What's the downside of High ISOs ?

[hdomke]

What does one give up with High ISO other than added noise? Does sharpness, tonal gradation or color fidelity suffer?

Having extremely high ISOs is tremendously helpful, but I wonder what I loose when I switch from ISO 200 to 6,400 on my Canon 1D Mk4.
As I do some "pixel-peeping" the only thing that is apparent is luminance noise (which looks like film grain) and color noise (which looks like christmas tree lights in the shadows).

Does high ISO cause any image distortion besides noise?

[ejmartin]

What does one give up with High ISO other than added noise? Does sharpness, tonal gradation or color fidelity suffer?

Having extremely high ISOs is tremendously helpful, but I wonder what I loose when I switch from ISO 200 to 6,400 on my Canon 1D Mk4.
As I do some "pixel-peeping" the only thing that is apparent is luminance noise (which looks like film grain) and color noise (which looks like christmas tree lights in the shadows).

Does high ISO cause any image distortion besides noise?

Sharpness, tonal gradation and color fidelity all suffer.  Any good demosaic algorithm uses tonal gradients or other measures of homogeneity to decide how to interpolate the missing colors in the raw data; noise generates false gradients that confuse the demosaic and lead to poorer demosaic.  Noise reduction leads to loss of contrast and detail.  Noise also leads to color fluctuations on all scales, though by smaller amounts on larger scales; color noise is a loss of color fidelity, more subtle on larger scales but still present.

[BernardLanguillier]

Sharpness, tonal gradation and color fidelity all suffer.  Any good demosaic algorithm uses tonal gradients or other measures of homogeneity to decide how to interpolate the missing colors in the raw data; noise generates false gradients that confuse the demosaic and lead to poorer demosaic.  Noise reduction leads to loss of contrast and detail.  Noise also leads to color fluctuations on all scales, though by smaller amounts on larger scales; color noise is a loss of color fidelity, more subtle on larger scales but still present.

Also DR, right?

Regards,
Bernard

[Ray]

Certain cameras seem to specialise in high ISO performance. The Nikon D3s springs to mind.

I wonder if there's any camera currently available on the market, of whatever format, that can produce a cleaner and sharper image than the D3s at ISO 12,800.

Would an 8x10" field camera with the best film currently available, push processed to ISO 12,800, produce a better image in any respect than the rather low resolution D3s with a mere 12mp?

I'd really like to know.

[ejmartin]

Also DR, right?

Regards,
Bernard

Yes.

[hdomke]

ejmartin,
Thanks for your suggestion that "Sharpness, tonal gradation and color fidelity all suffer"
Have you actually seen this in a print?
Or read technical papers which discuss it?

I've been "pixel peeping" at the images I am shooting at ISO 6,400 and I must say, after a little noise reduction, I can't detect any image flaws.
But perhaps I don't know what to look for.  

Sharpness, tonal gradation and color fidelity all suffer.  Any good demosaic algorithm uses tonal gradients or other measures of homogeneity to decide how to interpolate the missing colors in the raw data; noise generates false gradients that confuse the demosaic and lead to poorer demosaic.  Noise reduction leads to loss of contrast and detail.  Noise also leads to color fluctuations on all scales, though by smaller amounts on larger scales; color noise is a loss of color fidelity, more subtle on larger scales but still present.

[Guillermo Luijk]

I've been "pixel peeping" at the images I am shooting at ISO 6,400 and I must say, after a little noise reduction, I can't detect any image flaws.
But perhaps I don't know what to look for.

If you can see differences when 100% pixel peeping on your monitor, then there exists a print size from which those differences will become visible in the printed copy as well. Of course there will be too printing sizes for which the differences will become negligible; after all all images (film or digital) have certain degree of noise, and wonderful images have been printed in the history of photography.

[ImplantImages]

If you can see differences when 100% pixel peeping on your monitor, then there exists a print size from which those differences will become visible in the printed copy as well. Of course there will be too printing sizes for which the differences will become negligible; after all all images (film or digital) have certain degree of noise, and wonderful images have been printed in the history of photography.

I would add contrast to the previous sharpness and DR. Still modern cameras are unbelievable my D700 runs rings around my old D200.

[BJL]

If you can see differences when 100% pixel peeping on your monitor, then there exists a print size from which those differences will become visible in the printed copy as well.

Or rather, it will be visible at certain combinations of high print size and low viewing distance. The print size needed will probably be about the same PPI as the monitor's, so as a rough estimate 125PPI, which with a now entry level 40MP DMF sensor means about 58"x43" (about 1.5x1.1 meters), viewed from a typical computer screen viewing distance of 15 to 20 inches (40 to 50 cm).

What fraction of DMF images are printed at a lowly 125PPI, or at any computer display PPI level, and then viewed this closely?

One point to bear in mind is that increasing pixel counts are in good part used to allow printing at higher PPI, not to simply increase print area in direct proportion to the increase in pixel count.


Imperfections that are only visible at 100% on-screen or comparable viewing conditions are probably n the realm of scientifically measurable but insignificant by the practical criteria of most photographers' display and viewing intentions.

[Doug Peterson]

What does one give up with High ISO other than added noise? Does sharpness, tonal gradation or color fidelity suffer?

Having extremely high ISOs is tremendously helpful, but I wonder what I loose when I switch from ISO 200 to 6,400 on my Canon 1D Mk4.
As I do some "pixel-peeping" the only thing that is apparent is luminance noise (which looks like film grain) and color noise (which looks like christmas tree lights in the shadows).

Does high ISO cause any image distortion besides noise?

The "flexibility" of the file - namely the ability to recover (pleasing, natural, detailed, color-accurate, smoothly graded) shadows and highlights and to make large adjustments to the white balance suffer very soon after base ISO. This may or may not be important to you.

Note that there is a huge difference in the WAY (not just the speed) that CMOS and CCD images degrade at Higher ISOs, so experience with digital back files does not correlate well to experience with Canon files the other (except in so far as both get generally worse as ISO goes up).

Also note that you should check your files out with different raw processors. I don't have any significant experience with the 1D Mark 4 (it's not targeted at the markets we sell into) but with the 5D Mark 2 which also has great high ISO characteristics the high-ISO files look much nicer in Capture One 5 than in Lightroom 2/3-beta or Adobe Camera Raw. Part of this assessment is aesthetic - for instance each raw processor shows a different kind of noise and deciding which kind of noise looks best is only opinion. Part of this is objective - how much subject detail is still visible and how far into the shadows and highlights is there meaningful data? To me the biggest difference is Adobe's tendency to produce the christmas tree lights in the shadow in low-frequency (multiple pixel wide) blotches especially in areas of continuous shadow. Both Adobe and Capture One produce this christmas-tree noise but in my experience (5D Mark 2 between ISO 1600-6400) it is much less offensive in Capture One. In fact I'd go so far as to say I can often pick out when a image was processed in LightRoom because of that shadow christmas tree low-frequency color noise.

Of course the above paragraph should be read with the disclaimer that my company sells Capture One. Fortunately you don't have to take my word for it - the 30-day trial is free and fully functional (make sure to select "try Capture One PRO" when you first launch the application). Also to be a more fair (though still biased) informant you should be aware that other raw processors also have excellent reputations for high ISO files including Brian Griffith's Irrident Raw Developer.

Because the 1D IV is still quite new I think the community here would benefit from you posting either your own experiments processing high ISO files with different raw processors or posting a raw file for others to play with in different raw processors.

Doug Peterson
__________________
Head of Technical Services, Capture Integration
Phase One, Leaf, Cambo, Canon, Apple, Profoto, Eizo & More
National: 877.217.9870  |  Cell: 740.707.2183
Newsletter: Read Latest or Sign Up
RSS Feed: Subscribe
Buy Capture One at 10% off
Personal Work

[Doug Peterson]

Would an 8x10" field camera with the best film currently available, push processed to ISO 12,800, produce a better image in any respect than the rather low resolution D3s with a mere 12mp?

Thank goodness there are different tools for different jobs.

If I had to shoot the ultra-low-light action photography that ISO 12,800 implies with an 8x10 field camera I might kill myself.

Then again if I ever start shooting landscape or product photography with a small format camera set to ISO 12,800 I would be equally out in la-la land.

[hdomke]

dougpetersonci
"The "flexibility" of the file - namely the ability to recover ... shadows and highlights and to make large adjustments to the white balance suffer very soon after base ISO."
I was aware that at high ISO noise can be exaggerated when brightening shadows; hence "expose right". I was not aware of difficulty with white balance.

Do you know of any in depth reviews or research papers that document the image distortion seen at high ISO in the latest DSLRs?

I still see many photographers with these cameras keeping there ISO settings in the range of 200 or perhaps 400; it seems that they don't trust higher settings. Are they right or are they missing out on a great new feature?

[Ray]

Thank goodness there are different tools for different jobs.

If I had to shoot the ultra-low-light action photography that ISO 12,800 implies with an 8x10 field camera I might kill myself.

Then again if I ever start shooting landscape or product photography with a small format camera set to ISO 12,800 I would be equally out in la-la land.

I understand, but sometimes one might want to get the shot with whatever camera is at hand.

Let me rephrase the question. Is there any camera of any format, digital or film, which will produce a superior image in any respect, at ISO 12,800 (and/or ISO 25,600) to the Nikon D3s?

We know that MFDBs have much higher resolution than a D3s, but also much poorer high-ISO performance.

We also know that resolution is degraded at high ISO, and significantly graded with MFDBs at high ISO.

DXO Mark doesn't compare resolution, but I see that at ISO 6400, and at the downsampled size of 8x12", the D3s has better SNR, significantly better DR, better tonal range and better color sensitivity than the D3X with double the pixel count.

I see from the Imaging-Resource Comparator that despite the advantages of the D3s, the D3X still retains very marginally better resolution in some parts of the image at ISO 6400. However, at ISO 6400 underexposed one or two stops (ie. ISO 12,800 or 25,600), it's doubtful the D3X would have any resolution advantage over the D3s, just as it's doubtful that a P65+ would have any resolution advantage over the D3X at ISO 6400.

[Dick Roadnight]

Would an 8x10" field camera with the best film currently available, push processed to ISO 12,800, produce a better image in any respect than the rather low resolution D3s with a mere 12mp?

This is a very academic question, as you would use high ISO to give you high shutter speeds for hand-hold-ability and/or small aperture for DOF, and you do not get much hand-hold-ability or DOF with a 10 * 8... perhaps you are thinking about surf shots?

I think the nice thing about 800 ISO on a Hasselblad is that you have so much res for a start, that, hopefully, even after loosing res to ISO, you still get a better picture than you would with a D3X?  ...then, of course the D3 performs better than the D3X in low light?

[Dick Roadnight]

What does one give up with High ISO other than added noise? Does sharpness, tonal gradation or color fidelity suffer?

I think, perhaps, we are over concerned about noise... and we would be better off using higher ISO to get faster shutter speeds and more DOF when we need them.

[attachment=21434:Nadhim_only.jpg]

...presumably we do not need permission for politicians, and this is Nadhim Zahawi, our parliamentary candidate.

This I shot hand-held at

200 ISO
1/180
f5.6
50 -110 @ 110mm

The building behind is relevant to the picture, as it shows our candidate in our town.

His face is sharp, but his ears are not, so I think I got the compromise right, but, if I had moved in for head and shoulders, I think I should have used 800 ISO (max), F11 to have anywhere adequate DOF?

[Ray]

This is a very academic question, as you would use high ISO to give you high shutter speeds for hand-hold-ability and/or small aperture for DOF, and you do not get much hand-hold-ability or DOF with a 10 * 8... perhaps you are thinking about surf shots?

I think the nice thing about 800 ISO on a Hasselblad is that you have so much res for a start, that, hopefully, even after loosing res to ISO, you still get a better picture than you would with a D3X?  ...then, of course the D3 performs better than the D3X in low light?

There are two major reasons to use a fast shutter speed, (1) to freeze camera shake in the absence of Image Stabilization technology when the camera is hand-held, (2) to freeze subject movement. Neither IS nor use of a tripod can freeze subject movement.

There could be lots of situations where one has one's 8x10 field camera set up on a tripod to shoot landscapes, then suddenly a situation arises that requires a fast shutter speed. It might be the unexpected appearance of a surfer riding a high wave whilst the camera is set up for a beach scene, or it might be the unexpected appearance of a flock of geese leisurely flying in front of the setting sun.

But let's not get sidetracked by the impracticality of using an 8x10" field camera in such circumstances. I mentioned that format merely because it probably still retains a resolution advantage over the best MFDBs. I used the term as a form of shorthand in a question which is really asking, is there any camera in existence that produces higher resolution then the D3s at ISO 12,800 and/or 25,600, in real-world scenes (as opposed to line charts).

800 ISO is a far cry from 12,800 ISO. Even some of the latest P&S cameras deliver respectable resolution at ISO 800.