Why a shutter?

[Dez]

I hear lots of complaints about the loud shutter in the A7 and especially A7R. My question is why does a mirrorless digital camera need a shutter at all?

I can't think of any reason at all for a physical shutter at the beginning of the exposure, and the only reason for a trailing shutter that occurs to me is if the polling of the sensor elements goes on AFTER the exposure, and the sensor is not capable of latching its exposure value until it can be polled. If that is the case though, it is hard to see how TTL flash metering can be effective, as it would require measurement of the levels during the actual exposure. Presumably the sensor elements require a certain amount of time to integrate the light falling on them to come to an acceptable level, but why would this require them to be blacked out for an instant before and after the exposure?

How exactly does this work?

Cheers,
Dez

 

[sevo]

There is no realtime TTL flash metering in digital cameras (except for a few early examples that did a film-like off-the-sensor reflected light reading, a problematic approach given the mirror like reflections and near black colour of sensors) - they do it by measuring a pre-flash with the sensor, deleting the sensor and exposing then.

The cameras can do away with the first curtain (the sensor can be deleted faster than a shutter moves), but not with the second, unless they read the sensor line-by-line like a video camera - which causes "rolling shutter" issues, most notably motion distortion.

 

[Dez]

Thanks for the insight on TTL flash. I thought it might be something like that.

It would be necessary to poll the sensor elements one at a time in any case. There is a function that clears the sensor values, making the pre-exposure curtain unnecessary. If there is also a function that latches the light levels in the elements so they would not be disturbed by light hitting them during a post-exposure scan, this polling can be done relatively slowly, and a second shutter operation is not needed.

I understand the principle of a moving scan, and the distortion it causes, but focal-plane shutters have the same issue, and have been around pretty successfully for about 100 years now. Would we expect this distortion to be much worse in a device like a Sony A7?

Cheers,
Dez

 

[1joel1]

I haven't read about it in a while, but besides flash synching, which I would gladly give up, there are also some heat issues. Do a google search on "global shutters" and you can learn how this will someday, soon I hope, be applied to still cameras.

Joel

 

[rwintle]

This is one thing I very much liked about using our (now old) Sony DSC-R1 (apart from the yummy Zeiss Vario-Sonnar T* glass) - turn all the beeps and fake electronic shutter noise off and it's a silent shooter - with flash sync as fast as you please.

Not sure how this very fancy "bridge" camera differs electronically from modern mirrorless ILC's, but it sure was handy for shooting events where silence was desirable. Let down by horrid noise over ISO 400, unfortunately. 🙁

 

[Dez]

An interesting article here:
http://www.red.com/learn/red-101/global-rolling-shutter

It must cost a good deal more to add the mechanical shutter to the design, and as we have seen, it causes a good deal of annoyance in some potential customers. The "rolling shutter" as described in this article is like a focal plane shutter with a slit width of one picture element. Such a design applied to a digital camera would not work with electronic flash, so does not appear to be usable. But why isn't there full electronic control of the shutter? There must be an overwhelming technical issue. Clearing the memory is straightforward, but there must be a problem with electronic control; there needs to be the ability to gate the sensor elements' ability to gather light by a control line. Obviously this isn't easy to do.

Perplexed,
Dez

 

[Godfrey]

The DSC-R1 has a fixed lens and an in-lens leaf shutter, which is what makes it so quiet.

Sensors are serial devices: the data is read from them serially, line by line, in blocks of lines, or even "full frame transfer" (rare, implemented in the old Kodak CCD sensor that the Olympus E-1 uses). But always serial. (Note that high-end cameras like the Nikon D3/D4, etc, implement multiple DAs and multiple channels for readout which is what allows them to fire at very high sequence speeds successfully. But even this is not fast enough for a good all-electronic shutter implementation.)

There's a limit to how fast that serial read-out can operate, which is the root reason for why there are no large-sensor cameras without a mechanical shutter, why you get "rolling shutter" effects on continuous capture, etc. Even the Sony A7 EFCS (electronic first curtain shutter) implementation has some issues at very short exposure times ... the upper 1/3 of the image area will show a gradient darkening to the edge of the frame with EFCS enabled.

The serial nature of the data transfer has to observe a boundary condition due to the distance the signals have to travel. The larger the sensor, the more pixels there are, the more likely there will be rolling shutter and other limitations on implementing an all-electronic shutter. You can see these same limitations in every current largish sensor camera that implements electronic shutter or EFCS.

Until sensor and supporting data transfer technology improves another quantum leap, and read-out speeds double or triple at least, only fairly small sensors can be designed to operate with entirely electronic shutters.

G

 

[Dez]

Understood that the readout is of necessity serial, and can easily take a lot longer than the actual exposure, although less time than the second mechanical shutter operation leaves the sensor dark. The so-called "electronic first curtain" is presumably a pulse that zeros out the readings of all sensor elements. This is a parallel bus, and the clearing would take a very short time.

What I am hoping to see is a second parallel bus which would be an "expose enable" signal, which would only allow the sensor elements to respond to light while it is activated. If such a signal were possible, it would be turned off at the end of the timed exposure, and the readings could be gathered relatively slowly after the exposure. My problem is that I don't understand the physics of the sensor. It must be that such an "expose enable" control is impossible, or technically impractical for a reason I don't know. As an electrical engineer, that is a bit embarrassing, although that technology is nowhere near my own field.

Cheers,
Dez

 

[Godfrey]

Well, I'm not an electrical engineer. But several friends of mine are, and two of them are specialists in imager sensor design (one at Pelican, I think, the other at that HP spinoff whose name I can never remember). They've explained to me how sensors work a dozen times, but it doesnt stick as I haven't the background to understand details at that level.

Suffice it to say that they, and everyone else in that business, is looking for the next breakthrough. Global electronic shutter on large scale devices is a particular holy grail...

G

 

[Wenge]

I think rounded pixels via hex 6 or 8-sided would be interesting (vs. simple square px)

 

[pvdhaar]

Understood that the readout is of necessity serial, and can easily take a lot longer than the actual exposure, although less time than the second mechanical shutter operation leaves the sensor dark. The so-called "electronic first curtain" is presumably a pulse that zeros out the readings of all sensor elements. This is a parallel bus, and the clearing would take a very short time.

What I am hoping to see is a second parallel bus which would be an "expose enable" signal, which would only allow the sensor elements to respond to light while it is activated. If such a signal were possible, it would be turned off at the end of the timed exposure, and the readings could be gathered relatively slowly after the exposure. My problem is that I don't understand the physics of the sensor. It must be that such an "expose enable" control is impossible, or technically impractical for a reason I don't know. As an electrical engineer, that is a bit embarrassing, although that technology is nowhere near my own field.

Cheers,
Dez

Could it be that the gates and routing on the sensor is already so densely packed that there's no room for frills like that? I wouldn't be surprised if it isn't even synchronous logic. Think in terms of a DRAM memory cell versus a properly clocked register.. A sensor is probably quite similar: sacrifice everything possible for the pixel count and squeeze the reset together as far as it will go..

The EFCS can even be a bonus that's not intentionally designed in. Like, if we twist its ears and flip it over backwards, it does this weird thing like going blank..

 

[YYV_146]

I think eventually small system cameras will go to global, electronic shutters.

FYI there are cameras currently available that use all-electronic shutters, such as the GM1. It can do 1/16000 under the full electronic shutter, but there is significant rolling shutter effect when shooting moving objects.

 

[sevo]

It must be that such an "expose enable" control is impossible, or technically impractical for a reason I don't know.

So far, there is no known semiconductor whose photosensitivity can be electrically controlled and which is suitable for application at sensor scale and camera sensitivities. So that would imply a transfer to a on-chip temporary storage - possible, and has been done, but existing chips of the type have so many issues (like increased bleed, rolling shutter, overexposure vulnerability) that there are not that many situations where they chosen over a plain shutterless or mechanically shuttered variant (or a electronic global shutter as a separate system in front of the sensor).

 

[rwintle]

The DSC-R1 has a fixed lens and an in-lens leaf shutter, which is what makes it so quiet.

You know, we've owned that camera for many years and I never knew that(!). It really does shoot completely silently, as quietly as a pocket point-n-shoot. Thanks for the info.

(guess I should go and read the manual again... 😉)