The tweener settings are a convenience, and not a whole lot else. Sure, I can't think of a situation where I'd need to use them, but that said, it's always nice to know I have the option. If one can fine tune the exposure compensation in 1/3 increments, the shutter speed, and aperture in 1/3 stop increments, why not ISO? And I agree with ziggy, the difference is probably tiny.
That said, the difference had better be tiny, otherwise Canon really should put something in the manual mentioning it's more a "simulated" exposure, kind of like the expanded ISO 50 on some of the cameras can only be enabled in custom functions because it's not a "native" setting.
This is the kind of thing where I'm sure if shadow noise is a monster concern for an application, people should have a heads up, likewise for highlights, these being the only situations I can think of where in camera tweaking like this could really be measured (particularly in 14-bit raw)
That said, the difference had better be tiny, otherwise Canon really should put something in the manual mentioning it's more a "simulated" exposure, kind of like the expanded ISO 50 on some of the cameras can only be enabled in custom functions because it's not a "native" setting.
This is the second reference in this thread to "simulated" vs. "native" exposure, and I'm having trouble getting my mind around that distinction. Everything in digital is as "simulated" as everything else, no?
... Everything in digital is as "simulated" as everything else, no?
Not speaking for anyone else but the "assumption" floating on the Web is that fractional ISOs are calculated as adjustment factors from the calibrated ISOs. If true, that could explain the increase in noise and the decrease in dynamic range, which would also affect tonality.
Fractional f-stops and fractional shutter speeds are true hardware functions (presumably).
"Calibrated ISOs"
That's the heart of it -- calibrated off of what? Analog (ASA) presumably, but if they can calibrate accurately to the "traditional" ASAs why should the quality be any different at interpolated points?
I saw with my own eyes the results of the test that you posted, and they're hard to argue with. I'm drilling down on this simply because I have frequent occasion to shoot at high ISO and want to eliminate paranoia as a factor in my decisions on settings. Until I came upon this thread, my procedure had been to climb the ISO ladder as darkness conditions dictated, but now I have to fret about those tweener settings.
That's the heart of it -- calibrated off of what? Analog (ASA) presumably, but if they can calibrate accurately to the "traditional" ASAs why should the quality be any different at interpolated points?
...
"Calibrated ISO" is an industry term meaning that the ISOs for a given camera have been adjusted to meet published standard sensitivities. Apparently the ISO sensitivity ratings for imagers are determined by a combination of:
1) Chip design: CMOS, CCD and derivatives have design sensitivities.
2) On-chip low noise amplifiers: The first level of amplification is in the silicon of the imager itself. I believe that the output can be modulated by ISO rating. I believe this also where the analog/digital conversion occurs.
3) In camera processor: After the chip delivers its signal data that data is further processed by the in-camera processor. Apparently even RAW images are processed to some degree. It is assumed that the principal (calibrated) ISO sensitivities are established at this point.
3a) It may be that this is the point that the fractional ISO adjustments are factored in, after all of the other data has been massaged. If true, mathematical "shortcuts" may have to be applied in order to keep processing speeds practical. No one knows for sure because the manufacturers are not revealing the exact process, but this additional processing and mathematical shortcut could explain the odd behavior of the fractional ISO outputs. If only the calibrated ISOs are used then this step can be eliminated.
RAW conversion: Whether the linear RAW imge is processed in-camera or by a computer/software based RAW conversion, at some point a non-linear curve is applied to the color channels to produce what we know as an acceptable exposure (or an unacceptable exposure, but more appropriate data than the original RAW data). The output from this stage finally creates the data used for JPG, TIFF, etc.
If you accept the previous you can see multiple opportunities for the introduction of error and even noise.
A little old, but a bit of perspective from the other side of the fence.
Thanks for that link, I had missed it.
Chuck Westfall is a truly great guy and a straight shooter (pun intended). He and I agree in that I said,
"The fractional ISOs are not terrible. Most people will never see the reduction in quality. Few will measure the reduction in quality.
If I had a "need" to use fractional ISOs I would be concerned. Not a biggee IMO."
If anyone "needs" to use fractional ISOs, feel free to do so. The most affected zones are probably in the 3-4 stops below median, or the dark to deep shadow areas.
If you have to recover those regions it is because something happened and the exposure was blown anyway. If you don't have to recover those regions, you will never "see" the problem of fractional ISOs. Monitors won't display the problem and neither will printing.
It is also possible that the cameras with 14 bit capture and processing largely mitigate the problem entirely for any practical use. (The major benefit to 14 bit capture and processing is probably in the shadow zones.)
Got it. No problemo I had a night sports shoot last night and couldn't get you (or your avatar anyway) out of my brain. My new "stepping stones" only include 1250 because you blessed it!
Mostly unrelated, i accidentally shot at 1600 in brilliant sunshine for a while yesterday and got some great shots that, upon realizing the problem, I feared were a disaster. Turns out, no issue whatsoever. Clear as a bell.
Comments
That said, the difference had better be tiny, otherwise Canon really should put something in the manual mentioning it's more a "simulated" exposure, kind of like the expanded ISO 50 on some of the cameras can only be enabled in custom functions because it's not a "native" setting.
This is the kind of thing where I'm sure if shadow noise is a monster concern for an application, people should have a heads up, likewise for highlights, these being the only situations I can think of where in camera tweaking like this could really be measured (particularly in 14-bit raw)
__________________
www.browngreensports.com
http://browngreensports.smugmug.com
Not speaking for anyone else but the "assumption" floating on the Web is that fractional ISOs are calculated as adjustment factors from the calibrated ISOs. If true, that could explain the increase in noise and the decrease in dynamic range, which would also affect tonality.
Fractional f-stops and fractional shutter speeds are true hardware functions (presumably).
Moderator of the Cameras and Accessories forums
That's the heart of it -- calibrated off of what? Analog (ASA) presumably, but if they can calibrate accurately to the "traditional" ASAs why should the quality be any different at interpolated points?
I saw with my own eyes the results of the test that you posted, and they're hard to argue with. I'm drilling down on this simply because I have frequent occasion to shoot at high ISO and want to eliminate paranoia as a factor in my decisions on settings. Until I came upon this thread, my procedure had been to climb the ISO ladder as darkness conditions dictated, but now I have to fret about those tweener settings.
Anyway, Noiseware ROCKS!
__________________
www.browngreensports.com
http://browngreensports.smugmug.com
"Calibrated ISO" is an industry term meaning that the ISOs for a given camera have been adjusted to meet published standard sensitivities. Apparently the ISO sensitivity ratings for imagers are determined by a combination of:
1) Chip design: CMOS, CCD and derivatives have design sensitivities.
2) On-chip low noise amplifiers: The first level of amplification is in the silicon of the imager itself. I believe that the output can be modulated by ISO rating. I believe this also where the analog/digital conversion occurs.
3) In camera processor: After the chip delivers its signal data that data is further processed by the in-camera processor. Apparently even RAW images are processed to some degree. It is assumed that the principal (calibrated) ISO sensitivities are established at this point.
3a) It may be that this is the point that the fractional ISO adjustments are factored in, after all of the other data has been massaged. If true, mathematical "shortcuts" may have to be applied in order to keep processing speeds practical. No one knows for sure because the manufacturers are not revealing the exact process, but this additional processing and mathematical shortcut could explain the odd behavior of the fractional ISO outputs. If only the calibrated ISOs are used then this step can be eliminated.
RAW conversion: Whether the linear RAW imge is processed in-camera or by a computer/software based RAW conversion, at some point a non-linear curve is applied to the color channels to produce what we know as an acceptable exposure (or an unacceptable exposure, but more appropriate data than the original RAW data). The output from this stage finally creates the data used for JPG, TIFF, etc.
If you accept the previous you can see multiple opportunities for the introduction of error and even noise.
Simple?
Moderator of the Cameras and Accessories forums
For sake of discussion, I found a rebuttal article, including some not hugely informative information from Canon: http://digitaljournalist.org/issue0706/tech-tips.html
A little old, but a bit of perspective from the other side of the fence.
Thanks for that link, I had missed it.
Chuck Westfall is a truly great guy and a straight shooter (pun intended). He and I agree in that I said,
"The fractional ISOs are not terrible. Most people will never see the reduction in quality. Few will measure the reduction in quality.
If I had a "need" to use fractional ISOs I would be concerned. Not a biggee IMO."
If anyone "needs" to use fractional ISOs, feel free to do so. The most affected zones are probably in the 3-4 stops below median, or the dark to deep shadow areas.
If you have to recover those regions it is because something happened and the exposure was blown anyway. If you don't have to recover those regions, you will never "see" the problem of fractional ISOs. Monitors won't display the problem and neither will printing.
It is also possible that the cameras with 14 bit capture and processing largely mitigate the problem entirely for any practical use. (The major benefit to 14 bit capture and processing is probably in the shadow zones.)
Moderator of the Cameras and Accessories forums
Mostly unrelated, i accidentally shot at 1600 in brilliant sunshine for a while yesterday and got some great shots that, upon realizing the problem, I feared were a disaster. Turns out, no issue whatsoever. Clear as a bell.
__________________
www.browngreensports.com
http://browngreensports.smugmug.com