Is 100 a native ISO on 50D? Correct Dynamic Range values for 50D

Started by GregoryOfManhattan, July 15, 2013, 03:41:40 AM

Previous topic - Next topic

0 Members and 1 Guest are viewing this topic.

GregoryOfManhattan

magic lantern src/raw.c uses the DxOmark screen data for the dynamic_range of each camera at each ISO.
the code computation assumes the values begin at ISO 100, however available data on DxOmark does not include a value for ISO 100. http://www.dxomark.com/index.php/Cameras/Camera-Sensor-Database/Canon/EOS-50D

Measuered ISO,  manufacturer ISO,   dynamic range Ev
    ,100,   
158,    200,    10.94
306,    400,    10.6
618,    800,    10.05
1172,   1600,   9.19
2342,   3200,   8.26
4433,   6400,   7.26
9210,   12800,  6.33


in src/raw.c we had
static int dynamic_ranges[] = {1094, 1060, 1005, 919, 826, 726, 633};

what value should be used for ISO 100?
a 3rd degree polynomial fit to existing data gives a value of 11 for the ISO 100 value.
static int dynamic_ranges[] = {1100, 1094, 1060, 1005, 919, 826, 726, 633};

for the 100 ISO value DxO mark also reports almost the same "ISO Sensitivity" for manufacturer ISOs 100 and 200 - 157 and 158.

http://www.clarkvision.com/articles/digital.sensor.performance.summary/ Individual Sensor Data Table for 50D show lower dynamic range for ISO 100 10.95 vs. ISO 200 11.14

perhaps contrarily, Bill Claff reports higher photographic dynamic range - 8.77 at ISO 100  and 8.46 at ISO 200 http://home.comcast.net/~NikonD70/Charts/PDR.htm

shooting a set of images as on http://www.rawdigger.com/howtouse/pixel-capacity-and-amplifier-gain
does not show the behavior of ISO 50 on 5D2 with ISO100 on 50D.
data below. shooting a tungsten lit wall at f5.6

Shutter ISO G avg G std deviation
S 1/5, 100, 5386.8,  127.02
S 1/10, 200, 5337.39, 136.3
S 1/20, 400, 5361.66, 152.85
S 1/40, 800, 5346.62, 182.41
S 1/80, 1600,  5320.95, 229.62
S 1/160, 3200, 5229.87, 301.09
Test 2
S 1/5, 100, 5086.95, 119.03
S 1/5, 200, 10198.97, 244.73

this quick test shows lower standard deviation at ISO 100 with the same signal (Green average value).
the example shown on http://www.rawdigger.com/howtouse/pixel-capacity-and-amplifier-gain for the 5D2 at ISO 50 shows a much greater noise (standard deviation) at ISO 50 which is why it considered a non-native ISO.
on the 50D, ISO 100 appears (subject to a statistical validation of the quick test here) to be a native ISO for the camera with lower noise than ISO 200 and is a valid selection for shooting.

still leaves the question of what is the correct value for ISO100 dynamic range in the magic lantern code.
scaling the clarkvison number to match DxO at ISO200 would give 10.75 at ISO 100
scaling the Bill Claff number would give 11.34 at ISO 100

my quick and not-statistically validated test give a very similar noise level at ISO 100 and 200 (7 % lower standard deviation - 13% lower squared standard deviation). so the value should be expected to be close to the value at 200.

the current code is an improvement as otherwise, the dynamic range values would be lower for all ISOs and this would restrict the apparently available exposure range.

Audionut

Quote from: GregoryOfManhattan on July 15, 2013, 03:41:40 AM
http://www.clarkvision.com/articles/digital.sensor.performance.summary/ Individual Sensor Data Table for 50D show lower dynamic range for ISO 100 10.95 vs. ISO 200 11.14

He shows increased Gain e/DN with increased Maximum Signal (full well capacity).  By these results, ISO 100 is not simply ISO 200 with some digital manipulation as I assumed based on the DXO results.  His measured DR also fits in with the results of the full well capacity.  There wasn't a doubling of capacity with ISO 100 vs ISO 200.

Quote from: GregoryOfManhattan on July 15, 2013, 03:41:40 AM
perhaps contrarily, Bill Claff reports higher photographic dynamic range - 8.77 at ISO 100  and 8.46 at ISO 200 http://home.comcast.net/~NikonD70/Charts/PDR.htm

Looking at various Canon bodies, the digitally pulled ISO's (at the lower end) have better dynamic range then their base stops.  This suggests an error in the workflow.  Consider (5D3) that he shows ISO 320 as having (all but) the same dynamic range as ISO 100.

1%

A1ex said to look at CMOS register value.

reg 0 = 0x4 for iso 100 0x94 for iso 200...

It actually changes the value at 100, 160, 320, etc

Audionut

Quote from: 1% on July 15, 2013, 04:34:59 AM
It actually changes the value at 100, 160, 320, etc

That's different behavior to the 5D3.

Quote from: a1exAs expected, intermediate ISOs like 160 or 250 do not cause any changes in ADTG/CMOS configuration. These ISOs are obtained by applying some digital gain to the raw data acquired at the nearest full-stop ISO, and this gain is configured from the DIGIC register 0xC0F08030 (SHAD GAIN). In LiveView, the gain is only applied to the YUV image (it does not affect the 14-bit raw data at all), but in photo mode, the gain is burned into the raw data. Don't ask me why