"Native" ISO speeds for sensors?

^ Ahh, okay. That may very well be - and vary from body to body.

gooble wrote in post #5761500
I agree but am not sure the native ISO is 100. Based on the results of the tests and Canon's comments I think it's between 100 and 200. Most of those graphs show the lowest noise at 160 which is between those two.

That doesn't make it a "native ISO". Shoot a RAW at ISO 100 and reduce the "exposure" in the converter to the equivalent of ISO 80 and you will have less noise. That's what exposing to right is all about. Does that make 80 the "native" ISO?

I'm still sticking with the definition of native being what the chip itself delivers thus the native ISOs are 100, 200, 400, 600, 1600 for the 40D. I like to think of ISO like a gas pedal of a car. I press the pedal and the car goes faster, or I turn the knob and the output is a bit higher but it's still native to the chip itself.

What you are debating is what is the best ISO for noise a complete different discussion than what is the native ISO of the chip.

tzalman wrote in post #5763850
That doesn't make it a "native ISO". Shoot a RAW at ISO 100 and reduce the "exposure" in the converter to the equivalent of ISO 80 and you will have less noise. That's what exposing to right is all about. Does that make 80 the "native" ISO?

I'm not even sure that there'd be less noise and even if it was that would not make the native ISO 80. It'd be what it is.

The definition of native ISO is not the ISO at which you get the least noise, although that is most likely true in many cases because the noise is caused by boosting the signal of the native sensor state.

I'll also say it again, I don't believe you can have multiple "native" ISOs as people keep posting.

Each sensor has a certain sensitivity to light which is compared to the ISO standard and then that ISO which it matches is the native ISO. The other ISO's are determined by multiplying or dividing that native ISO's image data by 2.

Although the term “native” gets bandied around a lot, I really do not think that it is applicable when talking about digital camera sensors, whether CCD or even CMOS. In my opinion, it is closer to being a marketing type of description. I encountered the term used much more frequently in the software engineering department where I worked before retiring and it seemed to usually be in the context of relating to the basic instruction set of a particular microprocessor ... in other words, if you wrote an applet for a processor using the machine level instruction set which is even more fundamental than writing in assembly language, that would be referred to as writing in the native language of the processor. And, of course, any assembler or compiler must be able to convert human readable code to the native language of the processor that will be running the code.

References to the native resolution of a computer monitor were made and that is something that makes sense in describing digital monitor parameters, but no rationale was established to make a case for using the term “native” with respect to digital camera sensor ISO values other than the implication that everything has a “native” state of being.

A camera sensor, by itself, doesn't have any particular ISO equivalent value associated with it. The charge that the sensor accumulates during an exposure is infinitesimally small and without much more amplification than is on the CMOS chip itself, not capable of having any ISO equivalent. Noise (in its various forms) would be major part of the sensor limitations when deciding on the operating range of the sensor. It is the job of the design and development engineers to see how much amplification ought to be applied to get a decent output. Several causes of noise exist -- amplifier noise, thermal noise, variation in light sensitivity between individual sensor sites, and quantization noise. Digital sensors react differently to light than film does, but equivalent ISO standards for digital systems have been developed to define the characteristics of a particular ISO in addition to the basic exposure value that a light meter would calculate (for example, with an EV of 12, a “correct” exposure could be obtained from using a shutter speed of 1/250 second, aperture of f/4, and an ISO speed of 100 whether using film or digital).

Once the engineers determine the sensor’s usable range of equivalent ISO values that can be synthesized for a particular sensor/processor set, they make a design decision on what range of ISO values is "right" for that particular camera. Of course, "right" is a very imprecise term and I can imagine that it involves the subjective evaluation of many people, both technical and non-technical along with considering the target user category.

Anyway, this long roundabout response hopefully explains why I feel that “native” is an irrelevant term for use with digital camera sensors. To my mind, the term “native” would describe the physical characteristics of something and not be a performance descriptor.

Forgive me for being naive, but with the 40D, the link w/graph posted near the top of the page suggests that the following would be the best ISO settings for the lowest amount of noise at each exposure step?
160, 320, 640*, 800, 1000, 1250, 1600, 3200

The graph shows that ISO 160 is cleaner than both 100 & 125.
The graph shows that ISO 320 is cleaner than both 200 & 250.
The graph shows that ISO 640 is cleaner than 500 & almost equal to 400.

Is it safe to assume that below ISO 800, one should only use 160, 320, & 640 if they're after the best image quality w/the lowest noise levels?


*640 and 400 are so close, level-wise, on the graph, that I would think it would be better to go ahead and shoot at 640 to take advantage of faster shutter speed or smaller aperture for essentially the same amount of noise.

bill boehme wrote in post #5764708
...
A camera sensor, by itself, doesn't have any particular ISO equivalent value associated with it. The charge that the sensor accumulates during an exposure is infinitesimally small and without much more amplification than is on the CMOS chip itself, not capable of having any ISO equivalent.
...

The following link is very informative about ISO, noise, signal-to-noise ratios and much more. The charges that each pixel can accumulate during an exposure (the well capacity) is in the range of 10,000 to 70,000 electrons/well. If the well is given more photons than it can handle, it still reports the maximum value and that pixel is considered "blown" If you adjust the exposure conditions so that the highest value recorded is just below the maximum capacity (i.e. almost, but not quite, blown), you have a definite link between incoming light intensity and sensor response and hence a way of defining an intrinsic property of the sensor. The decision that then has to be made is "how far above the traditional 18% target gray value should that almost full well pixel be placed?" If the DR of the sensor is 10 stops, then it might be set at 4-5 stops above the 18% gray. This then provides the ISO value associated with a particular sensor (well capacity). Higher ISO shots will have a correspondingly lower capacity (double the ISO collect 1/2 as many electrons etc)
http://www.clarkvision​.com …el.size.matter/​index.html

Andy

AJSJones wrote in post #5770756
The following link is very informative about ISO, noise, signal-to-noise ratios and much more. The charges that each pixel can accumulate during an exposure (the well capacity) is in the range of 10,000 to 70,000 electrons/well. If the well is given more photons than it can handle, it still reports the maximum value and that pixel is considered "blown" If you adjust the exposure conditions so that the highest value recorded is just below the maximum capacity (i.e. almost, but not quite, blown), you have a definite link between incoming light intensity and sensor response and hence a way of defining an intrinsic property of the sensor. The decision that then has to be made is "how far above the traditional 18% target gray value should that almost full well pixel be placed?" If the DR of the sensor is 10 stops, then it might be set at 4-5 stops above the 18% gray. This then provides the ISO value associated with a particular sensor (well capacity). Higher ISO shots will have a correspondingly lower capacity (double the ISO collect 1/2 as many electrons etc)
http://www.clarkvision​.com …el.size.matter/​index.html

Andy

Thanks, Andy. I am familiar with Roger Clark's web site and have it bookmarked and have read most, if not all of his very useful information. Even though the sensor has quantifiable characteristics, I was addressing the question from an electronics perspective of getting the collected charges to something that has a usable voltage and impedance that will interface with the other electronics. The extremely small charge that can be collected in each sensor well or bucket is much too small to do anything with it until it is amplified to a usable level that can the be processed by the A/D converter.

bill boehme wrote in post #5771831
The extremely small charge that can be collected in each sensor well or bucket is much too small to do anything with it until it is amplified to a usable level that can the be processed by the A/D converter.

Bill, do we know what level that would be? I'm guessing normal CMOS logic levels, but that's just guessing.

An on-sensor amplifier followed by an op amp, thence to the A/D converter? The ISO gain adjustment in the op amp?

-js

number six wrote in post #5773357
Bill, do we know what level that would be? I'm guessing normal CMOS logic levels, but that's just guessing.

An on-sensor amplifier followed by an op amp, thence to the A/D converter? The ISO gain adjustment in the op amp?

-js

CMOS logic levels in custom devices such as this probably operate with low noise margins so I would not be surprised if the logic levels are in the 1 to 3 volt range (certainly not at TTL levels). The on-sensor amplification is not very much (due to real estate reasons) and probably serve mostly as isolation and buffering. I understand that there is an amplifier module mounted immediately behind the sensor chip. Your assumptions all sounds exactly what I would expect.

bill boehme wrote in post #5773487
CMOS logic levels in custom devices such as this probably operate with low noise margins so I would not be surprised if the logic levels are in the 1 to 3 volt range (certainly not at TTL levels). The on-sensor amplification is not very much (due to real estate reasons) and probably serve mostly as isolation and buffering. I understand that there is an amplifier module mounted immediately behind the sensor chip. Your assumptions all sounds exactly what I would expect.

Apologies in advance, but my interactions with electronic devices have always been anthropomorphic. It may seem odd, but when I was a lad television sets had 30 or 40 firebottles inside and transistors were still semi-experimental. I figured that I knew what my friend the electron was going to do in that triode or cathode ray tube.

---------

Never mind that: seems to me that the electrons in the sensor wells would be thin and lonely, very fragile, and the on-chip buffer would be enough to make them stand up like men. (Ummm. I may come back and edit that simile.)

But they would still be only a few and, though properly organized, not a force to be reckoned with.

Pass them on to the operational amplifier and then they're big and strong and they have their marching orders: they're ISO 400 or whatever the sergeant told them to be.

But they're still analog - they have personalities, all the way from 1 microvolt to 3,000 microvolts.

Then they get to the analog/digital converter. This is much like the slaughterhouse for the poor critters. They go in as individuals and their parts come out in cans. Millions of personalities cut down and stuffed into 16,384 cans.

The horror! The horror!

Bye. Time for my meds.

-js

number six wrote in post #5773674
Never mind that: seems to me that the electrons in the sensor wells would be thin and lonely, very fragile, and the on-chip buffer would be enough to make them stand up like men. (Ummm. I may come back and edit that simile.)
-js

So are you saying that Canon's secret weapon is a thin coating of Viagra while poor primitive Nikons have to make do with a data-base of "feelthy peectures"?

tzalman wrote in post #5773995
So are you saying that Canon's secret weapon is a thin coating of Viagra while poor primitive Nikons have to make do with a data-base of "feelthy peectures"?

Isn't that in a Canon White Paper?

-js