I was watching this slow motion video of a camera shutter and the scene at time 2:49 puzzled me. All of the mirrors flipped up in sync except for the one in the bottom left shot of the 1/2000s which was late, but still all four shutters opened at the exact same time. How come only one of them was delayed? I would think that for a machine as intricate as a camera, it would need to have extremely precise and consistent mechanisms, but it does not seem so, at least from watching this short segment of the video. Is there a good reason for this?
it would need to have extremely precise and consistent mechanisms
Pretty much, no. Does it really matter if the shutter is open for (say) 1/1900s rather than 1/2000s? That's a difference of about 7% of a stop1, which you're just not going to notice unless you're deliberately looking for it - it's far less than the 1/3 of a stop precision you generally have on a camera.
Stepping back a bit, the point here is more that the cameras we think about on this site are for art, not science. There are specialist devices which will give you very precise exposure times, but the cost of engineering that kind of precision into a device just isn't worth it.
The four slow motion clips are synchronised at the point the shutter curtain begin to move. In reality the "slow" mirror might have moved at the same time as the other three in relation to when the shutter button was pressed, but then for whatever reason began to open the shutter curtains after a shorter delay than the other three. Once the four clips are synchronized at the point where the shutter begins to open at the same time then it might only appear that the mirror of the one camera was slower as measured from the time the shutter button contact was closed. If you look closely you can also see the the mirror on the lower right is also fractionally "slower" than the two on top, which also have a near imperceptible difference between them.
There are a variety of reasons that could make the one mirror appear to move a little later than the other three. When a camera's shutter button is pressed a lot of things happen in a very short period of time. Some of those things occur simultaneously. Other things wait until a previous event is confirmed before the camera's processor gives the next "go" command.
Without a lens mounted on the camera any consideration of autofocus can be eliminated, but in situations where a lens is in place and AF is turned on it could affect the outcome. The way most PDAF systems work the initial reading is used to measure how far out of focus the lens is. The camera then sends an instruction to the lens to move the focus elements by that amount. In the meantime the mirror begins to swing up out of the way. Since the mirror is no longer all the way down, the PDAF sensor is now blind. Since it is not possible to take another reading with the PDAF sensor the camera will either 1) assume the lens followed the instruction correctly and take the photo as soon as the mirror position has been confirmed all the way up or 2) use a focus position sensor in the lens to confirm that the AF system moved the lens the correct amount and move the lens an additional amount if needed before allowing the shutter to begin to open once the mirror position has been confirmed.
As the mirror is moving up out of the way the camera would also be stopping down the aperture to the selected value. If the metering is set to Evaluative (or Matrix in the Nikon system) the camera could still be comparing the reading taken by the light meter when the mirror was down with a preloaded library of scenarios used to determine recommended or automated exposure. With the electrical communication between camera and body that is used by Canon's EOS system there is likely a confirmation of the position of the aperture diaphragm required before the shutter begins to release. Of course in our case since there is no lens this can also be eliminated as a possible cause.
With as many things controlled by microprocessors as there are in the modern digital camera, differences may also be attributable to the load on the camera's processing power at the time the shutter button is pressed. Battery voltages can also affect the speed at which a microprocessor runs. As the battery voltage drops, many processors reduce the frequency at which they process information. Thus, it takes longer for the camera to go through the long sequence of instructions and sensor readings that must take place each time a photo is taken.
Finally, there may well have been a slight mechanical difference that could explain the variation. Perhaps the lubrication of the wheels, gears, and linkages between the motor that moves the mirror and the mirror itself was a little thicker for the one sequence. Perhaps a brush in the motor itself was a little weaker than the others and the starting position of the armature in the motor can explain the difference.
Notice that the total variation between the time the mirror is up for each clip are still very slight compared to the length of the longer time gap between when the mirror is all the way up and when the shutter curtains actually begin to open. At 400X time dilation, the "slow" mirror is about 2.5-3 seconds slower than the others. Then there's about a 7 second gap between the time all of the mirrors are up and the time the shutters begin to move. Notice also that the transit time of the first and second shutter curtains across the sensor is near identical for all four clips, even when observed at 400X time dilation. So is the return of the mirror to the down position.
At most what we have observed is a difference of about 7 microseconds (1/140 second) between the time the mirror is up and the time the shutter begins to open. It then takes about 18 microseconds (1/60 second) before the shutter actually begin to open after the "slowest" mirror is up. Then only roughly 4 microseconds (1/250 second) are used for the first curtain to transit across the sensor.
All things considered, that's plenty precise enough for photography where the variation between target and actual shutter times and aperture openings are measured in increments of around 1/6 stop.
Cameras are designed to be as precise as possible. Yet, with mechanical moving elements, even when they are driven by electronics, there are always the issues of inertia, friction and wear. Thus, it is not surprising that in this particular case, one mirror decided to take a gentler approach in responding to commands. There is almost no way to provide 100% assurance when mechanical elements are involved.
In really old film cameras, which have not been maintained and stored well, you can see this in many elements, but more pronounced in shutter blades and mirrors.
This is one reason why camera manufacturers are touting mirrorless cameras, electronic shutters and so on.