This test shows that, using the override feature, on my A570 the fastest possible shutter speed is about 7/100K sec and the shutter speed can be altered down to about 10/100K sec while still operating pretty linearly (i.e the exposure at 10/100K sec is 1/2 the exposure at 20/100K sec and 1/4 the exposure at 40/100K sec).
The procedure I used is substantially the same as I used for testing the ISO override feature in
http://chdk.setepontos.com/index.php/topic,3373.0.html.
However, the results I got from my first attempt to do the shutter speed test using an incandescant lamp didn't seem very linear so I made two refinements.
For my light source I used sunlight on a bright cloudless day.
For this test I calculated the mean value of a 200x200 selection of just one of the green pixels from the raw photo (the even numbered pixels on each odd numbered row). This is what Clark does in the article I'm copying my approach from
http://www.clarkvision.com/imagedetail/evaluation-1d2/index.html but I skipped it for the ISO override test and instead calculated the mean value of all four RGGB raw pixels in a 200x200 pixel selection.
The Test SetupI taped 3 pieces of copy paper to the lid of my laptop and angled the lid so that it was a right-angles to the sunlight.
I set up the camera on a tripod focused on the centre of the sheet of paper and as close as I could get without casting a shadow on the paper.
- Auto-everything off in the Canon menu
- Manual Mode
- Sunny WB
- Colours Off
- Manual focus zoomed in to the maximum (23.2mm)
- Aperture f/5.5 (I set this as open as possible at max zoom because I would be using the fastest shutter speeds)
- Exposure 1/2000 sec.
- CHDK RAW enabled. I kept the small JPEG files as well so that I had a record of the EXIF settings for each shot and to cut down on the typing I had to do to setup the charts in the spreadsheet later on.
The Test PhotosI took a set of photos with the shutter speed override enabled at 1/100K x (40, 38, 36, 34, 32, 30, 28, 26, 24, 22, 20, 18, 16, 14, 12, 10, 8, 6, 4, 2, 1) and repeated it to get 3 sets of photos.
The Processing/AnalysisConvert the raw file to a TIFF using "dcraw -D -4 -j -t 0 -T ".
Extract just one of the Bayer Array green pixels from the TIFF file (effectively creating a monochrome version of the image using just that green pixel) and then get the mean value for a 200x200 pixel selection at the centre of that monochrome image.
Clark uses commercial software that I don't own. ImagesPlus to split the raw photo into the red, green, and blue monochrome versons and Photoshop to get the mean values of the selection.
I love ImageJ. I was able to write a single ImageJ macro that processed all the .CRW files in a directory. It opens the photo splits out the green pixels, selects 200x200 pixels at the centre of that green image, calls a standard function to get the mean for the current selection, opens the matching JPEG image and gets the EXIF data I want (using someone else's ImageJ plugin) and write all this info for each photo to a log window in a tab delimited format that I can save to disk and open in OpenOffice Calc. If anyone is interested I can post the macro.
The ResultsI've included the charts of my results as attachments to this post so you'll need to be logged in to see what I'm talking about.
Shutter Speed Override Chart 1.JPG shows the results for all three sets of photos. It shows good linearity down to 20/100K or 10/100K and a definite bottom limit to the fastest possible shutter speed.
Shutter Speed Override Chart 2.JPG has all the results for just the shutter speeds between 40/100K and 20/100K in one series so that I can put a trendline through all of them. The trendline formula is 743420.27x + 35.72. When I do the same for shutter speeds between 40/100K and 10/100K the trendline formula is 708925.48x + 46.52.
I took a pair of dark frames at 1/2000 sec and the mean raw data values were 32.2 & 32.3. So the value of 35.72 I get in the first trendline is pretty consistent with a value of 32 being no exposure. I'm pretty sure I've also seen references to this value of 32 for a zero exposure in the raw file somewhere else on the CHDK wiki. So I'm taking the 40/100K to 20/100K trendline as more probably more accurate.
The mean raw pixel values for 1/100K, 2/100K and 4/100K sec all vary between 85 and 95. If I use 90 as the average raw pixel value for the fastest possible shutter speed and plug it into the trendline formula from Shutter Speed Override Chart 2.JPG then the fastest possible exposure works out at 7.3/100K sec.
