1.4 shot. Food for thought....

20110818_131442_000.jpg
 
Sorry for the thread revival, but yesterday I came across a discussion which raises a whole new explanation for increased noise at (very) wide apertures.....

http://photography-on-the.net/forum/showthread.php?t=950999&highlight=t-stop

I tried some experimental shots of my own. Now these were in overcast daylight, so not really a low light situation, and probably with a fair spectrum of light wavelengths, so not requiring a big white balance shift to correct for a spectral imbalance. The exposures here should be equivalent in terms of photons passing through the lens, but if the assertion in the article is accurate then quite possibly the sensor is not recording as many of them at f/1.4 as at f/4, requiring a secret ISO boost to compensate. One thing is certain - the shot at f/1.4 looks a lot noisier than the one at f/4. While DOF effects may well contribute to the appearance, I'm prepared to accept that there could be more going on here than our earlier discussions considered.

I've deliberately selected an area of the frame away from the centre, where vignetting may also have an impact, and I also chose to include areas which were OOF in order to further allow any noise issues to be revealed. The biggest tell-tale is in the fence at the rear, which is almost in focus, and upon which the noise is far more noticeable at f/1.4. The staked plant at the rear also shows a marked difference, even though they aren't very OOF in the f/1.4 shot. Food for thought....

<snip image>

It's vignetting Tim, and under exposure away from the centre. Canon 50 1.4 is two stops down in the corners, which is typical of very low f/numbers and shorter focal lengths. See here http://www.the-digital-picture.com/Reviews/Lens-Vignetting-Test-Results.aspx?Lens=115

I think we have talked before about the accuracy of f/numbers, which as you know are purely theoretical, and take no account of vignetting or any other real world factors.
 
I'd be a bit shocked if an EF lens showed vignetting of 2 stops on a 7D. Whatever vignetting there might be it's surely not going to be that severe. In any event, those crops are not from the corner, but just nearby. According to DXOMark the maximum vignetting for this lens on a 7D approaches 1 stop at f/1.4 - http://www.dxomark.com/index.php/Le....4-USM/(camera)/619/(cameraname)/Canon-EOS-7D

FWIW I have applied no adjustment in Lightroom. These are unedited raw comparison. Hopefully Lightroom is not applying different processing to the two raw files, including vignetting correction.

Here are comparable crops from the centre. The noise is still far more apparent in the wide open shot. Also note the difference in colour fidelity between the two images. They look equally bright to my eyes, but as well as the noise there is a difference in colours in the tomatoes...

20110818_185931_000.jpg



Oh, and for context here are the full images....

20110818_191335_000.jpg
 
Last edited:
Yes, about one stop vignetting on a 7D tallies with the-digital-picture link.

I see what you're getting at, and it's well known that sensors are sensitive to the angles at which light strikes them, but I also think your test image is very deceptive.

It needs to be an area of plain tone, possibly deliberately completely out of focus to eliminate the optical shortcomings at f/1.4?
 
Back
Top