First, a little background.

A couple of weeks ago I discovered the Clarkvision.com site for the second time and, this time, I studied his article on measuring Sensor Noise, Dynamic Range, and Full Well Capacities for the Canon 1D in detail.

Clark uses an incandescant light globe shining on a sheet of copy paper to generate an even light source for measuring the performance of cameras and their sensors.
http://www.clarkvision.com/imagedetail/evaluation-1d2/index.html

I was keen to use the CHDK ISO override feature to get a slower ISO than stock on my A570 IS and it struck me that Clark's was a method that I could use to try and calibrate the ISO override feature for my camera. I highly recommend Clark's site, check it out.

The Test Setup.

I clipped 3 pieces of copy paper to a book so that I could stand them up vertical.
I warmed up an incandescent lamp for >30 minutes and then set it up about 2' away from the copy paper to start with.
I set up the camera on a tripod focused on the centre of the sheet of paper.
- Auto-everything off in the Canon menu
- Manual Mode
- Incandescant WB
- Colours Off
- Aperture f/8
- Manual focus zoomed in to the maximum (23.2mm)
- Exposure 1/60 sec. The AC power in Oz is 60Hz so I wanted to make sure I always captured the light from one full AC cycle in case the lamp brightness varies during the cycle.
- CHDK RAW enabled. I kept the JPEG files as well so that I had a record of the settings for each shot and to cut down on the typing I had to do to setup the charts in the spreadsheet later on.
- CHDK Overrride "Off"

I set the ISO to1600, which is an option in Manual Mode on the A570, and adjusted the distance of the lamp from the paper until the histogram on half-press showed no overexposure.

Before I took any pictures I also turned off all the fluorescent lights in the room.

After the setup, all I change for the test photos is either the ISO setting or the ISO override value.

The Test Photos

I took a first set of photos with the Canon ISO settings to compare against the shots taken later using ISO Overrride and also to check how reproducible the method is using an incandescent lamp + copy paper as my light source. I took a photo at all the Canon ISO settings (50, 80, 100, 200, 400, 800, 1600) and repeated this 4 times.

I took a second set of photos with the ISO Override enabled at 10x (2, 3, 4, 5, 6, 7, 8, 9) and 100x (1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15). I only repeated this 3 times because my SD card filled up.

Finally, I disabled the ISO Override and I took a 5th set of photos at all the Canon ISO settings (50, 80, 100, 200, 400, 800, 1600) again so I could compare these results at the end of the session with the ones at the start of the session.

The Processing/Analysis

Convert the raw file to a TIFF using "dcraw -D -4 -j -t 0 -T ". The object here is to get the raw values without any adjustments or conversions.
-D Shows the totally raw data with no interpolation and no colour scaling.
-4 Write 16-bit linear samples (fixed white level, no gamma).
-t 0 By default, dcraw applies the flip specified by the camera. -t 0 disables all flipping.
-j this option guarantees that each output pixel corresponds to one raw pixel.
-T Write TIFF

Get the mean value for a 200x200 pixel selection at the centre of each picture. Clark uses Photoshop to do this step, which I don't own. A google search for some free "photo mean std deviation" software quickly found ImageJ and, not only could I get the min, max, mean and std deviation for a selection using a menu option but I could write a script to open the image, get the image height & width, calculate the position of 200x200 pixels at the centre, select that range, call a function to get the mean for the current selection, open the matching JPEG image and get the EXIF data (using someone else's plugin), write all this info to a log window in a tab delimited format (that I could later save to disk and open in OpenOffice Calc) and run the script for all the pictures in a directory. If you haven't seen ImageJ yet it is worth checking out.

The result is good news and bad.

The good news is that this test gave a pretty unambigous picture of what possible range of ISO values my camera is capable of. It also showed that the ISO override feature allows you to choose from a bigger set of ISO values. eg 100, 125, 160, 320, 500, 640, etc instead of just the stock values 80, 100, 200, 400 and 800.

The bad news is that the lowest ISO I achieved using the ISO Override feature on my A570 IS was equivalent to the stock ISO80 setting and the highest was only fractionally higher than the stock ISO800 setting. This highest effective ISO800 also applies to the stock ISO1600 that you can choose under Manual mode.

I've included charts of my results as attachments to this post so you'll need to be logged in to see what I'm talking about.

Stock ISO Chart.JPG plots the mean of the raw data values plotted against the stock ISO settings. It shows two things:
- That the incandescent lamp did provide a reliable, reproducible light source for testing, at least at the 1/60 sec exposure I used.
- That ISO1600 is exactly the same as the ISO800 on my A570 IS

ISO Override Chart 1.JPG and ISO Override Chart 2.JPG are the results for the ISO override settings.

ISO Override Chart 1.JPG plots the mean raw data values plotted against the ISO override value. It show:
- That the lowest ISO bottoms out at a mean raw data value of around 62. This is exactly the same value that you get with the stock ISO80
- That the highest raw data value is around 360. This is only slightly higher than the stock ISO800 which gives a mean raw data value around 330.
- That for an ISO override which gives a value of ISO80 in the JPEG EXIF the the raw data value of the image is a good match with the stock ISO80 setting.
- That for an ISO override which gives a value of ISO800 in the JPEG EXIF the the mean raw data value of the image is slightly higher than the stock ISO800 setting. The stock ISO800 gives a mean raw data value around 325 - 330 while an ISO override that gave an ISO800 in the JPEG EXIF had a mean raw data value of 330 - 350.

ISO Override Chart 2.JPG plots the mean of the raw data values plotted against the ISO that is recorded in the JPEG EXIF data. This is easier to compare with Stock ISO Chart.JPG because some ISO override values I used produce the same ISO in the EXIF data. You can see this in ISO Override Chart 2.JPG where some ISO values have 6 data points instead of 3.

The mapping between the ISO override values I used and the ISO the camera recorded in the JPEG EXIF data was:

ISO ISO
override in
value EXIF
10x 2 32
3 50
4 64
5 80
6 100
7 100
8 125
9 125
100x 1 160
2 320
3 500
4 640
5 800
6 1000
7 1000
8 1250
9 1250
10 1600
11 1600
12 2000
13 2000
14 2000
15 2500