Purpose
The purpose of this page is to describe an experiment I performed to measure the exposure consistency of a consumer grade digital still camera, in order to determine whether this hardware is a viable solution for professional-quality stop-motion animation. In question is the camera's ability to capture identical images with solid color consistency. In other words, when converting the captured images into a movie, will that movie exhibit flicker?
Setup
For this experiment, I used the Canon PowerShot S50 digital camera connected to my computer via a USB 1.1 connection. The capture process was driven by Breeze Systems' PSRemote v1.1.
As you can see in the picture above, the USB cable connects the camera directly to the computer. All of the camera's settings can be controlled through this connection, therefore there is no need to physically touch the camera during animation, which is a Good Thing. The camera is powered by an AC adapter rather than running on battery power. This is done to ensure that the camera won't shut down in the middle of the experiment.
All images were captured in Canon's RAW file format at full, five-megapixel resolution (2592x1944). The Canon S50 does support JPEG compression as well, but I chose not to use JPEG because it is a lossy file format, and therefore could introduce imprecision into the captured frames. The RAW files were converted to 16 bit per channel .tiff images, which were then downres'd in Photoshop to 640x480. (Yes, I know this defeats the purpose of using a five-megapixel camera, but I wanted to generate a movie file that could be viewed on today's computers.) The 640x480 images were then imported into Gromada's VideoMach v2.7.2 to create a 24-bit, 24 fps MPEG-1 movie.
Test 1: Small Aperture
For the first test, I set the aperture to its smallest radius. For the Canon S50, the minimum aperture size is F8.0. (Remember, the higher the aperture value, the smaller the lens aperture.) A small aperture is desirable for stop-motion work, because it results in maximum depth of field. A shallow depth of field destroys the illusion that your miniature set is life-sized.
The lighting, unfortunately, was not as controlled for the first two tests as I would have liked. I used a single, standard incandescent light bulb to illuminate my subject. This light source was not connected to any kind of power conditioner, so there may have been fluctuations in the light levels. However, the first two tests were conducted in the middle of the night when there was no major drain on the power supply in my studio.
To maintain a consistent exposure, I locked the shutter speed to 1.3 seconds, and turned on the AF Lock.
The above screenshot shows the main window of PSRemote. All of the camera settings are easily accessible, and there are customizable shortcut keys to take a preview shot (F9), or release the shutter (F8). There is also a time lapse feature, which allows you to take a picture at fixed intervals - not especially useful for stop-motion work, but I found it handy in conducting this experiment!
Unfortunately, there is indeed visible flicker in the resulting movie, as you can see by clicking on the thumbnail above. Some might not be bothered by this amount of flicker, but others undoubtedly will be. This flicker exists in the source RAW files, as can be seen in the histogram movie. Therefore, the flicker was introduced either by the camera or the lighting, and not by the .tiff conversion, downres'ing, or MPEG compression.
Test 2: Wide Aperture
For the second test, I set the aperture to its widest opening. For the Canon S50, the maximum aperture size is F2.8. I again set the camera to manual exposure, this time with a shutter speed of 1/5 second, and turned on the AF Lock for consistent focus.
Results:
Results:
With the larger aperture, the flicker is almost completely eliminated in the final movie. The histogram movie still shows noise, but to a lesser degree. Notice that, although the histogram graph values do vary on the Y axis, the entire curve does not shift left and right as it did in the first experiment.
At this point, I am seeing a correlation between aperture and flicker. However, I do not know for certain yet that the lighting is not to blame.
Test 3: Small Aperture, with UPS and Volt Meter
In an attempt to rule out voltage fluctuations, I plugged the lights and the camera AC adapter into an APC UPS (Model #BK500MC). I also connected an analog volt meter to the UPS to see if I could detect any voltage changes. To further reduce the chances of stray light interfering with the experiment, I sealed the camera up in a room with no windows, and ran a thirty-two-foot USB cable (two cables daisy-chained together) under the door and into my office. The set was illuminated by two standard incandescent bulbs, as opposed to the single incandescent bulb used for the first two tests.
The camera was again set to its smallest aperture, F8.0, with full manual control of the shutter speed, etc. (I noticed too late that I'd forgotten to turn on AF Lock, but since nothing in the scene is moving, that should not be a factor.)
Results
The UPS does appear to help the situation some, as the test movie exhibits less flicker than test 1. However, the UPS did not eliminate all the voltage irregularities. To better illustrate this point, I cropped a 320x240 area of my 5-megapixel source images to create a closeup movie of the volt meter needle. However, the voltage does not exactly correlate to the exposure variance. In other words, it appears that the camera is not providing a consistent exposure.
Test 4: Wide Aperture, with UPS and Volt Meter
This test was identical to Test 3, except that the camera was set to its widest possible aperture, F2.8. I also remembered to turn on AF Lock, but again, that setting should not affect this experiment.
Results:
Again, with the iris open wide, the resulting movie was much more stable. There is one pronounced dip right at the beginning of this test, but if you look at the closeup movie, you can clearly see that the voltage dips right at that moment.
Conclusion
With and without a UPS, the tests done with a small aperture had more flicker than those done with a wide aperture. As Jim Arthrus pointed out on the stopmotionanimation.com message board, this is probably due to the fact that this camera model uses a single mechanism for both the iris and the shutter.
You can judge for yourself how extreme you think this flicker is. I have a feeling that it will be an acceptable degree of error for some animators, but that professionals will want to look for a camera with more consistent performance. But it's important to note also that the voltage fluctuations were at least as responsible for exposure inconsistencies as the camera.