Nerd Wars...Part 1

What's surprised me is not that people are confused - I expect that because the idea does go against the grain of the 'virtual film' paradigm that digital photography has developed. It's just the level of hostility amongst some, of which chuxter is the most extreme example, and has now made it his mission to expose my 'false religion'. It's not even as if anyone says that photographers must do it this way - we just say if you know this, you can do it this way, if you like.
he seems to have misunderstood what the demonstration was about. i looked at the example and immediately said, "that's not what it shows." it pretty clearly shows why it's better to boost ISO in-camera than it is to correct it after the fact.
I understood. I very much agree that the caption to Guillermo's crops is, as you say. I even suggested to Emil that a good one line summary of his verbose text might be:

" Analog gain results in less apparent noise than digital gain. "

I still think that is a good generalization, but Emil didn't reply. :-(
A bit overly broad generalization. Doesn't apply for instance to P65+ backs, as I showed in a recent post. Rather

" Analog gain results in less apparent noise than digital gain, up to a point of diminishing returns. "

The ISO point of diminishing returns X is camera-dependent, and I showed how to determine it using DxO's data. Beyond it, one may opt to retain more highlight headroom by using ISO X, without worrying that increased noise will result.

--
emil
--



http://theory.uchicago.edu/~ejm/pix/20d/
 
What's surprised me is not that people are confused - I expect that because the idea does go against the grain of the 'virtual film' paradigm that digital photography has developed. It's just the level of hostility amongst some, of which chuxter is the most extreme example, and has now made it his mission to expose my 'false religion'. It's not even as if anyone says that photographers must do it this way - we just say if you know this, you can do it this way, if you like.
he seems to have misunderstood what the demonstration was about. i looked at the example and immediately said, "that's not what it shows." it pretty clearly shows why it's better to boost ISO in-camera than it is to correct it after the fact.
I understood. I very much agree that the caption to Guillermo's crops is, as you say. I even suggested to Emil that a good one line summary of his verbose text might be:

" Analog gain results in less apparent noise than digital gain. "

I still think that is a good generalization, but Emil didn't reply. :-(

I react strongly to what Thomas said because I have heard it too often. I have gotten sensitized to messages like that...they immediately get my attention. Just as you immediately said (to your self, I presume...not having any other intelligent life-form handy), " that's not what it shows. ", I immediately thought, "here they go again trying to confuse everybody."
now, much of his OP is about how "technologists" (you, et al?) need to explain better. so perhaps by his own admission, he doesn't get it. and that's okay. this is pretty complicated stuff. i can't say i even get it completely after reading your posts.
So far, Thomas has never written anything that helped me understand. He doesn't write in a way that helps me...it's like he's more interested in using impressive words and complicated constructions (perhaps to impress others in his group) than making a serious attempt to help me and others improve our knowledge. I have learned from others, though.
I am sorry to hear that. Possibly I am not a great communicator, or my style of writing does not suit you. That does not however justify in any way the lies, distortions, bashing, accusations or any other part of your very poor behaviour. Even this thread had an ignoble motive. Happily, it's turned against you now it appears that some other people do understand. Hence, back you come, trying to justify yourself on the basis that you don't understand me. As I said before, pathetic.

--
thomas
 
What's surprised me is not that people are confused - I expect that because the idea does go against the grain of the 'virtual film' paradigm that digital photography has developed. It's just the level of hostility amongst some, of which chuxter is the most extreme example, and has now made it his mission to expose my 'false religion'. It's not even as if anyone says that photographers must do it this way - we just say if you know this, you can do it this way, if you like.
he seems to have misunderstood what the demonstration was about. i looked at the example and immediately said, "that's not what it shows." it pretty clearly shows why it's better to boost ISO in-camera than it is to correct it after the fact.
I understood. I very much agree that the caption to Guillermo's crops is, as you say. I even suggested to Emil that a good one line summary of his verbose text might be:

" Analog gain results in less apparent noise than digital gain. "

I still think that is a good generalization, but Emil didn't reply. :-(
A bit overly broad generalization. Doesn't apply for instance to P65+ backs, as I showed in a recent post. Rather

" Analog gain results in less apparent noise than digital gain, up to a point of diminishing returns. "

The ISO point of diminishing returns X is camera-dependent, and I showed how to determine it using DxO's data. Beyond it, one may opt to retain more highlight headroom by using ISO X, without worrying that increased noise will result.
The caveat being that the point of diminishing returns could be to the left of the ISO curve. This would happen if the read chain was designed so the late chain noises (ADC noise and so on) was negligible with respect to the early chain noise. Then there would be no point applying analog gain at all (as I think you pointed out somewhere)
--
thomas
 
It was exactly this phrase that I find confusing/amazing/bemusing.

"For Guillermo's 350D, that point of diminishing returns is somewhere between ISO 800 and 1600; for the D300, it's considerably lower."

Which I read as, you feel that the 350d is good to 1600 ISO and the D300 only up to a considerably lower ISO. Beyond that, the noise tradeoff is not worthwhile.
 
Here is the one we're interested in:



I shows what you described.

I use 800 ISO a lot on my D300. Subjectively, it has less noise at that setting than at 400 ISO. I think some objective tests (like the one here on dpr) have shown that too. I use my flashes a lot , even in daylight. A pragmatic side benefit to using 800 ISO is that I extend battery life in the flashes.
Fine for determining the point of diminishing returns. Shows it to be about 1600 on the 350D and about 400 on the D300.

Note however if one wants to compare camera noise between models, one should use the "print" tab from DxO rather than the "screen" tab as you have done. The latter disadvantages the D300 by about a third of a stop due to its higher resolution.
You taught me something here. I had seen those two tabs on the DxO screen, but stupidly thought it did something else! Like format the screen for printing or something. @#$% frogs!!! I looked at that other one and see what you are saying.

The graph qualitatively says that the D300 is a better low ISO performer, while the 350D is slightly better at high ISO.

Yes, it's expected that it would have lower noise at high ISO. The 350D has 37% lower pixel density than the D300. [sorry, I just could not help myself ;-)] The D300 has an unusual characteristic as far as noise is concerned...800 ISO is a "sweet spot"...800 ISO has less noise than the settings preceding it, like 400 ISO. This is true of both JPEG and NEF formats. On the DxO graphs (both the print and screen versions), the 800 ISO points are slightly below the "line". This indicates lower DR, which seems counter to the vast number of subjective opinions. Any thoughts?
However, this is where my caveat comes in -- Canon cameras exhibit more pattern noise than most other brands, so there is more to it than is simply shown in the graph -- the pattern noise will limit the image quality at high ISO, and so one may prefer the D300 still at these ISO even though it does not show as well on this particular performance metric.
Yes, I agree. The subjective appearance of the "Canon" noise and the "Nikon/Sony" noise is different. People react to them differently (ie, not everybody likes the way Nikon/Sony noise appears). As you know, we need both objective and subjective judgments.

I noticed that you used a different expression, above. In the past, you confused Harry and others when you said the D300 wasn't "useful" above 800 ISO. Now you have a different way of expressing that...you said, 400 ISO was " the point of diminishing returns " for the D300. Hmmm...

I think that's better, but let's see what Harry thinks. It's still a bit ambiguous...

--
Charlie Davis
Nikon 5700, Sony R1, Nikon D50, Nikon D300
HomePage: http://www.1derful.info
“...photography for and of itself – photographs taken
from the world as it is – are misunderstood as a
collection of random observations and lucky moments...
Paul Graham
 
It was exactly this phrase that I find confusing/amazing/bemusing.

"For Guillermo's 350D, that point of diminishing returns is somewhere between ISO 800 and 1600; for the D300, it's considerably lower."

Which I read as, you feel that the 350d is good to 1600 ISO and the D300 only up to a considerably lower ISO. Beyond that, the noise tradeoff is not worthwhile.
That is not the intent. Diminishing returns means that you will get very nearly as good results in terms of noise by using a lower ISO and push-processing in the raw converter, rather than continuing to raise the ISO. It does not mean that it is not worth using the camera for low-light photography, nor does it imply that the D300 is worse in that regard than the 350D. It simply means that the low-light image one gets is rather insensitive to the ISO setting in the camera beyond a certain point.

And in fact, I would draw the opposite conclusion -- that the fact that raising the ISO on the D300 doesn't help the noise much relative to using a lower ISO and push-processing, means that the camera electronics is much less of a burden in the overall noise budget as the ISO is lowered . Thus the D300 is a stellar performer in deep shadows at low ISO (the tonal range at that ISO where the camera's contribution to overall noise is most important).

--
emil
--



http://theory.uchicago.edu/~ejm/pix/20d/
 
On the DxO graphs (both the print and screen versions), the 800 ISO points are slightly below the "line". This indicates lower DR, which seems counter to the vast number of subjective opinions. Any thoughts?
The deviation is within the margin of error of the measurements. I would not give it significance in the absence of other evidence of something going on.
I noticed that you used a different expression, above. In the past, you confused Harry and others when you said the D300 wasn't "useful" above 800 ISO. Now you have a different way of expressing that...you said, 400 ISO was " the point of diminishing returns " for the D300. Hmmm...
I said ISO's above a certain point weren't "useful" because they didn't improve the S/N. We were talking about noise, weren't we? They might be useful for ancillary reasons, as DSPographer mentioned (not-too-dark review image in-camera, live view, etc; though all these could be fixed with better firmware); but not for improving noise performance.

And as I replied in another post, this does not mean the camera should not be used for low light photography or other situations where the exposure is low. It just means that the noise in the images obtained will be rather insensitive to which ISO was used to take them, so long as that ISO is above a certain point.

--
emil
--



http://theory.uchicago.edu/~ejm/pix/20d/
 
I have, as a wedding shooter occasionally had images a stop or so under at 800 or 1600 ISO.

When adjusted in RAW and compared to the correctly exposed shot at 1600 - 3200 ISO, the underexposed shots were absolutely terrible in comparison.

Now, I don't even think about keeping the ISO lower. If it needs 3200, so be it.
 
What's surprised me is not that people are confused - I expect that because the idea does go against the grain of the 'virtual film' paradigm that digital photography has developed. It's just the level of hostility amongst some, of which chuxter is the most extreme example, and has now made it his mission to expose my 'false religion'. It's not even as if anyone says that photographers must do it this way - we just say if you know this, you can do it this way, if you like.
he seems to have misunderstood what the demonstration was about. i looked at the example and immediately said, "that's not what it shows." it pretty clearly shows why it's better to boost ISO in-camera than it is to correct it after the fact.
I understood. I very much agree that the caption to Guillermo's crops is, as you say. I even suggested to Emil that a good one line summary of his verbose text might be:

" Analog gain results in less apparent noise than digital gain. "

I still think that is a good generalization, but Emil didn't reply. :-(
A bit overly broad generalization.
:-) I knew you'd think that. You are not a guy who makes generalizations easily.
Doesn't apply for instance to P65+ backs, as I showed in a recent post.
The exception that proves the rule? After all, how many P65+ backs did they sell? 791 perhaps? ;-) Whatever the number, it's miniscule and the Kodak technology in those MF sensors is SO different than all other sensors, it's not surprising or significant they are flyers to the broader data sset.
Rather

" Analog gain results in less apparent noise than digital gain, up to a point of diminishing returns. "
I instinctively shudder when you add 7 words to my perfectly composed 10 words...a 70% increase in word count, but it didn't add 7% validity, IMO.
The ISO point of diminishing returns X is camera-dependent, and I showed how to determine it using DxO's data. Beyond it, one may opt to retain more highlight headroom by using ISO X, without worrying that increased noise will result.
I want a real example. If I accept your approach that proclaims the D300 has "diminishing returns" after 400 ISO, how do I use that in my daily photography? Let's use your expanded generalization as a basis for an experiment. You said:

" Analog gain results in less apparent noise than digital gain, up to a point of diminishing returns. "

which for the D300 is 400 ISO.

Thus, if I was to take a series of shots with different sensitivity/gain/ISO settings with my D300 and compare them with another series of shots that were at the base sensitivity/gain/ISO, but with digital gain applied in PP...I would find that 1) the base ISO shots were (of course) identical, 2) the 400 ISO shots had less noise for the in-camera version, and 3) all the rest of the shots had less noise for the PP pictures. Right?

In making these pictures, I would use the fixed illumination model as used by Guillermo and Gabor. Hmmm...didn't I do that test already? No, not really...I'll work on that...

--
Charlie Davis
Nikon 5700, Sony R1, Nikon D50, Nikon D300
HomePage: http://www.1derful.info
“...photography for and of itself – photographs taken
from the world as it is – are misunderstood as a
collection of random observations and lucky moments...
Paul Graham
 
" Analog gain results in less apparent noise than digital gain. "

I still think that is a good generalization
A bit overly broad generalization.
:-) I knew you'd think that. You are not a guy who makes generalizations easily.
No, I am happy to make rules of thumb, but they should be accurate ones. If there is a broad class of applications, or a broad class of cameras, to which the rule does not apply (as is the case here), then the scope of the rule should be stated.
Doesn't apply for instance to P65+ backs, as I showed in a recent post.
The exception that proves the rule?
Hardly.
After all, how many P65+ backs did they sell? 791 perhaps? ;-) Whatever the number, it's miniscule and the Kodak technology in those MF sensors is SO different than all other sensors, it's not surprising or significant they are flyers to the broader data sset.
It's not that the sensor is in a boutique camera, it's that it's a CCD sensor. For instance, Nikon D200, Pentax K10D, Canon G11. Typically benefit very little from raising the ISO beyond a stop or so from base ISO when shooting raw.
I want a real example. If I accept your approach that proclaims the D300 has "diminishing returns" after 400 ISO, how do I use that in my daily photography?
I do nature photography. Light is often variable. I don't shy from setting ISO 800 and keeping it there if the light goes down, knowing that if I bring the image brightness up in post I will not be losing much if anything relative to a shot taken at ISO 1600 or 3200, and I will not have to worry about blown highlights. If the image looks noisy, it's because I had less light to work with, not because I had the camera set wrong.
Let's use your expanded generalization as a basis for an experiment. You said:

" Analog gain results in less apparent noise than digital gain, up to a point of diminishing returns. "

which for the D300 is 400 ISO.

Thus, if I was to take a series of shots with different sensitivity/gain/ISO settings with my D300 and compare them with another series of shots that were at the base sensitivity/gain/ISO, but with digital gain applied in PP...I would find that 1) the base ISO shots were (of course) identical, 2) the 400 ISO shots had less noise for the in-camera version, and 3) all the rest of the shots had less noise for the PP pictures. Right?
You get some benefit from bumping the ISO from base ISO (which is 200, or perhaps a bit lower IIRC) up to 400. The shots at ISO 400 should look little different from those above 400, apart from having more retained highlights if the scene needs highlight headroom.

--
emil
--



http://theory.uchicago.edu/~ejm/pix/20d/
 
Actually, I don't think it is complicated, but I do think I have difficulty getting it across. One reason is probably that I don't communicate as well as I might. Another though is that the facts get blurred amongst the hostile and devious debates which inevitably seem to go with it. Knowing that the debate is bound to come I try to phrase what I say carefully so that it isn't vulnerable to simple attacks. So I wouldn't say 'high ISO reduces noise', what I'd say is that 'high ISO reduces visible read noise'. However, this language doesn't really help people that much, and gets dismissed as technobabble. I think when I reduce the subject to its basics, it's pretty clear.
i think people just don't know what that means. i don't know what that means, even though you've clearly explained it here at least one, and what the effective difference actually is.
The extra visible noise you see in a low light image is almost entirely photon shot noise.
...or that.
You see it because the dynamic range of the display medium exceeds the dynamic range that the camera can give with that little light going into the sensor.
though this perhaps makes sense.
I can't see why you'd want to do that, unless you were looking for high noise in your image.
or, you just wanted a lot of black. either are acceptable goals.

i shot pictures for years that were about pronounced and noticeable JPEG compression as texture. and "grainy" is much more well-established as a look than that.
Otherwise, if you're looking for a dark image with plugged shadows, you're etter darkening ind adjusting the shadow curve in processing. In noise terms, there is never a reason to reduce the exposure.
unless you want the noise. which is fine too. and i agree that plugging the shadows in post is probably better, but some people are "in-camera" photographer. of course, you can always add noise in post, too. but that might not look the same.
That's pretty much what I do too. Except I've come round to just setting the exposure I want, set the ISO to get it to the right and shoot away (in M), just checking the meter in case I need to change the ISO. I just wish I had auto ISO (and that auto ISO's were properly designed as 'exposure priority' mode) it would make life simpler.
i believe most auto-ISO functions only go up, right? shouldn't that set the maximum ISO for you, and you set the minimum manually?

okay, so, for instance, tonight i was shooting candids at a high school prom. in the dark.

i set the exposure i wanted, either f/2.8 or f/1.8 (wide open, depending on the lens i was using), 1/5th of a second when it was lighter, close to 1 second when it was dark. popped on the flash to lock in the foreground and provide a little illumination here and there. from there, i set the ISO up to what i needed while keeping the atmospheric qualities i wanted . i didn't want it to look like daytime in the ballroom, you know? but i wanted the background to have some light. so, it wasn't exactly exposing to the right. ETTR just looked too painfully bright and didn't represent what i wanted. so, keeping the ISO lower, between 100 and 400, depending what the DJ was doing with the lights, actually resulted in more apparent noise in some images that were slightly under exposed. but, of course, only when the images were slightly under exposed.

so, in practice, i was really sort of doing what you recommend, but because i was trying to give the maximum exposure i could get away with, i was also keeping the ISO i needed to use relatively low. i know a lot of photographers would have just cranked it up to 1600 or 3200 and snapped away. and i know a lot more that would have just left it at base ISO, and used the flash to light everything .

of course, "completely in the dark" is kind of a nightmare situation.
 
" Analog gain results in less apparent noise than digital gain, up to a point of diminishing returns. "

which for the D300 is 400 ISO.
i hate to be stubborn about this, but it really depends a lot on exposure. i can shoot usable pictures on my d200 up to about 640 or 800, and maybe passable up to say 1000. but if i under-expose even slightly, the grain comes out of the woodwork. and the d200 has a horribly noisy sensor.

similarly, i'm sure if you look, you'll find posts here on the d300 being noisy @200 iso , its base. well, yeah, if you under expose it.
 
I understood. I very much agree that the caption to Guillermo's crops is, as you say. I even suggested to Emil that a good one line summary of his verbose text might be:

" Analog gain results in less apparent noise than digital gain. "

I still think that is a good generalization, but Emil didn't reply. :-(

I react strongly to what Thomas said because I have heard it too often. I have gotten sensitized to messages like that...they immediately get my attention. Just as you immediately said (to your self, I presume...not having any other intelligent life-form handy), " that's not what it shows. ", I immediately thought, "here they go again trying to confuse everybody."
i strongly suggest that you calm down a little bit, and try to wrap your head around it a bit. i know this is complicated, and i'm trying to understand the point they're making -- the obvious one you would almost certainly agree with -- from their technical standpoint. it's the technical stuff that seems to be throwing you off. i don't think you'd be fighting if they told you that "under exposure affects noise a lot more than ISO." i hear that particular mantra, in various forms, quite frequently on this board. and that's what the demonstration in the OP shows -- although it's not clear from the picture alone that exposure is different.
So far, Thomas has never written anything that helped me understand. He doesn't write in a way that helps me...it's like he's more interested in using impressive words and complicated constructions (perhaps to impress others in his group) than making a serious attempt to help me and others improve our knowledge. I have learned from others, though.
perhaps there technical distinctions that are actually relevant. he seems like he's trying to help us understand, and indeed, i'm getting at least some of what he's saying.
 
but every time i've over exposed even just a little, those highlights are gone.
I'm not going to argue with that, I'd guess it depends a lot on the specific camera and JPEG settings.
probably, yes. but i've just never seen one that actually has a nice shoulder where you can recover highlight detail from, because, as you say, digital is linear.

i'm willing, and indeed incredibly interested in changing my mind on this matter, however. i would love to see something on digital pull-process, especially if it increases DR. i would be all over that technique like nobody's business.
 
Just concentrate on one thing for this post
So I wouldn't say 'high ISO reduces noise', what I'd say is that 'high ISO reduces visible read noise'. However, this language doesn't really help people that much, and gets dismissed as technobabble. I think when I reduce the subject to its basics, it's pretty clear.
i think people just don't know what that means. i don't know what that means, even though you've clearly explained it here at least one, and what the effective difference actually is.
For the photographer, there is no difference. For some who know a bit of electronics there is a difference.

For the photographer, what is usually meant by 'noise' is the speckly stuff in the picture they look at. For them the statement is true without further qualification++. However, if someone with a little electronics knowledge+++ looks at the statement, they object, because they understand that high ISO involves more amplification, and that amplification will increase the amplitude of the noise measured in volts, or measured in analog to digital units, which is true. What's going to happen in processing, though, is that the amplified signal gets mapped back to a normalised range of output tonality, which has the effect of reducing the visible noise in the final output down below where it would have been without the ISO adjustment. Therefore I need to qualify my statement with 'visible'. I've chosen that in the hope that photographers might still understand it while it still satisfies the technical quibble.

Now, the next point is to do with the type of noise we're talking about. There are two sources of noise by and large. The photon shot noise, which is caused by the random arrival of photons at the sensor, and the read noise, which is the electronic noise generated within the camera. Photographically speaking, the shot noise is generally the noise in bright and mid tone areas of the image, shot noise is the shadow noise. The amount of the shot noise in the image++++ depends absolutely on the exposure, so as the exposure reduces the visible shot noise increases. Generally, though the read noise decreases as ISO setting increases, and this is what I'm talking about, so I make sure I refer to 'read noise' and not just noise.

++ although it still confuses those who believe that there is an unbreakable link between ISO setting and selected exposure. To make it absolutely clear for these people one should say also for the same exposure , although the risk then is then that many, like chuxter, misunderstand exposure to include the ISO setting as well incident light, aperture and shutter speed. My good friend Joe James talks about 'exposure' (excluding ISO) and 'apparent exposure' (including ISO) to distinguish them, but I'm not so sure that makes it any less confusing for those who don't know what exposure means, and it just irritates those that do.

+++ the problem here is those with a little knowledge. Those with a lot tend to understand from the context that one is talking about the photographer's perspective.

++++ and even here I need to make a qualification, and say 'so far as the photographer is concerned', because the amount of shot noise goes up as exposure increases, but the noise to signal ratio goes down , and in terms of visible noise in the image, it's the noise to signal ratio that you see.
--
thomas
 
I can't see why you'd want to do that, unless you were looking for high noise in your image.
or, you just wanted a lot of black. either are acceptable goals.
The best way to achieve a lot of black is to do so when you make your processing (tone mapping) choices. As I suggested, lowering exposure only increases visible noise, so, if noise (or lack of it) is a concern, you shouldn't reduce exposure to make the output image darker, you should do that in processing.
i shot pictures for years that were about pronounced and noticeable JPEG compression as texture. and "grainy" is much more well-established as a look than that.
Sure, all the advice goes out the window if you want effects to do with exploiting side effects of specific workflows. That's perfectly legitimate, and if you need a particular workflow to do it, go with it.
Otherwise, if you're looking for a dark image with plugged shadows, you're etter darkening ind adjusting the shadow curve in processing. In noise terms, there is never a reason to reduce the exposure.
unless you want the noise. which is fine too. and i agree that plugging the shadows in post is probably better, but some people are "in-camera" photographer. of course, you can always add noise in post, too. but that might not look the same.
That's pretty much what I do too. Except I've come round to just setting the exposure I want, set the ISO to get it to the right and shoot away (in M), just checking the meter in case I need to change the ISO. I just wish I had auto ISO (and that auto ISO's were properly designed as 'exposure priority' mode) it would make life simpler.
i believe most auto-ISO functions only go up, right? shouldn't that set the maximum ISO for you, and you set the minimum manually?
I don't have auto ISO on either of my cameras, and I haven't looked in detail at what any of the auto ISO functions does, apart from the need to do so for discussions here. What I'd want is to set my f-number and shutter speed, and have the camera adjust the ISO to suit as the light changes, also allowing EC so I can choose to push the histogram to the right (as against what the camera chooses as its preferred mapping from exposure to raw file values). So far as I can see no camera offers that. For 'manual' what I'd like is to be able to put ISO onto one of the control wheels. Once I've chosen f-number and shutter speed, I'm much more likely to want to change the ISO than either of those, then probably the f-number (for different DoF effects) and lowest priority the shutter speed. More particularly, I'm only likely to change the shutter speed when going into a completely different sequence of shots, so I can afford a more clunky control sequence to do it.
okay, so, for instance, tonight i was shooting candids at a high school prom. in the dark.

i set the exposure i wanted, either f/2.8 or f/1.8 (wide open, depending on the lens i was using), 1/5th of a second when it was lighter, close to 1 second when it was dark.
That's a bit I don't understand. If you can live with 1 second, why not use it for everything, whether the light is brighter or not? The problem with both for candids is that you'll find many shots spoiled by subject movement. This is the time when FF really comes into play, because with the big sensor and a large aperture lens, you can gain more than two stops of light on the sensor over what you had to live with.
popped on the flash to lock in the foreground and provide a little illumination here and there. from there, i set the ISO up to what i needed while keeping the atmospheric qualities i wanted . i didn't want it to look like daytime in the ballroom, you know?
Sure, but whether it looks like daytime in the ballroom or not isn't a matter of exposure if you're shooting raw, it's a matter of how you choose to process. Here's an example. Here's an image shot in mixed light (natural + artificial + fill in ring flash) as the ex-camera jpeg.





Here's its histogram, so you can see it is a reasonable choice of exposure (the light through the window has blown, but that's reasonable in this context)





It's entirely due to processing whether this is presented as a moody low light image





or a brighter daylight image





The only difference between these images is the tone mapping selected in raw conversion (although I have downsampled them all from the native resolution). If I had chosen to make the 'low light' image by reducing the exposure, it would have had much more visible noise.
but i wanted the background to have some light. so, it wasn't exactly exposing to the right. ETTR just looked too painfully bright and didn't represent what i wanted. so, keeping the ISO lower, between 100 and 400, depending what the DJ was doing with the lights, actually resulted in more apparent noise in some images that were slightly under exposed. but, of course, only when the images were slightly under exposed.
The apparent noise being from the lower exposure, rather than the 'under' exposure.
so, in practice, i was really sort of doing what you recommend, but because i was trying to give the maximum exposure i could get away with, i was also keeping the ISO i needed to use relatively low. i know a lot of photographers would have just cranked it up to 1600 or 3200 and snapped away. and i know a lot more that would have just left it at base ISO, and used the flash to light everything .

of course, "completely in the dark" is kind of a nightmare situation.
Sure, but the best way to handle it in IMO is to decide beforehand how big an exposure you can tolerate (in terms of DoF and motion blur) then go with it, the ISO follows behind.
--
thomas
 
decide beforehand how big an exposure you can tolerate
:)

I swear....... none of this makes much sense if any. It seems as though you are trying to sell this as like you have reinvented the wheel. Expose to the right of the histogram and recover the highlights in PP at the lowest possible ISO. Hasn't this been always been the way ?

And the samples you show won't convince anyone of the correctness of your technique is this is the best result you can show.

--
Nick in Shanghai.
 
" Analog gain results in less apparent noise than digital gain. "

I still think that is a good generalization
A bit overly broad generalization.
:-) I knew you'd think that. You are not a guy who makes generalizations easily.
No, I am happy to make rules of thumb, but they should be accurate ones.
That's what I meant by "easily". There are degrees of accuracy. You strive for more accuracy in your "rules-of-thumb" than most people do. I suspect that sometimes your generalizations resemble a book. ;-)
If there is a broad class of applications, or a broad class of cameras, to which the rule does not apply (as is the case here), then the scope of the rule should be stated.
I agree with that. But always we should strive for clear, compact language..."bullet points" must fit on one line of rather large text. There can be supporting and illuminating words of any size.

You and Thomas had "bullet points" something like, "Higher ISO gives less noise" and then wrote "books" to justify that position, I agree with you that it would have been better to state, "Higher ISO gives less noise in some cases ", then clearly describe the cases and explain why that happens. I was disappointed and disturbed that this didn't happen.
Doesn't apply for instance to P65+ backs, as I showed in a recent post.
The exception that proves the rule?
Hardly.
After all, how many P65+ backs did they sell? 791 perhaps? ;-) Whatever the number, it's miniscule and the Kodak technology in those MF sensors is SO different than all other sensors, it's not surprising or significant they are flyers to the broader data sset.
It's not that the sensor is in a boutique camera, it's that it's a CCD sensor. For instance, Nikon D200, Pentax K10D, Canon G11. Typically benefit very little from raising the ISO beyond a stop or so from base ISO when shooting raw.
Interesting "bullet point": " CCD sensors benefit little from ISO gain ".
I want a real example. If I accept your approach that proclaims the D300 has "diminishing returns" after 400 ISO, how do I use that in my daily photography?
I do nature photography. Light is often variable. I don't shy from setting ISO 800 and keeping it there if the light goes down, knowing that if I bring the image brightness up in post I will not be losing much if anything relative to a shot taken at ISO 1600 or 3200, and I will not have to worry about blown highlights. If the image looks noisy, it's because I had less light to work with, not because I had the camera set wrong.
Let's use your expanded generalization as a basis for an experiment. You said:

" Analog gain results in less apparent noise than digital gain, up to a point of diminishing returns. "

which for the D300 is 400 ISO.

Thus, if I was to take a series of shots with different sensitivity/gain/ISO settings with my D300 and compare them with another series of shots that were at the base sensitivity/gain/ISO, but with digital gain applied in PP...I would find that 1) the base ISO shots were (of course) identical, 2) the 400 ISO shots had less noise for the in-camera version, and 3) all the rest of the shots had less noise for the PP pictures. Right?
You get some benefit from bumping the ISO from base ISO (which is 200, or perhaps a bit lower IIRC) up to 400.
The camera calls the base sensitivity setting 200 ISO, but the data from DxO (and others) indicates that when set to 200 ISO, the D300's sensitivity is actually more like 150 ISO.
The shots at ISO 400 should look little different from those above 400, apart from having more retained highlights if the scene needs highlight headroom.
That's the strange part. The 800 ISO shots look better, noise wise, than 400 ISO shots do. Some people think that Nikon is applying some mild variety of NR to NEF images above 800 ISO...others discount that.

--
Charlie Davis
Nikon 5700, Sony R1, Nikon D50, Nikon D300
HomePage: http://www.1derful.info
“...photography for and of itself – photographs taken
from the world as it is – are misunderstood as a
collection of random observations and lucky moments...
Paul Graham
 
I understood. I very much agree that the caption to Guillermo's crops is, as you say. I even suggested to Emil that a good one line summary of his verbose text might be:

" Analog gain results in less apparent noise than digital gain. "

I still think that is a good generalization, but Emil didn't reply. :-(

I react strongly to what Thomas said because I have heard it too often. I have gotten sensitized to messages like that...they immediately get my attention. Just as you immediately said (to your self, I presume...not having any other intelligent life-form handy), " that's not what it shows. ", I immediately thought, "here they go again trying to confuse everybody."
i strongly suggest that you calm down a little bit, and try to wrap your head around it a bit. i know this is complicated, and i'm trying to understand the point they're making -- the obvious one you would almost certainly agree with -- from their technical standpoint. it's the technical stuff that seems to be throwing you off. i don't think you'd be fighting if they told you that "under exposure affects noise a lot more than ISO." i hear that particular mantra, in various forms, quite frequently on this board. and that's what the demonstration in the OP shows -- although it's not clear from the picture alone that exposure is different.
Although it's perhaps difficult to tell from my posts, I'm calm. Your last line is the crux of the problem...there were not proper explanations of what the pictures really showed.
So far, Thomas has never written anything that helped me understand. He doesn't write in a way that helps me...it's like he's more interested in using impressive words and complicated constructions (perhaps to impress others in his group) than making a serious attempt to help me and others improve our knowledge. I have learned from others, though.
perhaps there technical distinctions that are actually relevant. he seems like he's trying to help us understand, and indeed, i'm getting at least some of what he's saying.
I agree completely that Thomas is trying ... ;-)

Lately, he seems to have gotten the message that he needs to use fewer, better constructed sentences if he wants to be a part of the solution.

--
Charlie Davis
Nikon 5700, Sony R1, Nikon D50, Nikon D300
HomePage: http://www.1derful.info
“...photography for and of itself – photographs taken
from the world as it is – are misunderstood as a
collection of random observations and lucky moments...
Paul Graham
 
On the DxO graphs (both the print and screen versions), the 800 ISO points are slightly below the "line". This indicates lower DR, which seems counter to the vast number of subjective opinions. Any thoughts?
The deviation is within the margin of error of the measurements. I would not give it significance in the absence of other evidence of something going on.
I noticed that you used a different expression, above. In the past, you confused Harry and others when you said the D300 wasn't "useful" above 800 ISO. Now you have a different way of expressing that...you said, 400 ISO was " the point of diminishing returns " for the D300. Hmmm...
I said ISO's above a certain point weren't "useful" because they didn't improve the S/N. We were talking about noise, weren't we? They might be useful for ancillary reasons, as DSPographer mentioned (not-too-dark review image in-camera, live view, etc; though all these could be fixed with better firmware); but not for improving noise performance.

And as I replied in another post, this does not mean the camera should not be used for low light photography or other situations where the exposure is low. It just means that the noise in the images obtained will be rather insensitive to which ISO was used to take them, so long as that ISO is above a certain point.
BP: " Image noise is constant above some ISO gain value "...

But don't we also need to say that this is only true if the "exposure" is "fixed" in PP?

--
Charlie Davis
Nikon 5700, Sony R1, Nikon D50, Nikon D300
HomePage: http://www.1derful.info
“...photography for and of itself – photographs taken
from the world as it is – are misunderstood as a
collection of random observations and lucky moments...
Paul Graham
 

Keyboard shortcuts

Back
Top