today I had some spare time to think about the technical properties of modern cameras and it became clear to me that I am not fully understanding the concept of the mechanical shutter in cameras which use a digital sensor instead of film.
I understand that when using the electronic shutter, the rolling shutter effect is noticeable when capturing fast moving objects which can be reduced with short readout times of the sensor, but I don't understand why using a mechanical shutter can prevent this!
Is the sensor read out in a different way (not line by line) when using a mechanical shutter? Why does it make a difference to quickly shut off any further light by closing the curtains? Do the sensor pixels itself have a kind of "memory" which can then be read out in a way that is not time critical, not having the risk that the sensor data is being changed afterwards (since everything is dark?)
Like ... does the sensor act like a collection of many tiny "batteries" which are being charged by incoming light? And when the incoming light stops, they keep their charge?
I checked out some explanation videos on YouTube, but many of them only seem to compare the pros and cons of e-shutter vs. mechanical, without explaining the "why" behind it! I would love if someone could explain it to me. Thanks in advance!
Usually the mechanical shutter traverses the frame in a much shorter time (e.g. 1/250s) than the electronic shutter (e.g. 1/25s). The mechanical shutter still produces rolling shutter effects, but they are much less noticeable.
Both MS and ES are forms of focal plane shutters, although they don't both actually operate in the focal plane -- the MS operates in front of it. When the scan times of the two are the same, they have similar distortion (but not quite the same distortion) when capturing fast moving objects. But most mechanical shutters scan faster than most electronic shutters.
I think it's a very good question. From MY understanding, the OP is right - while the sensors are exposed to light for less time when using a mechanical shutter, the actual readout takes just as long as an electronic shutter - it's just that the the theory says that the data captured is from (effectively) the same point in time and so there is much less rolling shutter.
But what I don't understand (and maybe this is the point of the original post) is how a mechanical shutter capture can do it so much quicker than an electronic one. What am I missing?
Let me explain what I don't understand:
ES - the button is pressed at (let's say 1/10th of a second shutter speed). At that moment in time and for a 100 milleseconds afterwards, the sensor is read - one line at a time (OK, maybe a few) and it takes a finite time to read all the lines. Which produces rolling shutter. This I understand.
MS - the button is pressed (at 1/10th second). At that moment in time the first and second curtain start traversing the sensor exposing light to each of the lines of the sensor. BUT (and here's my dilemma), the sensor cannot be read any faster than using ES and yet, the sensor is producing less rolling shutter. So how is the data stored until it can be read? and why cannot than method be used for ES?
ES: There's a rolling reset that travels across the sensor at the rate it will be read out. That starts the exposure. There's a rolling read that travels across the sensor at the same rate after a delay set by the shutter speed. That ends the exposure.
MS: There's a rolling curtain that opens. That starts the exposure. There's another curtain that covers the sensor that follows after a delay set by the shutter speed. That ends the exposure. Then the sensor is dark and the electrons are sitting there, and the readout takes place at a pace that is not critical.
Ah, thank you. So the sensor is not read in real time with MS.
So, why can't the do that with ES? Is it because the sensor reads continuously and you can't shut it off (so, with MS, the sensor is still reading, but with no light, there's nothing added).
[edit] - not that's not right. With ES the sensor is read, a pixel (or a line) at a time and you have a 'switch on and switch off' time relating to "shutter speed". But you can only switch the sensor on a line at a time. No - that doesn't make sense either because that's not how MS works.Unless with MS, you turn on every line in the sensor, then expose it all, then read what has been exposed. Then turn the sensor off, a line at a time.
Absent global shutter, the only way to end the exposure with ES and today's CMOS sensors is to read the data out. Slower sensors are read a line at a time. Faster ones read more than one line at a time.
With MS, and with ES, you don't have to turn off the sensor after taking a picture, but you do have to put it into a different mode between exposures if you have an EVF.
With MS, the reset takes place before the first curtain is opened. So it's not time-critical either.
So, with ES, the sensor is reset (which clears out any old data, I assume) and then it's read, line by line. which then ends the exposure, line by line. And it's the delay in reading each line which determines the 'shutter' speed.
This assumes that the reset happens line by line as well (if the reset happened to the whole sensor at the beginning, thne the last line to be read would have gathered more light than the first one).
With MS, the whole sensor is reset, shutter is opened and closed and then the sensor is read.
[I'm trying to get it im my head visually - which is how I think]
No. For mechanical, electronic, or EFCS shutter, a key design aim is to expose each point of the sensor for an equal amount of time, and this is acheived quite effectively, in spite of the engineering difficulties in implementing mechanical and EFCS shutters.
The manufacturer may - or may not [!!] - have the same design target exposure times for a given "shutter speed" setting, when electronic or mechanical shutter is selected.
Traditionally, exposure times scale in powers of 2, but to avoid scaring the horses, nominal shutter time settings are displayed.
For example, a setting of 1/1000 s is taken to mean 1/1024 s (1024=2^10), and a setting of 30 s is taken to mean a 32s exposure (which catches out folk new to intervalometers). See: www.scantips.com/lights/fstop2.html
I have a camera which, at a shutter speed setting of "1/30", gives an exposure time of 1/(29.92 +/- 0.01) s with electronic shutter, and a shutter speed of 1/(31.995 +/- 0.005) s with mechanical shutter. This "feature" would be much more useful if it were documented.
Maybe, maybe not. It can depend on the manufacturer, on whether you are shooting stills or video, perhaps on the camera drive mode, or other things.
On Sony and Canon (AFAICT) readout for stills takes longer with mechanical or EFCS shutter than with electronic shutter.
On Panasonic Full-Frame, readout for stills takes the same time for mechanical, EFCS, or electronic shutter (~51ms for S1[H]), but is faster for "open gate" (i.e. full-frame 3:2) video (~30ms for S1[H]).
Canon gives 12-bit RAW output with electronic shutter; Sony gives 12 or 14-bit output, depending on model; Panasonic S-series (Full-Frame) gives 14-bit.
I don't think that these differences reflect a fundamental difference in technology, rather design choices.
(Getting rather abstruse) I suspect this circle could be squared by using variable slope ADCs, but there are quite a few non-public electronic details that could get in the way.
That's confused - I'm not sure what it means.
The technogy of good quality mechanical shutters is quite amazing. This technology has been continuously refined over the last 50 years, particularly the last couple of decades of the 20th century.
(With a conventional active-pixel CMOS sensor, and with some other kinds of sensor):
Mechanical shutter decouples sensor readout from sensor exposure: The shutter exposes the sensor, then the sensor is read out. Then the shutter is re-cocked.
With electronic shutter, the "First Curtain" is implemented by bringing rows of the sensor out of reset. The second curtain is implemented by row readout and associated analogue-to-digital conversion. Readout is much slower than reset, so the "rolling shutter", or "shutter travel time" is limited by the speed of row readout.
All the pixels in a row (or perhaps more than one row) are read out, and converted to digital values, in parallel. So, for a given readout mode, the readout time is proportional to the number of rows read. This is why high video frame rates often involve a sensor crop.
Between the mechanical second curtain and readout, the pixel data is stored in the capacitance associated with each pixel's photodiode. It's desirable to keep the time between mechanical second curtain and readout short, as "dark current" noise accumulates in the pixel capacitors. But this isn't much of a problem at normal readout timescales.
This can't work with electronic shutter because the readout is the second curtain.
Edits & Corrections:
1. Sony MILCs aren't always 12-bit with electronic shutter (Jim Kasson)
2. My measurement for Panasonic S1 electronic shutter speed at a "1/30" setting was 1/(29.92 +/- 0.01) s, not 1/(29.995 +/- 0.005) s. (I misread my spreadsheet).
