CMOS sensors read each line sequentially. How long each line is "on" and "off" determines exposure time. But just as with mechanical focal plane shutters, the total time needed to take an exposure is the same, regardless of the exposure time. It takes any given sensor the same amount of time to read out each line of the sensor regardless of how long each line was "on". This provides a limit to the maximum number of frames that may be captured per second (assuming the exposure time is shorter than the readout time - otherwise the exposure time is the limit on fps).
Since the sensors in smartphones do double duty to not only record images, but to also give an almost real time preview on the phone's screen in lieu of an optical viewfinder, exposures longer than the frame rate of the video feed supplied to the phone's screen remove the ability to provide a constantly updated stream of frames to the screen during exposure.
If the operation is electronic, does that mean there's no inherent fastest or slowest "shutter speed" for a given smartphone?
That depends upon what one means by inherent. There are physical limits to how quickly a line or photosite on a sensor can be switched on and then off. There are also limits to how short an exposure time can be and still allow enough time to collect meaningful information (i.e. collect enough photons to rise above the noise floor for a given sensor). Currently in most typical lighting conditions and scenarios, the latter is reached at a longer exposure time than the limit imposed by the former.
Some background information for those who may not be knowledgeable about how mechanical focal plane shutters and sensor with "electronic shutters" that read the output sequentially line-by-line rather than a sensor with a "global shutter" that reads the entire sensor all-at-once. (Hint: There are a good number of existing questions here that go into great detail about how various types of mechanical and electronic shutters operate.)
So I don’t get of on the wrong foot, could you please address my confusion on the first few sentences? 1. CMOS sensors read each line sequentially. How long each line is "on" and "off" determines exposure time. The first sentence is about reading; the second is about exposure. I’m not grasping how they relate to each other. 2. But just as with mechanical focal plane shutters, the total time needed to take an exposure is the same, regardless of the exposure time. I haven’t been able to parse this.
The sensor can not be collecting photons that accumulate charges in each photosite (exposure) and move those charges off the sensior (readout) at the same time, Just like you can't pour the water out of a rainbucket into a measuring cup and catch rain falling on your yard at the same time. When a sensor is read it is "cleared" all charges are removed. When it is "turned on" it begins accumulated charges created by photons falling on each photosite. When it is "turned off" it stops accumulating charges created by photons falling on each photosite.
The time difference between when each photosite is turned on and turned off is the exposure time. That is, it's the total amount of time the photosite was collecting charges.how long it takes the sensor to sequentially turn on each line on the sensor is fixed. It (usually) starts at the bottom (which is the top of the inverted image) and goes line by line to the top (the bottom of the inverted image).
In much the same way mechanical shutter curtains in modern cameras transit the sensor at a fixed rate. It's the difference between when the first curtain begins opening and the second curtain begins following it in the same direction across the sensor that determines exposure time. At short exposure times the entire sensor is never uncovered at the same time. It may take 4 milliseconds (1/250) for each shutter curtain to transit across the sensor. If the second curtain begins closing 1 millisecond after the first curtain begins opening, the first curtain has only revealed 25% of the sensor.
As the first curtain continues travelling across the face of the sensor the second curtain is "chasing" it 1ms (6mm for a 24mm high FF sensor) behind. When the first curtain finishes opening, revealing the bottom of the sensor to light, the second curtain has already covered up the first 75% of the sensor. 1ms later it finishes closing to completely cover the sensor. The GIF in the linked question and linked video (below) shows how it takes about 2.5ms (1/400) to take exposures of 1/1000, 1/2000, 1/4000, and 1/8000 by decreasing the size of the slit between the two curtains as they transit the sensor (The transit time for the Canon EOS 7D is about 2.5ms).
If a rolling shutter travels from top to bottom, why does this image seem to show skew in the other direction?
The accepted answer includes a link to the entire video.
With long exposure times it's the same thing. Let's say we use 1 second exposure with a camera that has a shutter transit time of 2.5ms (1/400). The first curtain opens and takes 2.5 milliseconds to cross from the top to the bottom of the sensor and uncover the sensor. Exactly one second after the first curtain began opening the second curtain begins to close. It also takes 2.5ms to cross the sensor and cover it back up. Thus from the time the first curtain began to open until the second curtain completely closes is 1,002.5 ms. But no part of the sensor has been uncovered more than 1,000ms.
When the exposure time is shorter than the transit time, the second curtain begins closing before the first curtain is completely open. The total time it takes to make a 1/1000 exposure is the transit time (2.5ms) + the delay between first and second curtain (1.0ms), or 3.5ms. No spot on the sensor is uncovered for more than 1ms, but the photosites on one edge are exposed 2.5ms earlier in time than the photosites on the far edge are exposed.