This page was inspired by the discussion on the (now-defunct) Canon EOS mailing list in November 1996 regarding whether the the EOS shutter speeds are stepless. The only convincing argument so far had been a sentence in an A2 advertisement which claims that the camera is capable of a 1/137 sec exposure if needed.
I performed some measurements in November 1996, and here are the results. The short answer is: yes, they are stepless. But there is more.
Method: Removed film from camera, opened the back and pushed in the small lever which tells the camera that the back is open. Now the operation of the shutter can be observed by looking at it. The camera was mounted on a tripod and pointed at a brightly lit corner of the room. Shutter was released using the RC-1 remote control with 2-sec delay. Time was measured with a handheld stop-watch. All results below are averages from three measurements.
Since one cannot reliably measure short shutter speeds using a stopwatch, only longer shutter speeds were explored. I was surprised to get a 0.1-0.2 second reproducibility from my consecutive measurements, and I was typically within 0.2 seconds from the camera settings.
Result 1 (unexpected): The shutter speeds in M(anual) and Tv modes are very close to a perfect "square root of 2" sequence. I have not seen this to be claimed by anyone (including Canon) before!
(*) assuming sqrt(2) steps
I expect the same to continue towards shorter shutter speeds as well, so we should get, e.g. 1/32, 1/64, 1/128, etc. when we dial 30, 60, 125 manually.
Result 2 (expected): The actual shutter speeds in Avmode are stepless (but please read the note below!). By adjusting the aperture value and film speed, and pointing the camera in slightly different directions, a reading which was just on the border of flipping between 8" and 10" was obtained. The actual shutter speed was measured to be 8.71 sec, 9.45 sec, and 9.50 sec with three slightly different camera orientations. You can notice that such speeds can not be set manually.
Generalizations: I believe the above to be true for the Elan II, A2, and 1N cameras. Most probably the Rebel also shares similar shutter design. I also believe that the shutter speeds are obtained by counting regular clock pulses in the software. Therefore there actually might exist a "step" of, for example, 1/10000 seconds [just a guess]. For all practical purposes, however, the speeds can be considered stepless. The displayed readings are rounded to the closest standard value.
Note added on 07-July-1999. Canon has revealed that the light metering system of all EOS cameras uses 1/8-stop steps internally. A re-calculation of the experimental data above agrees with this. The accurately calculated 1/8 step row covering the 1/2 stop interval from 8 to 10 is as follows:
(the first being displayed as "8" and the last one being dispayed as "10" on the 1/2 stop increment bodies in M mode). What I apparently had measured, were the 8.724 and 9.513 values! At that time I did not bother to check how close they were to the 1/8 step series.
Thus the numerical results posted on this page remain standing, but the interpretation slightly changes: the light metering system uses 1/8 steps internally. I still guess that the physical shutter could be controlled in even shorter increments, but since the software never does so, it is in no way to our benefit.
Next tip: Are EOS apertures stepless?