 # How does the D3 achieve such high ISO?

Started Feb 17, 2008 | Discussions thread
Re: Why? And, 1Ds3?

SDRebel wrote:

I'm not certain I understand the implications of all the numbers.
Nevertheless, at the very least I have several questions.

First, why is the D3 producing better marks in your analysis?

Second, how the does 1Ds3 stack up (I assume you were reporting on
the 1D3 numbers)?

According to figures from

the electron count at raw saturation for the 1Ds3 is about 31000 electrons. The read noise is about 5 electrons at ISO 1600. Adding to Marianne's list:

Sensor DR at ISO 200 illumination limit:
D3 full-scale e- count = 65,600 (Roger's number)
D3 read noise = 6 e- (my number, more conservative than Roger's)
Mk III full-scale e- count = 38,000 (Emil's number)
Mk III read noise = 4 e-
1Ds3 full-scale e- count = 31,000
1Ds3 read noise = 5 e-

D3 sensor DR = 65,600/6 = 11,000 = 13.4 stops
Mk III sensor DR = 38,000/4 = 9,500 = 13.2 stops
1Ds3 sensor DR = 31,000/5 = 6,200 = 12.6 stops

But here is where one has to ask what is being measured here. These figures refer to pixel capacity, pixel read noise, and pixel DR. The 1Ds3 has 75% more pixels; the higher pixel density means that its DR and the D3 dynamic range are being measured at different spatial scales. We should refer the 1Ds3 values to the coarser spatial scale of the D3 pixels to provide a fair comparison. The capacity scales as the area, so we should multiply the 1Ds3 e- count by 1.75: 31000 x 1.75 = 54,200; the read noise scales roughly linearly with pixel size when combining pixels, so the 1Ds3 scaled read noise is 5 x 8.46/6.4 = 6.6; the dynamic range of the 1Ds3 sensor, when scaled to the size of the D3's pixels, is 54,200/6.6 = 8212 = 13.0 stops for "sensor DR" as defined by Clark.

Shot noise at the same spatial scale is thus 65,600/54,200 = 1.21 = .3 stop better for the D3 over the 1Ds3. The 1D3 fares worse here due to the 1.3 crop factor (smaller sensor). Using Marianne's scaling logic to get the ISO 800 read noise (which is not wholly accurate; read noise does not necessarily scale this way)

For example, at a moderate sensitivity of ISO 800:
D3 read noise = 6/16,400 * 16,383 DN = 6 DN
Mk III read noise = 4/9,500 * 16,383 DN = 7 DN

1Ds3 read noise at ISO 800 = 5/7750 * 16383 DN = 10.6 DN

Finally, do I understand correctly that for an ISO setting (200 or
above) that your analysis means that the D3 produces a cleaner file,
i.e., a file with less noise? But, what about resolution and dynamic
range?

The data indicates that DR favors the D3, though when referred to a level playing field by compensating for the different pixel densities, the difference with the 1Ds3 is only about 1/3 stop. The 1D3 fares somewhat worse due to the 1.3 crop factor. The D3 has better resolution as a percentage of frame size than the 1D3 due to its higher pixel count, but 1/3 worse than the 1Ds3 in linear resolution due to the latter's 75% higher pixel count.

-- hide signature --

emil
-- http://theory.uchicago.edu/~ejm/pix/20d/

Complain
Post ()
Keyboard shortcuts: