I think the poster was asking a more generalized question than you
responded to, but hey, who am I to stop you from making a jerk of
yourself.

Usually the lowest ISO setting of your digital camera is going to be
where you get the best picture with the lowest noise.  Higher settings
are amplified, I believe, so noise is increased.

They are only in increments of 100, 200, 400 to remain reciprocal with
shutter speed and aperture settings.

The signal is amplified to reach the desired level.

    As I just got through typing (and thus, you have not yet had
time to read it), and perhaps I can add a bit more detail this time, now
that I know the focus of your question.

1)    The sensor cells (CCD or CMOS) have capacitors which store a
    charge during the exposure.  The maximum change which can be
    stored defines the low-end ISO for the sensor, as more exposure
    fills the cell to the maximum, and highlight detail is lost.

2)    Once the exposure is done, the charge is moved from the sensor
    cells towards circuitry on the border of the sensor.

3)    The first bit of circuitry will be a gain stage, so if the
    exposure was not sufficient to fill *any* of the cells, the
    range of voltage resulting from the charges which *were* stored
    is increased by applying amplification, to provide a reasonable
    range of voltage to the next stage.

4)    The second bit of circuitry will be the A/D converter (Analog to
    Digital converter.)  This converts the analog information from
    the sensor cell which came through the amplifier into a number
    which represents that level in a digital form

    Note that there is always *some* noise in the sensors, though it
may be small enough so it is not noticed in the presence of the signal
at low gains.

    However, if the gain is high, it is amplifying both the noise
(which is approximately a constant level for a given exposure time) and
the signal.  This brings the noise up to a level where it is obvious
(typically on darker areas of the photograph).

    There are at least three sources of noise:

1)    "Thermal" noise, which is a function of the resistance
    (impedance) that the sensor and circuitry up to the amplifier
    sections and the temperature.  The higher the temperature, the
    more thermal noise, and the higher the gain (ISO) the more
    visible the noise from a given temperature will be.  Far
    Infrared sensors are often operated at very cold temperatures,
    such as 70K (the temperature of liquid nitrogen).  You've
    probably seen far IR images in the video from missile strikes
    from aircraft from the Gulf War footage.

2)    The digital "clocking" noise (which is left-overs from the signals
    used to move the information from the sensors to the A/D
    converters.

3)    "Fixed Pattern" noise, resulting from minute variations in the
    sizes of the sensors, and slight variations in the accuracy of
    the A/D converters.

    This last one has the advantage of being somewhat predictable,
    and many cameras have a mechanism to subtract this out from the
    image data.  This is commonly done (for long exposures) by
    closing the shutter, and collecting noise for the same time as
    the exposure just collected, and then digitally subtracting the
    noise from the signal data.

    The noise from these various sources all tends to be worse for
long exposures or high ISO, and as a result, it behaves in a manner
somewhat similar to film pushed to high ISOs.  However, the film won't
have the Fixed Pattern or the clocking noise sources, just the
equivalent of the thermal noise.  (There is a difference.  As you push
the ISO in a digital sensor, the brightness variation becomes greater,
but the size remains that of a single pixel.  As you push the ISO in
film, however, individual grains become much larger than the pixel size
from the digital image at the same ISO, so the noise (grain) from
seriously pushed film is more obvious than that from the sensors at
higher ISOs.  Yes, you can see the noise when you push a Nikon D70 to
1600 ISO, but it is not nearly as obnoxious as the grain in film pushed
to the same speed.

    Also, in digital images, there are various algorithms to reduce
the noise.  Some are built into the cameras (including that for
correcting fixed-pattern noise), while other may be in the camera, or
may be in the  computer during post-processing.

    The D70 is not particularly aggressive in in-camera noise
reduction processing, and (from what I have read) the Cannon is much
more aggressive.  One side effect is that the Cannon (at least the
Digital Rebel cameras) appears to produce softer images, as sharpness is
one of the casualties of aggressive noise reduction.  I am rather glad
that the Nikon leaves it to post-processing (in the computer), so I can
make my own choices between sharpness and noise.

    I'm not sure how much of this noise reduction may be done in a
Cannon camera in RAW mode.

    I do know that this last 4th of July, I opted to shoot with the
in-camera long exposure noise reduction turned off, so I was not stuck
waiting for the same time as the last exposure to gather the noise
reduction information before I could shoot again.  For most images, it
came out rather nice, but for some (when I held the shutter open for
longer than I expected, thanks to delays in the launching of the next
fireworks), I wound up with serious fixed-pattern noise in those images.

    To see some selected examples, visit the following URL:

        http://www2.d-and-d.com/EXAMPLES/july-4th/index.html

    I hope that this helps,
        DoN.