What Creates Noise/Grain At Higher ISO Speeds?

What is the actual reason why photos shot at higher ISO speeds create
noise/grain?

In film, I can understand that to produce films which is more sensitive to
light, the chemistry may be different, but why is this evident with digital?

If you look here, noise is very evident in this digital shot, but I am
puzzled why it occurs:
http://www.pbase.com/jps_photo/image/36330285/original

 

In message <cn858s$5f1$>,

Do you realize what you are looking at there? (Most of) the image was
highly sharpened with no noise-reduction, and this is at ISO 1600. You
are looking at what would be a 40 inch by 27 inch full image, if your
monitor were big enough to display it; that's over 3 feet wide. Also,
the optics are at the limits of their resolution, so there is no
high-contrast image detail to compete with the noise. The noise got
sharpened more than the detail. Look at the left and bottom edges;
there is no sharpening there.
--

 

The point of this example was not to say this was a poor photo, just to give
an example of noise. It was chose randomly only as an example of noise.

If I knew this would upset people, I would not have included the link. I
included the link purely as an example.

I hope I have not offended anyone.

 

In message <cn868u$s56$>,

No one is upset or offended. I was just bringing to your attention that
this is a crop; it is less than 5% of the full image, are-wise, and
about 22% of the image in each dimension. There are woods surrounding
the carton, and a female Cardinal looking at the container, in the
original image.

You would not see this noise in an 8x10 inch print! You would barely
see it in a 16x20, if most of the original image was included.

I used a small, 100% crop because I was demostrating a lens combo at the
pixel level, and didn't need the rest of the image.
--

 

Amplification of the signal, they also amplify the noise thats present , as
well as the low level signal at higher ISOs.

 

I can guess why by way of analogy, and I don't know if it's
accurate or not, but until a better explanation comes along it'll
do, at least for me. If you consider AM/MW radios, reception is
good when stations are nearby and produce strong signals. When
signals are weak, you turn up the volume, but probably notice that
the fidelity isn't quite so good. When signals are *really* weak,
as for example when trying to listen to distant shortwave signals,
you might have a better radio that has an additional RF gain control
that can be used to make the radio more sensitive. But now the
station's signal might be really weak and be "buried in the noise".
Some of this noise consists of weak natural or man-made signals that
are also picked up by the radio. There's not a lot that can be done
about this. But not all of it is. Some of the low level noise is
created by the radio's own circuits (and might be due to quantum
phenomena). Very good radios use "low noise" components.
Transistors, resistors, etc. A look at manufacturer's data sheets
will show which "low noise" transistors should be used in RF
circuits if noise is to be minimised. I imagine that the same thing
happens in the circuits that the camera sensors are used in. You
can increase the amplification (boost the ASA speed), but at the
cost of making the background noise (which was always there) more
apparent, since it is relatively constant, while the desired image's
signal becomes weaker (if it wasn't weaker you wouldn't have boosted
the film speed).

 

Many of the explanations for noise in electronic sensors are based on the
technical limits of the early digital technology we are still using. As such
they are misleading.

One of the misconceptions is that smaller digital sensors can never yield
the lower noise now seen in larger sensors. The fact is that noise levels
will decrease in all digital sensors as more experience is gained in the
design and manufacture of these devices. So will the dynamic range of light
values the sensors are able to capture. Claims to this effect are already
being made for the newest Canon and Fuji dSLRs.

Manufacturer's claims, like those of politicians, occasionally coincide with
reality.

The most important limiting factor in reducing the size and performance
(regardless of sensor size) of high resolution digital sensors has nothing
to do with sensor technology but has to do with the inherent properties of
light, reflection (above and within the sensor), diffaction and the
inescapable problems of bringing all color light rays into focus at the same
plane. The sort of optical noise that is caused by the intereacting
properties of light, lenses and the flat plane of digital sensors
(regardless of sensor size) is already underappreciated in evaluating the
performances of current digital cameras. Because of these effects the
designs of lenses for high end digital cameras are going in directions that
film cameras did not require.

Software may be part of the solution. Programs already exist for
post-exposure tweaking of vignetting, barrel and pincusion distortion but
they require manual tweaking. Who knows what the future will bring? Olympus
has publicly stated it is moving toward building this sort of post-exposure
processing into its dSLR line.

 

Film suffers larger physical grain (dye clouds) in the plane of the film (call
it x,y) as the ISO goes up. Faster speed = larger 'grain' or dye clouds. This
results in the grain that you see. It's not really that cut and dry, but the
end result is coarseness in the negative image that we call "grainy".

In digital, the x,y is constant regardless of the sensitivity... the grain size
will always be the same for a given sensor. However, the noise level will be
different from pixel to pixel, and this difference is amplified by the
increasing sensitivity... that is what gives the noisy 'grainy' look on high ISO
digital shots. The 'grains' remain the same size, but there is a greater
dynamic difference between them due to noise. This is more complex than that,
as each pixel that you see from most SLR's is a composite of several pixels of
different color. The noise response in, eg, red, green and blue channels is
different for each color and this further makes pixel to pixel noise differences
more apparent in the resulting end pixel.

Hope that helps.

--
-- r.p.e.35mm user resource: http://www.aliasimages.com/rpe35mmur.htm
-- r.p.d.slr-systems: http://www.aliasimages.com/rpdslrsysur.htm
-- [SI gallery]: http://www.pbase.com/shootin
-- [SI rulz]: http://www.aliasimages.com/si/rulz.html
-- e-meil: there's no such thing as a FreeLunch.

 

See my other reply.

And a good thing you did.

Nah. Don't worry about assorted crankies.

Cheers,
Alan

--
-- r.p.e.35mm user resource: http://www.aliasimages.com/rpe35mmur.htm
-- r.p.d.slr-systems: http://www.aliasimages.com/rpdslrsysur.htm
-- [SI gallery]: http://www.pbase.com/shootin
-- [SI rulz]: http://www.aliasimages.com/si/rulz.html
-- e-meil: there's no such thing as a FreeLunch.

 

I believe the gent wanted to know "why is this so?" and not have a blast over
his particular posted example.

Cheers,
Alan.


--
-- r.p.e.35mm user resource: http://www.aliasimages.com/rpe35mmur.htm
-- r.p.d.slr-systems: http://www.aliasimages.com/rpdslrsysur.htm
-- [SI gallery]: http://www.pbase.com/shootin
-- [SI rulz]: http://www.aliasimages.com/si/rulz.html
-- e-meil: there's no such thing as a FreeLunch.

 

What happens in a digital camera is that light entering the sensor creates a
voltage at each site. If we assume that any sensor site peaks at say 1
volt, then that's the point of white saturation. In a proper exposure, it
would be nice to have no voltages at 1 but at least some close to 1 for the
very brightest details in the scene. Now, suppose the scene is darkish and
the highest voltage recorded is only 1/10 of a volt. Not a big problem in a
digital camera, since amplifiers with a multiplication factor of 10 (called
a voltage gain of 10) can beef up all the sensor signals before the
information is converted to digital form. However, the gain of 10 also
multiplies the noise by a factor of 10. So, when you shoot a scene at a
high ISO, you are using high gain and seeing noise that you ordinarily would
not see. Hope this helps. Tune an FM radio receiver to a frequency where
there is no station and the hiss that you hear is aural noise. Look at a TV
displaying a weak signal and the grain that you see is noise.

 

While the responses so far are close to the answer, what I've seen
is not quite complete. Noise is due to several things, including
read noise, thermal/shot noise, and photon noise. Read and
thermal noise will dominate in longer exposure images. In short
exposure images, photon noise can dominate (probably depends
on the camera). Newer cameras, and especially DSLRs, which
use low noise electronics, most will likely be photon noise
limited for daytime scenes (exposures up to at least a fair
fraction of a second.

Photon noise limit: what does it mean? When counting photons,
the noise is the square root of the number of photons
counted. This is a fundamental physical limit.
For electronic sensors, the number of photons
counted is dependent on the quantum efficiency (which is
very high), and the size of the pixel (proportional
to the number of photons that can be collected per unit
time), and the electron well capacity of the device
(also generally scales with the size of the pixels).

A high-end camera, like the Canon 1D Mark II, which has
pixel spacing of about 8 microns, has a full well
capacity of a little over 52,000 electrons. That means
the noise at maximum signal is sqrt(52000) = 228, and the
maximum signal to noise is 52000/sqrt(52000) = 228.
At lower signal, the noise is less, but so is the
signal. For example, at half capacity, 26000, the signal
to noise is 161.

For a camera with smaller pixels, and smaller full well,
the total noise is less, but so is the signal. Let's assume
a sensor with 4-micron pixels and a full well capacity
of 13,000 electrons. The maximum signal to noise is
only 114.

I've described the signal to noise of the 1D Mark II at:
http://clarkvision.com/imagedetail/digital.signal.to.noise
and show that it is working at the photon noise limit.
In particular, wee figure 2. This means
photon nose is the fundamental limit for that sensor.
The only way to improve on these values is to increase
the pixel area and full well capacity. Like I said above,
other digital cameras are likely working at the photon noise
limit too. The smaller sensors will never improve on this
fundamental limit, at least as we understand physics today.

Film is probably also photon noise limited, but since its
quantum efficiency is very low, the signal to noise is much
lower too. See figure 1 on the above page. For faster film,
the larger grains are collecting fewer photons, so one would
expect that the signal to noise would drop. My web page
shows noise levels ~4 times worse for film compared to
electronic sensors at similar ISO values, consistent with
the lower quantum efficiency.

So, if you want lower noise images, buy a camera with a
larger sensor to collect more photons for a given exposure.

Roger Clark
Photos, other digital info at: http://clarkvision.com

 

In message <GSOld.14324$>,

No one is being cranky. I replied as I did because he may have thought
that it was an entire image shrunk to web size.

Noise of that magnitude with strong sharpening in a pixel-for-pixel crop
is to be expected.
--

 

Hi

A higher ISO speed allows you to shoot better pictures in low light
conditions. A higher ISO speed allows you to use a faster shutter
speed and/or a smaller aperture in a given light level.

The drawback of higher ISOs are as follows:
- Increased noise
- Larger file sizes — you'll get far fewer shots per memory card at
ISO 800 than at ISO 100.
- Reduced shadow detail and sharpness

The same problem exists for analog cameras when you use fast film.

Regards
Gary Hendricks
www.basic-digital-photography.com

 

sigh... again, he wants the "why" not the "what".

Cheers,
Alan


--
-- r.p.e.35mm user resource: http://www.aliasimages.com/rpe35mmur.htm
-- r.p.d.slr-systems: http://www.aliasimages.com/rpdslrsysur.htm
-- [SI gallery]: http://www.pbase.com/shootin
-- [SI rulz]: http://www.aliasimages.com/si/rulz.html
-- e-meil: there's no such thing as a FreeLunch.

 

???????

 

Hi...

Me too - all I can possibly think of is that he's
suggesting that noisy pictures don't jpeg compress
quite as much as clean ones?

Take care.

Ken

 

In message <sRUld.25081$>,

Look, I have no responsibility to choose between giving a person exactly
what they asked for, or to remain silent. There are plenty of people
here capable of answering questions. I only say something when I think
something *needs* to be said.
--

 

Digital: nominal background radiation (random noise) from many sources,
including the Big Bang, is always stimulating the sensor. At low ISO
settings, the strength of the light coming throught the lens (signal) is so
much greater than this background noise that you don't see it. The higher
the ISO, the lower the signal to noise power ratio.

Film: film is made more sensitive by simply putting more silver halide
(thicker) on the film. When wet and developing, the sensitized silver tends
to be cohesive in the emulsion and clump together. The thicker the
emulsion, the bigger the clumps or "grains". Also thicker emulsions
required longer developing times and this gives the silver more time to move
around and form bigger grains.

 

Harvey wrote:

Yeah ... for JPG's esp. as compression cannot be efficient when every pixel is
different due to noise. In RAW, there would be less such difference (if any).


--
-- r.p.e.35mm user resource: http://www.aliasimages.com/rpe35mmur.htm
-- r.p.d.slr-systems: http://www.aliasimages.com/rpdslrsysur.htm
-- [SI gallery]: http://www.pbase.com/shootin
-- [SI rulz]: http://www.aliasimages.com/si/rulz.html
-- e-meil: there's no such thing as a FreeLunch.