What do you considering the upper limit for ISO?

On Wed, 21 Oct 2009 20:25:17 -0700 (PDT), RichA wrote:
: Lets say you aren't pretending, like so many do that you can take
: really good images at high ISO. Where do you draw the line? Canon
: and Nikon are now boasting 100,000+ ISOs for their latest offerings,
: the Canon isn't even FF! But lets say you shoot with a D700 and you
: have the luxury of going with either 3200 or 6400 ISO. Can you
: produce a shot worthy of keeping at 6400 ISO? Or are we just kidding
: ourselves that the output at that ISO, or from any camera is
: acceptable?
: I shoot a D300. IMO, beyond 800 ISO I have trouble accepting the
: images as really good. I'm talking about non-noise reduced raw
: images. Sometimes, it's possible to use NR and not make the picture
: look like crap, but the leeway is narrow.

With my 50D I've gotten some excellent results at ISO 1000, when the light is
a mixture of incandescent and dim daylight. I let Digital Photo Pro apply the
default Canon NR, but it's pretty conservative and has little effect on
sharpness.

Bob

Robert Coe wrote:

With my 50D I've gotten some excellent results at ISO 1000, when the light is
a mixture of incandescent and dim daylight. I let Digital Photo Pro apply the
default Canon NR, but it's pretty conservative and has little effect on
sharpness.

What does the ISO setting actually do on a digital camera? Is there some
way to make the sensor more sensitive, or is it purely a software function?

Roy Smith wrote:
Robert Coe wrote:

With my 50D I've gotten some excellent results at ISO 1000, when the light is
a mixture of incandescent and dim daylight. I let Digital Photo Pro apply the
default Canon NR, but it's pretty conservative and has little effect on
sharpness.

What does the ISO setting actually do on a digital camera? Is there some
way to make the sensor more sensitive, or is it purely a software function?

The sensor itself is an analog device which outputs an
analog voltage. For most sensors the "base" ISO
sensitivity is between 100 and 200.

Hence if the maximum input signal of the Analog-to-
Digital-Converter (ADC) is 1 volt and the maximum
output of the sensor is 1 volt, the ISO of the camera
would be the base ISO if the sensor output was connected
directly to the ADC. But instead there is an analog
"ISO amplifier" between them.

If that base ISO is 200, and the amplifier is set for
unity gain (acts like it isn't there), the camera will
provide a 1 volt signal to the ADC at maximum brightness
levels. So the camera then has ISO 200 sensitivity.

If the amplifier is set for 2x gain, the ADC will get
that same 1 volt signal at a brightness level that
produces only 0.5 volts from the sensor. That's ISO
400. If the amplifier is set for 4x gain, the 1 volt to
the ADC occurs for a 0.25 volt signal from the sensor,
and that is ISO 800.

There is a practical limit to increasing the amplifier
gain however. Noise from the sensor gets amplified too.
And indeed, noise from the amplifier increases also. At
some point there isn't enough difference between the
noise and the signal (Signal to Noise Ratio) and the
image becomes less than useful. At that point, psuedo
ISO increases are used in some cameras.

If the maximum ISO that is acceptable is 6400, a psuedo
ISO of 12,800 can be obtained using software by simply
doing a 2x digitial multiply of each output value from
the ADC. The noise generated by such an operation is
relatively high, which makes the results less than
ideal.

--
Floyd L. Davidson http://www.apaflo.com/floyd_davidson
Ukpeagvik (Barrow, Alaska)

Floyd L. Davidson wrote:

If the amplifier is set for 2x gain, the ADC will get
that same 1 volt signal at a brightness level that
produces only 0.5 volts from the sensor. That's ISO
400. If the amplifier is set for 4x gain, the 1 volt to
the ADC occurs for a 0.25 volt signal from the sensor,
and that is ISO 800.

There is a practical limit to increasing the amplifier
gain however. Noise from the sensor gets amplified too.
And indeed, noise from the amplifier increases also. At
some point there isn't enough difference between the
noise and the signal (Signal to Noise Ratio) and the
image becomes less than useful. At that point, psuedo
ISO increases are used in some cameras.


To a point that's true. But in another sense, no.

That point is reached when the amplifier gain is high
enough, and the S/N high enough, that it can reliably
detect the actual number of electrons, which then get counted.

This point has been reached for small numbers of pixels,
but not for camera-sized sensors, especially ones with
respectable dynamic range.

Doug McDonald

Doug McDonald wrote:
Floyd L. Davidson wrote:

If the amplifier is set for 2x gain, the ADC will get
that same 1 volt signal at a brightness level that
produces only 0.5 volts from the sensor. That's ISO
400. If the amplifier is set for 4x gain, the 1 volt to
the ADC occurs for a 0.25 volt signal from the sensor,
and that is ISO 800.
There is a practical limit to increasing the amplifier
gain however. Noise from the sensor gets amplified too.
And indeed, noise from the amplifier increases also. At
some point there isn't enough difference between the
noise and the signal (Signal to Noise Ratio) and the
image becomes less than useful. At that point, psuedo
ISO increases are used in some cameras.


To a point that's true. But in another sense, no.

That point is reached when the amplifier gain is high
enough, and the S/N high enough, that it can reliably
detect the actual number of electrons, which then get counted.

This point has been reached for small numbers of pixels,
but not for camera-sized sensors, especially ones with
respectable dynamic range.

I'm not sure what you are talking about, but the discussion
was about DSLR camera sensors. Saying something isn't
so and then citing an unrelated (and unspecified) example
is non-productive.

--
Floyd L. Davidson http://www.apaflo.com/floyd_davidson
Ukpeagvik (Barrow, Alaska)

Floyd L. Davidson wrote:

enough, and the S/N high enough, that it can reliably
detect the actual number of electrons, which then get counted.

This point has been reached for small numbers of pixels,
but not for camera-sized sensors, especially ones with
respectable dynamic range.

I'm not sure what you are talking about, but the discussion
was about DSLR camera sensors. Saying something isn't
so and then citing an unrelated (and unspecified) example
is non-productive.


Not quite .... the technology is very close to being usable in
actual camera sensors. I suspect that in the not too distant future
the electronics, and the sensors, will be good enough that,
at very low light levels, they actually WILL count photons and
thus render read noise zero. This will likely not extend, at
least for short exposures, up very far in intensity, where they
will revert to the present methods.

It is indeed meaningless for present cameras.

Doug