ISOless sensors, read noise and photography - many questions!

Started Oct 3, 2013 | Discussions
Steen Bay
Veteran MemberPosts: 6,397
Like?
Re: full circle: "just raise ISO, when necessary!"
In reply to bobn2, Oct 7, 2013

bobn2 wrote:

boardsy wrote:

From http://www.sensorgen.info/SonyNEX-C3.html

- no visible benefit in keeping ISO at the low read noise minimum (800 for this sensor), and possibly struggling with a dim view/LCD in camera, and raising brightness later.

I'd be very interested if bobn2, Great Bustard, Gollywop, Detail Man et al would agree?

It depends on the scene. If the scene has a higher tonal range than can be handled by the 'correct' ISO, then there is advantage to using the lower ISO.

-- hide signature --

Bob

Could be argued that the ISO isn't 'correct' if important highlights are blown. Then the 'correct' ISO would be a lower ISO where the highlights are protected.

Reply   Reply with quote   Complain
boardsy
Senior MemberPosts: 2,204Gear list
Like?
Re: full circle: "just raise ISO, when necessary!"
In reply to Steen Bay, Oct 7, 2013

Steen Bay wrote:

bobn2 wrote:

boardsy wrote:

From http://www.sensorgen.info/SonyNEX-C3.html

- no visible benefit in keeping ISO at the low read noise minimum (800 for this sensor), and possibly struggling with a dim view/LCD in camera, and raising brightness later.

I'd be very interested if bobn2, Great Bustard, Gollywop, Detail Man et al would agree?

It depends on the scene. If the scene has a higher tonal range than can be handled by the 'correct' ISO, then there is advantage to using the lower ISO.

-- hide signature --

Bob

Could be argued that the ISO isn't 'correct' if important highlights are blown. Then the 'correct' ISO would be a lower ISO where the highlights are protected.

@ bobn2 - yes, that's been mentioned already, and part of what I assumed by "when necessary" - so raise ISO as required for a bright image with protected highlights (if they're a priority for the shot - better-exposed shadows could be more important). But aside from protecting highlights it seems here's no other benefit to keeping it low and raising brightness in RAW later.

@ Steen Bay - isn't that pretty much what bobn2 is saying here? I.e. if the EV=0 suggests a high ISO ('correct' ISO according to the metering) that blows a few highlights, the 'really correct' ISO should be lower (if highlights are a priority for the shot - better-exposed shadows could be more important).

-- hide signature --

Regards,
Alan
my Flickr

Reply   Reply with quote   Complain
Steen Bay
Veteran MemberPosts: 6,397
Like?
Re: full circle: "just raise ISO, when necessary!"
In reply to boardsy, Oct 7, 2013

boardsy wrote:

Steen Bay wrote:

bobn2 wrote:

boardsy wrote:

From http://www.sensorgen.info/SonyNEX-C3.html

- no visible benefit in keeping ISO at the low read noise minimum (800 for this sensor), and possibly struggling with a dim view/LCD in camera, and raising brightness later.

I'd be very interested if bobn2, Great Bustard, Gollywop, Detail Man et al would agree?

It depends on the scene. If the scene has a higher tonal range than can be handled by the 'correct' ISO, then there is advantage to using the lower ISO.

-- hide signature --

Bob

Could be argued that the ISO isn't 'correct' if important highlights are blown. Then the 'correct' ISO would be a lower ISO where the highlights are protected.

@ bobn2 - yes, that's been mentioned already, and part of what I assumed by "when necessary" - so raise ISO as required for a bright image with protected highlights (if they're a priority for the shot - better-exposed shadows could be more important). But aside from protecting highlights it seems here's no other benefit to keeping it low and raising brightness in RAW later.

@ Steen Bay - isn't that pretty much what bobn2 is saying here? I.e. if the EV=0 suggests a high ISO ('correct' ISO according to the metering) that blows a few highlights, the 'really correct' ISO should be lower (if highlights are a priority for the shot - better-exposed shadows could be more important).

-- hide signature --

Regards,
Alan
my Flickr

Using a lower ISO in order to protect the highlights won't affect the shadow noise if the actual exposure (f-stop, shutter speed and lighting) is the same. But yes, think we agree here. No need to use an ISO that's so high that important highlights are blown, and no need either to use an ISO that's so low that the highlights are e.g. 3 stops below the clipping point (which would give unnecessarily dark JPEGs).

Reply   Reply with quote   Complain
boardsy
Senior MemberPosts: 2,204Gear list
Like?
Re: full circle: "just raise ISO, when necessary!"
In reply to Steen Bay, Oct 7, 2013

Steen Bay wrote:

boardsy wrote:

Steen Bay wrote:

bobn2 wrote:

boardsy wrote:

From http://www.sensorgen.info/SonyNEX-C3.html

- no visible benefit in keeping ISO at the low read noise minimum (800 for this sensor), and possibly struggling with a dim view/LCD in camera, and raising brightness later.

I'd be very interested if bobn2, Great Bustard, Gollywop, Detail Man et al would agree?

It depends on the scene. If the scene has a higher tonal range than can be handled by the 'correct' ISO, then there is advantage to using the lower ISO.

-- hide signature --

Bob

Could be argued that the ISO isn't 'correct' if important highlights are blown. Then the 'correct' ISO would be a lower ISO where the highlights are protected.

@ bobn2 - yes, that's been mentioned already, and part of what I assumed by "when necessary" - so raise ISO as required for a bright image with protected highlights (if they're a priority for the shot - better-exposed shadows could be more important). But aside from protecting highlights it seems here's no other benefit to keeping it low and raising brightness in RAW later.

@ Steen Bay - isn't that pretty much what bobn2 is saying here? I.e. if the EV=0 suggests a high ISO ('correct' ISO according to the metering) that blows a few highlights, the 'really correct' ISO should be lower (if highlights are a priority for the shot - better-exposed shadows could be more important).

-- hide signature --

Regards,
Alan
my Flickr

Using a lower ISO in order to protect the highlights won't affect the shadow noise if the actual exposure (f-stop, shutter speed and lighting) is the same.

Yes, of course! (Unless ISO is part of exp... Sorry, unless ISO is being used to influence aperture or shutter speed.)

But yes, think we agree here. No need to use an ISO that's so high that important highlights are blown, and no need either to use an ISO that's so low that the highlights are e.g. 3 stops below the clipping point (which would give unnecessarily dark JPEGs).

Yes - I'd thought in the wake of the ISO wars that under-exposing when necessary, keeping ISO to a minimum best read noise level (likely 200/400/800) and raising brightness in RAW was somehow preferable to high ISO wrt noise, but it doesn't seem so, not for the NEX sensor anyhow.

-- hide signature --

Regards,
Alan
my Flickr

Reply   Reply with quote   Complain
bobn2
Forum ProPosts: 32,212
Like?
Re: full circle: "just raise ISO, when necessary!"
In reply to boardsy, Oct 7, 2013

boardsy wrote:

Steen Bay wrote:

bobn2 wrote:

boardsy wrote:

From http://www.sensorgen.info/SonyNEX-C3.html

- no visible benefit in keeping ISO at the low read noise minimum (800 for this sensor), and possibly struggling with a dim view/LCD in camera, and raising brightness later.

I'd be very interested if bobn2, Great Bustard, Gollywop, Detail Man et al would agree?

It depends on the scene. If the scene has a higher tonal range than can be handled by the 'correct' ISO, then there is advantage to using the lower ISO.

-- hide signature --

Bob

Could be argued that the ISO isn't 'correct' if important highlights are blown. Then the 'correct' ISO would be a lower ISO where the highlights are protected.

@ bobn2 - yes, that's been mentioned already, and part of what I assumed by "when necessary" - so raise ISO as required for a bright image with protected highlights (if they're a priority for the shot - better-exposed shadows could be more important). But aside from protecting highlights it seems here's no other benefit to keeping it low and raising brightness in RAW later.

That's another 'it depends'. It depends on how you 'raise brightness in raw' and how the camera does it. On a computer, potentially you can do it with high quality extended word or floating point arithmetic which at least in theory could produce better results than the 16bit integer arithmetic commonly applied in image processing pipeline hardware. That of course doesn't mean that any particular raw processing program is guaranteed to perform better than any particular camera.

-- hide signature --

Bob

Reply   Reply with quote   Complain
boardsy
Senior MemberPosts: 2,204Gear list
Like?
Re: full circle: "just raise ISO, when necessary!"
In reply to bobn2, Oct 7, 2013

bobn2 wrote:

boardsy wrote:

Steen Bay wrote:

bobn2 wrote:

boardsy wrote:

From http://www.sensorgen.info/SonyNEX-C3.html

- no visible benefit in keeping ISO at the low read noise minimum (800 for this sensor), and possibly struggling with a dim view/LCD in camera, and raising brightness later.

I'd be very interested if bobn2, Great Bustard, Gollywop, Detail Man et al would agree?

It depends on the scene. If the scene has a higher tonal range than can be handled by the 'correct' ISO, then there is advantage to using the lower ISO.

-- hide signature --

Bob

Could be argued that the ISO isn't 'correct' if important highlights are blown. Then the 'correct' ISO would be a lower ISO where the highlights are protected.

@ bobn2 - yes, that's been mentioned already, and part of what I assumed by "when necessary" - so raise ISO as required for a bright image with protected highlights (if they're a priority for the shot - better-exposed shadows could be more important). But aside from protecting highlights it seems here's no other benefit to keeping it low and raising brightness in RAW later.

That's another 'it depends'. It depends on how you 'raise brightness in raw' and how the camera does it. On a computer, potentially you can do it with high quality extended word or floating point arithmetic which at least in theory could produce better results than the 16bit integer arithmetic commonly applied in image processing pipeline hardware. That of course doesn't mean that any particular raw processing program is guaranteed to perform better than any particular camera.

-- hide signature --

Bob

Ok - all we can say for sure, then, is that Russell has demonstrated that LR (at least his use of it) isn't any better at raising the brightness of a NEX C3's under-exposed ISO800 shot than the C3's ISO3200.

Is there any particular raw processing program that is known to use high quality extended word or floating point arithmetic?

-- hide signature --

Regards,
Alan
my Flickr

Reply   Reply with quote   Complain
Jack Hogan
Veteran MemberPosts: 4,018
Like?
Re: full circle: "just raise ISO, when necessary!"
In reply to boardsy, Oct 7, 2013

boardsy wrote: Is there any particular raw processing program that is known to use high quality extended word or floating point arithmetic?

RawTherapee V4 and up. It being open source and free it is an excellent testing ground although not as polished as some commercial converters for a day-to-day workflow.

Reply   Reply with quote   Complain
boardsy
Senior MemberPosts: 2,204Gear list
Like?
Re: full circle: "just raise ISO, when necessary!"
In reply to Jack Hogan, Oct 7, 2013

Jack Hogan wrote:

boardsy wrote: Is there any particular raw processing program that is known to use high quality extended word or floating point arithmetic?

RawTherapee V4 and up. It being open source and free it is an excellent testing ground although not as polished as some commercial converters for a day-to-day workflow.

Thanks - I remember RT was well-regarded for its Beyer de-mosaicing algorithm; do you have any idea if it's better than Adobe Camera Raw for brightness/exp. comp./shadow noise? Must try it again.

-- hide signature --

Regards,
Alan
my Flickr

Reply   Reply with quote   Complain
Jack Hogan
Veteran MemberPosts: 4,018
Like?
Re: full circle: "just raise ISO, when necessary!"
In reply to boardsy, Oct 7, 2013

boardsy wrote:

Thanks - I remember RT was well-regarded for its Beyer de-mosaicing algorithm; do you have any idea if it's better than Adobe Camera Raw for brightness/exp. comp./shadow noise?

As far as I can tell, all raw converters are more or less equivalent in terms of what they do and pretty close in 'IQ'. The stuff isn't rocket science:

  • squeeze captured information into 16 bits of a colorimetric color space per your wishes (white balance, black and white point)
  • have a number of appropriate pre-cooked rendering recipes that look good (brightness/contrast curves, saturation, sharpening)
  • deal with CA, moirè and the like

That's pretty well it, the rest (like global or local corrections to squeeze in tones that would otherwise fall off either side of the histogram on output, lens corrections, noise reduction, sharpening) you can do in any decent editor - in most cases much better than a raw converter, also thanks to dedicated plugins (e.g. InFocus, Denoise, Tonal Contrast).

In the end it's just a matter of how convenient they make it and how well they address your preferred workflow. I find that Capture NX2 fits well my Nikonian low-volume high-IQ approach with a minimum of fuss so I like it.  Others have different approaches and prefer other converters. But as far as brightness/exp. comp./shadow noise is concerned, they are all pretty well equivalent imho.

Jack

Reply   Reply with quote   Complain
RussellInCincinnati
Senior MemberPosts: 3,197
Like?
am happy with your starting this helpful thread
In reply to boardsy, Oct 25, 2013

original poster boardsy wrote: ...if the sensor is effectively ISOless then is base ISO always optimal ...or are you as well off raising it in camera (providing highlights are preserved) rather than in RAW/pp?

Boardsy my takeaway a few weeks ago from our tortuous thread, was to notice that a Nex C3 sensor has minimum read noise at around ISO 800. Which you made so easy to see by including a handy noise graph in your original post.

Now the C3's read noise is not a lot less at ISOs besides 800, but hey why not use the sensor at that most efficient-in-some-ways ISO, even if you theoretically "need" a higher ISO to make the little preview look "good"? Since you can so easily "raise the ISO" (i.e. move an "exposure" slider in say Lightroom) after the fact. The only downside is that the LCD does indeed get dim.

Well have finally brought that concept into the field. In the scene below, the flight was going to be at ISO 3200 and 1/10th of a second. Around F/8 to get complete cross-frame clarity. Everything looked fine in the electronic viewfinder (i.e. clip-on eyepiece-magnified LCD).

But then remembered at the last minute "hey, why sweat the exact exposure when I'm worried about blowing out highlights in the white fabrics? Why worry about how many bracketing exposures needed, at what intervals, just in case? Why not move the ISO back to 800, the screen looks kind of dark but everything's still perfectly viewable/manual focusable, and then can have the luxury of deciding in post-processing exactly how bright to raise the scene highlights? Instead of letting the camera firmware possibly over-amplify the highlight brightnesses in an attempt to make the screen look nicer in the field?"

So did it, moved the ISO back to 800, screen got dark but it's not like everything disappeared. Manual focused, took the single exposure, and went home. Then played around with the exposure and highlight sliders in Lightroom for, don't know, about 15 seconds or so. What a thrill to get a better result with exactly the highlight preservation or blowout desired, decided after the fact in the optimal editing environment of the "darkroom." No exposure bracketing in the field, welcome simplification when there are about a zillion other things to think about and adjust. I.e. bracketing completely pointless since there is absolutely no chance of blowing highlights on a greatly underexposed image, and the "final exposure" is set at editing time. Really in a sense the rawest of all possible raw files.

Now leaving your camera at some lower ISO when you "need" a higher ISO only makes sense if (a) the ISO you leave the camera at indeed has better or no worse read noise and (b) you have verified that everything looks exactly the same or better whether you let the camera firmware "amplify" the brightnesses for you in the raw file, or you let your post-processing software do the "amplifying" arithmetically. Both of which have verified are true in the Nex 3/Lightroom case--even if you need ISO 12800, you can get "the same" or better results by post-processing an image from a camera set to ISO 800. Thus the Nex C3 is a good deal of the way towards being Iliah Borg's "true ISO-less camera."

Upshot of all this is to make overexposing a low-light photo, or bothering to bracket it, is quaint history.

Reply   Reply with quote   Complain
RussellInCincinnati
Senior MemberPosts: 3,197
Like?
oops posted non-final version of sample photo
In reply to RussellInCincinnati, Oct 25, 2013

ISO 3200 indicated as needed in camera, but photo taken "underexposed" at ISO 800, viewfinder dim. Then image "pushed' in post-processing with the Lightroom exposure slider set to +1.88 or so.

Reply   Reply with quote   Complain
boardsy
Senior MemberPosts: 2,204Gear list
Like?
so not quite back where we started!
In reply to RussellInCincinnati, Oct 30, 2013

RussellInCincinnati wrote:

original poster boardsy wrote: ...if the sensor is effectively ISOless then is base ISO always optimal ...or are you as well off raising it in camera (providing highlights are preserved) rather than in RAW/pp?

Boardsy my takeaway a few weeks ago from our tortuous thread, was to notice that a Nex C3 sensor has minimum read noise at around ISO 800. Which you made so easy to see by including a handy noise graph in your original post.

Now the C3's read noise is not a lot less at ISOs besides 800, but hey why not use the sensor at that most efficient-in-some-ways ISO, even if you theoretically "need" a higher ISO to make the little preview look "good"? Since you can so easily "raise the ISO" (i.e. move an "exposure" slider in say Lightroom) after the fact. The only downside is that the LCD does indeed get dim.

Well have finally brought that concept into the field. In the scene below, the flight was going to be at ISO 3200 and 1/10th of a second. Around F/8 to get complete cross-frame clarity. Everything looked fine in the electronic viewfinder (i.e. clip-on eyepiece-magnified LCD).

But then remembered at the last minute "hey, why sweat the exact exposure when I'm worried about blowing out highlights in the white fabrics? Why worry about how many bracketing exposures needed, at what intervals, just in case? Why not move the ISO back to 800, the screen looks kind of dark but everything's still perfectly viewable/manual focusable, and then can have the luxury of deciding in post-processing exactly how bright to raise the scene highlights? Instead of letting the camera firmware possibly over-amplify the highlight brightnesses in an attempt to make the screen look nicer in the field?"

So did it, moved the ISO back to 800, screen got dark but it's not like everything disappeared. Manual focused, took the single exposure, and went home. Then played around with the exposure and highlight sliders in Lightroom for, don't know, about 15 seconds or so. What a thrill to get a better result with exactly the highlight preservation or blowout desired, decided after the fact in the optimal editing environment of the "darkroom." No exposure bracketing in the field, welcome simplification when there are about a zillion other things to think about and adjust. I.e. bracketing completely pointless since there is absolutely no chance of blowing highlights on a greatly underexposed image, and the "final exposure" is set at editing time. Really in a sense the rawest of all possible raw files.

Now leaving your camera at some lower ISO when you "need" a higher ISO only makes sense if (a) the ISO you leave the camera at indeed has better or no worse read noise and (b) you have verified that everything looks exactly the same or better whether you let the camera firmware "amplify" the brightnesses for you in the raw file, or you let your post-processing software do the "amplifying" arithmetically. Both of which have verified are true in the Nex 3/Lightroom case--even if you need ISO 12800, you can get "the same" or better results by post-processing an image from a camera set to ISO 800. Thus the Nex C3 is a good deal of the way towards being Iliah Borg's "true ISO-less camera."

Upshot of all this is to make overexposing a low-light photo, or bothering to bracket it, is quaint history.

This is an interesting result Russell - did you need to handle noise carefully in LR or just use defaults?

I'd still wonder about ""the same" or better results by post-processing an image from a camera set to ISO 800" - if the same, why bother (unless of course to preserve crucial highlights, as above!), if better to "do ISO in pp" then it's worth the trouble to focus and frame with a dim LCD... hmmm... this may be a NEX-specific issue, but maybe use a raised flash to trigger the always-bright LCD, and block the flash (hood, or whatever) from influencing the exposure?

-- hide signature --

Regards,
Alan
my Flickr

Reply   Reply with quote   Complain
RussellInCincinnati
Senior MemberPosts: 3,197
Like?
underexposure's post-processing burden barely worth mentioning
In reply to boardsy, Oct 30, 2013

original poster boardsy wrote: ...if the sensor is effectively ISOless then is base ISO always optimal ...or are you as well off raising it in camera (providing highlights are preserved) rather than in RAW/pp?

Russell: Well have finally brought that concept into the field. In the scene below, the flight was going to be at ISO 3200 and 1/10th of a second. Around F/8 to get complete cross-frame clarity. Everything looked fine in the electronic viewfinder (i.e. clip-on eyepiece-magnified LCD).

But then remembered at the last minute "hey, why sweat the exact exposure when I'm worried about blowing out highlights in the white fabrics? Why worry about how many bracketing exposures needed, at what intervals, just in case? Why not move the ISO back to 800, the screen looks kind of dark but everything's still perfectly viewable/manual focusable, and then can have the luxury of deciding in post-processing exactly how bright to raise the scene highlights?

This is an interesting result Russell - did you need to handle noise carefully in LR or just use defaults?

There was no careful anything. The only thing I did to handle the "underexposed" image taken with cam set to ISO 800 above, was spend somewhere between 3 and 10 seconds moving the normally-little-used "exposure" slider in Lightroom 4.x to the right, to about +1.88 EV or whatever.

After doing that, there wasn't the slightest difference or extra trouble involving unusual noise suppression or anything else. Well the difference was it was kind of fun to feel absolutely zero stress about having blown highlights in an unbracketed scene with lots of high-key detail, the God-like comfort of just moving the exposure slider as little or as much as wanted until the highlights were just shy of burning out.

I'd still wonder about ""the same" or better results by post-processing an image from a camera set to ISO 800" - if the same, why bother (unless of course to preserve crucial highlights, as above!), if better to "do ISO in pp" then it's worth the trouble to focus and frame with a dim LCD...

Firstly, it's only in the electronic viewfinder era that we have the possibility of seeing an image at the "right" brightness. You and I have been looking at optical viewfinders for years, where the brightness of the previewed image has been arbitrarily darker or brighter than/irrelevant to the final print. So it's not exactly like a fresh outrage that the electronic image preview is a couple of stops darker than the final print.

On a more complex note, we can break all scenes in the world down into 2 categories. Scenes that have nothing bright in them at all, and scenes that do. The scenes that have nothing bright in them at all, you can just set the cam to ISO 3200 with impunity, heck you could set the ISO too high if you felt like it, since we just said the scene has nothing bright in at all to blow out. The scenes that do have something bright in them, well at least the bright parts of the scene (which visually are usually the most important) are easily discernible in an EVF, even if you're underexposing by a couple of f-stops.

It's not like everything disappears just because you're 2 stops dark. Let's also recall that in a near-ISO-less cam like a Nex, you only need to leave the viewfinder a stop or two dark, just dark enough that there is no possibility of blowing highlights. If I have an ISO 3200 scene, it's quite fun to set the ISO to 800, because that's just a huge safety margin, the 2-stop-under relieves me from bracketing, with no troublesome penalty in post-processing. And since I'm only 2 stops under, the view isn't so bad. After all don't typical scenes have 8-12 stops or more of dynamic range? 2 stops is really a hill of beans most of the time.

hmmm... this may be a NEX-specific issue, but maybe use a raised flash to trigger the always-bright LCD, and block the flash (hood, or whatever) from influencing the exposure?

Am intending for this discussion to apply to users of any raw-capable camera.

In the specific Nex case, it's only miserly folks like me who are happy to work with $225 dollar used Nex x3x bodies, who don't have access to a menu setting for always-bright finder. This whole issue of dark finders dries up and blows away for Nex 5+ cams that have a permanent setting for always-bright finder. Hmm but then doesn't that mess up the on-screen histogram?

Reply   Reply with quote   Complain
RussellInCincinnati
Senior MemberPosts: 3,197
Like?
quit trying to avoid taking credit for useful thread
In reply to boardsy, Oct 30, 2013

boardsy: I'd still wonder about ""the same" or better results by post-processing an image from a camera set to ISO 800" - if the same, why bother?

Because (a) no more bracketing (b) never a blown highlight, highlights always visually top priority if you don't want your photos to look like unfinished artworks (c) simplification of always using ISO 800 when the light's bad, no matter exactly how bad it is. One less field decision that could result in a screwup if you set a real high ISO and then forgot to reset it later. Which mistake have done approximately one zillion times. (d) Am wagering if you're using a camera known to your raw processing software, the raw processing software can push the data (beyond the minimum read noise ISO such as Nex C3 ISO 800) with more computational resources and customizability than camera firmware. Am thinking here about RawTherapee working with all 64-bit numbers when extracting the last bit of highlight detail, all the work put into Lightroom etc. (e) this is getting real unimportant but it is psychologically less stressful knowing that you're using your sensor at the 4%-or-whatever-better minimum read noise setting, rather than the most-perfect-EVF-preview setting.

Ordinary interior light, quite small F-stop for perfect clarity throughout scene depth. ISO 800 didn't cut it for 1/6th second exposure, needed ISO 3200 or so. But left cam at ISO 800 and moved an exposure slider in raw post-processing software. Feels like the only low-stress, no-bracketing file clutter, infinite-highlight-detail way to fly now.

Reply   Reply with quote   Complain
John Sheehy
Forum ProPosts: 16,283
Like?
Re: am happy with your starting this helpful thread
In reply to RussellInCincinnati, Oct 31, 2013

RussellInCincinnati wrote:

Thus the Nex C3 is a good deal of the way towards being Iliah Borg's "true ISO-less camera."

Well, except for Canon DSLRs and some non-Sony sensors used in Nikons like the D4, most cameras with RAW are pretty "ISO-less" by ISO 400 or less.  It would be really nice, however, if the cameras actually used it under the hood, like the Sigmas and many medium format cameras do, or let advanced users decide what gain to use at what ISO settings, and what bit depth the RAW files use.  Shooting an exposure index of 6400 from the camera's ISO 800 gain not only gives you more headroom, but if that headroom is not put to use, lossless-compressed RAWs are smaller, because unlike the least significant bits which are filled with noise, always, the high-order bits contain all or mostly zeros, which can be severely compressed.

Upshot of all this is to make overexposing a low-light photo, or bothering to bracket it, is quaint history.

Well, I have Canons which aren't so-called "ISO-less" until 800 to 3200, but I have always balked at people's suggestions that "proper exposure is very important at high ISOs"; the fact is, the absolute exposure, not the relative exposure at a high ISO, is what determines noise first and foremost.  It seems that many people have trouble distinguishing between relatives and absolutes, and seeing less noise at ISO 3200 with +2/3 EC, and more noise at ISO 3200 and -2/3 EC, they falsely assume that the relative exposure was the thing that made the former better, when, in fact, the only thing that made it better was the 1.33 stops extra absolute exposure, and -1/3 stop at ISO 1600 would give the same noise benefit, and 1 stop more highlight headroom.

Reply   Reply with quote   Complain
RussellInCincinnati
Senior MemberPosts: 3,197
Like?
interesting that underexposed files can be smaller
In reply to John Sheehy, Oct 31, 2013

Thus the Nex C3 is a good deal of the way towards being Iliah Borg's "true ISO-less camera."
John Sheehy wrote:

Well, except for Canon DSLRs and some non-Sony sensors used in Nikons like the D4, most cameras with RAW are pretty "ISO-less" by ISO 400 or less. ...Shooting an exposure index of 6400 from the camera's ISO 800 gain not only gives you more headroom, but if that headroom is not put to use, lossless-compressed RAWs are smaller...high-order bits contain all or mostly zeros, which can be severely compressed.

Fascinating. Instead of letting the camera firmware throw meaningless/random lower-order bits into every data point, which is basically writing meaningful higher-order bits pointlessly padded on the right with random lower-order bits, you are just storing string-of-zero-prefixed "significant figures" in underexposed raw files. So that in some cases the raw files are smaller.

Upshot of all this is to make overexposing a low-light photo, or bothering to bracket it, is quaint history.

Well, I have Canons which aren't so-called "ISO-less" until 800 to 3200, but I have always balked at people's suggestions that "proper exposure is very important at high ISOs"; the fact is, the absolute exposure, not the relative exposure at a high ISO, is what determines noise first and foremost.

Sure. If the camera isn't usefully amplifying the analog signal (in some way that's better than digital manipulation) past some certain ISO, it's the greater or lesser quantity of photons in the image that are going to determine how much arithmetic multiplication you're going to be doing to the noise as you multiply the data. The fewer the photons hitting the sensor, the more you're going to be amplifying the inherent sensor noise on your way to making the real data show as full brightness.

It seems that many people have trouble distinguishing between relatives and absolutes, and seeing less noise at ISO 3200 with +2/3 EC, and more noise at ISO 3200 and -2/3 EC, they falsely assume that the relative exposure was the thing that made the former better, when, in fact, the only thing that made it better was the 1.33 stops extra absolute exposure, and -1/3 stop at ISO 1600 would give the same noise benefit, and 1 stop more highlight headroom.

It is weird to set a fixed, low ISO and watch the screen get arbitrarily dim as you reduce the shutter speed and/or close down the lens, in the field. Your instinct in the moment is the uncomfortable feeling that your image is falling off a cliff, because of the viewfinder-dimming-feedback you are getting (since your viewfinder doesn't know you're going to be raw post-processing). It takes some to get over that and realize that simply the less or more you expose the image, you're really only deciding to have more or less noise...with the final brightness always ending up fine no matter how deep you are diving.

Reply   Reply with quote   Complain
Steen Bay
Veteran MemberPosts: 6,397
Like?
Re: am happy with your starting this helpful thread
In reply to RussellInCincinnati, Oct 31, 2013

RussellInCincinnati wrote:

original poster boardsy wrote: ...if the sensor is effectively ISOless then is base ISO always optimal ...or are you as well off raising it in camera (providing highlights are preserved) rather than in RAW/pp?

Boardsy my takeaway a few weeks ago from our tortuous thread, was to notice that a Nex C3 sensor has minimum read noise at around ISO 800. Which you made so easy to see by including a handy noise graph in your original post.

Now the C3's read noise is not a lot less at ISOs besides 800, but hey why not use the sensor at that most efficient-in-some-ways ISO, even if you theoretically "need" a higher ISO to make the little preview look "good"? Since you can so easily "raise the ISO" (i.e. move an "exposure" slider in say Lightroom) after the fact. The only downside is that the LCD does indeed get dim.

Note that the graphs in the OP are a bit inaccurate at the highest ISO settings. If a camera has reached its lowest read noise (e-) at for example ISO 800, then the read noise will be the same at higher ISOs. The read noise will always decrease or remain the same if the ISO is increased, it'll never be higher at a higher ISO than it was at a lower ISO.

Reply   Reply with quote   Complain
boardsy
Senior MemberPosts: 2,204Gear list
Like?
Re: quit trying to avoid taking credit for useful thread
In reply to RussellInCincinnati, Oct 31, 2013

RussellInCincinnati wrote:

boardsy: I'd still wonder about ""the same" or better results by post-processing an image from a camera set to ISO 800" - if the same, why bother?

Because (a) no more bracketing (b) never a blown highlight, highlights always visually top priority if you don't want your photos to look like unfinished artworks (c) simplification of always using ISO 800 when the light's bad, no matter exactly how bad it is. One less field decision that could result in a screwup if you set a real high ISO and then forgot to reset it later. Which mistake have done approximately one zillion times. (d) Am wagering if you're using a camera known to your raw processing software, the raw processing software can push the data (beyond the minimum read noise ISO such as Nex C3 ISO 800) with more computational resources and customizability than camera firmware. Am thinking here about RawTherapee working with all 64-bit numbers when extracting the last bit of highlight detail, all the work put into Lightroom etc. (e) this is getting real unimportant but it is psychologically less stressful knowing that you're using your sensor at the 4%-or-whatever-better minimum read noise setting, rather than the most-perfect-EVF-preview setting.

Ordinary interior light, quite small F-stop for perfect clarity throughout scene depth. ISO 800 didn't cut it for 1/6th second exposure, needed ISO 3200 or so. But left cam at ISO 800 and moved an exposure slider in raw post-processing software. Feels like the only low-stress, no-bracketing file clutter, infinite-highlight-detail way to fly now.

Ok, I'll take the credit for this thread! Though let it be known that I'm just riding the coat-tails of bob2n, Great Bustard et al's schooling in the recent "ISO is not part of exposure" DPR-wars, and the concomitant ISO-less approach. A big education for me, as there is plenty of misinformation out there on the mythical "exposure triangle", raising ISO "increases the light available to the sensor"(< I actually read that in a tutorial! ) etc.

And thanks again for persevering further than I did since the previous "full circle" apparent dead-end. I put this in practice yesterday evening - grabbing a few street shots in the dark, I set ISO at 800, set shutter speed at 1/50 (min. necessary for my FDn 50/1.4 + Lens Turbo hand-held) and f2 so under-exposed, and fired away without worrying about the LCD or ISO/brightness. Haven't got around to pp yet, but I'm confident about highlight preservation and brightness-raising in ACR.

Like this uninspired snap, underexposed at hand-held imits of 1/30 & f2 on FDn 35/2 + LT, base ISO, and pushed in ACR to least as bright as the scene was to the eye (could take more perhaps if desired), with highlight detail preserved:

-- hide signature --

Regards,
Alan
my Flickr

Reply   Reply with quote   Complain
RussellInCincinnati
Senior MemberPosts: 3,197
Like?
yes you've got the religion
In reply to boardsy, Oct 31, 2013

oardsy wrote: grabbing a few street shots in the dark, I set ISO at 800, set shutter speed at 1/50 (min. necessary...and f2 so under-exposed, and fired away without worrying about the LCD or ISO/brightness.

That's it, that's it. And the street scene you showed again was a good deal of what originally got my attention about this topic. For the life of me could not see anything wrong with that "pushed" photo.

...at least nothing wrong with this photo that a bit of color balancing couldn't fix.

(also interesting to read your description of the optics used for this photo)

Reply   Reply with quote   Complain
RussellInCincinnati
Senior MemberPosts: 3,197
Like?
so just use an ISO a stop or two below what's "needed"
In reply to Steen Bay, Oct 31, 2013

Steen Bay wrote: Note that the graphs in the OP are a bit inaccurate at the highest ISO settings. If a camera has reached its lowest read noise (e-) at for example ISO 800, then the read noise will be the same at higher ISOs. The read noise will always decrease or remain the same if the ISO is increased, it'll never be higher at a higher ISO than it was at a lower ISO.

Now that you mention it, that must be true. Because why would the manufacturer do anything to the camera to increase its read noise at increasingly high ISO's? Wouldn't make sense. Guess the very slight bounce upward in read noise in the original post graphs is simply "noise in the data" or measurement drift. Thanks for pointing this out, it seems obvious now.

So with your observation, if you want to preserve highlights and get rid of the need for bracketing above ISO 800 on a Nex C3, all you have to do is make sure you're using an ISO that's 800 or higher, but still lower than what you need for a perfect electronic viewfinder preview and playback. By that logic, if need to take a Nex C3 raw picture at ISO 25600, leaving the ISO at 6400 would eliminate the need for bracketing, preserve highlights, and be "just as good" as leaving the ISO at 800. And by setting an ISO just a stop or two lower than what you "need", ISO 6400, for a good viewfinder image, leaves you with an usable viewfinder image without being pointlessly dim ISO 800.

Point well taken. To recap, in the field when you need a photo above ISO 800, set the camera to an ISO at least 1 and perhaps 2 stops lower than what you think is indicated. Viewfinder will still be usable, and you can fix the ISO in raw post-processing.

Reply   Reply with quote   Complain
Keyboard shortcuts:
FForum MMy threads