[ Pending a proper review. ]
One of the selling points of the Fujifilm X100T, besides being a gorgeous camera that improves on its predecessor in so many ways, is the new Electronic Shutter mode (a feature that’s coming to the X-T1 as well) which enables shutter speeds up to 1/32000s, which is a ridiculously short amount of time.
This is possible because the camera is able to turn the sensor on and off that quickly. While normally the camera would turn the sensor on, open the mechanical shutter, close the shutter, turn the sensor off and then read the data, now the mechanical shutter remains open and the sensor is turned on for, say, a ludicrously short amount of time.
However! If you’re imagining that all the entire sensor’s switch is flipped on and off to take a picture, you’d be mistaken.
Instead, the camera turns on and off each line of pixels in the sensor, one at a time, from the top of the frame to the bottom. So the bottom line of pixels actually records the image a little later, resulting in an effect called rolling shutter. This can be pretty funky, as you can see here.
So how long does that actually take?
Since there was no official documentation on this topic that I could find, I enlisted my boyfriend’s help. He retired to the attic he calls his Engineering Laboratory and reassembled components he’d been tinkering with for a portable light modifier I’d requested, wrote some code, and an hour later our experiment could start.
Here’s the result:
And now the explanation.
The array of LED lights was hooked up to the computer via an Arduino and instructed to flash 100 times per second. The banding you see in the image is due to the rolling shutter effect: the lights were on when the first few lines of the sensor were being scanned, and off during the reading of the next few lines of the sensor, and on again for the next few lines, etcetera. As you can see, the red numbers on the mysterious device on the right were bright in their own right, so they were visible both when the LEDs were on and off.
In this photo, which is straight out of the camera, the shutter speed is reported at 1/1600s, but now we have a simple way to see how much time elapsed in the real world while each line of the sensor was getting it’s 1/1600s’ worth of light: count the bands.
I see seven bands of lines where the light was on, seven where it was off. The light flashed 100 times per second, so between the reading of the first line of the sensor and the bottom line, about 7/100ths of a second had passed. And ~0.07 seconds is more commonly known to photographers as:
~1/15th of a second.
Not terribly quick, really. I don’t have the steadiest of hands and would struggle to get a sharp shot at 1/15s, so even though the picture may be sharp, the image will likely be distorted because the camera moved a bit while the sensor was being scanned.
We investigated a bit further, to see if our conclusions were valid. First we bumped the frequency of the LEDs from 100 Hz to 200 Hz, producing twice as many flashes per second:
I count 14 bands here. 14/200 = 7/100 = 0.07s = ~1/15s, same deal. And this was at 1/4000s on the electronic shutter.
Next up: does the electronic shutter speed affect the duration of the scan? Let’s try at 1/32000s:
Out of focus and poorly framed, a thousand apologies – but we can still draw conclusions: flashing at 100 Hz, we see six visible bands of illumination and can deduce that there was a seventh at the bottom of the frame (but since there was no light there, we see nothing – a thousand more apologies).
And, for the hell of it, 1/30s:
Now we’re getting more ambient light, but even at 1/30s we can still identify seven bands of greater illumination.
Turns out, it doesn’t matter what you do. Using the electronic shutter on the X100T, regardless of the shutter speed you choose, you’re capturing 1/15s of real time – but you get to choose how many photons your sensor absorbs in that time. Still pretty cool.