Electronic Shutter on the SL

[Jared]

Advertisement (gone after registration)

On another thread an interesting topic came up a couple days ago with regard to the implementation of the electronic shutter on the SL.  At least, it's interesting to me as I really get into the technical aspects of photography.  If you are only interested in the artistic aspects of photography, I recommend you stop reading here as what I'm going to cover has very limited practical applications, and will probably not help you take better pictures.  It's more an academic exercise.

 

On the other thread, someone mentioned that he/she was avoiding using the electronic shutter (1/12,500s and above) on the SL since users of some Sony cameras had noticed a problem with image quality when using the same feature on their cameras.  I decided to dig a little deeper and actually measure the performance of the electronic shutter on the SL to see if I could find a difference.

 

First some background.  What really is an electronic shutter, and how could it possibly impact image quality?  With a normal mechanical shutter on a camera with an EVF like the SL, here is the basic process:

 

1) Shutter is wide open when the camera is on in order to collect light for the EVF.  It just stays wide open.  The CMOS chip is read out one line at a time, and then the chip is emptied of charge electronically and the process starts over again.  This happens something like 30 to 90 times per second (depending on the amount of light available) in order to give a "real time" view to the EVF.  Mechanical shutter is not used at all when you are looking at the EVF.

2) The photographer presses the shutter release.  The shutter is closed.  The lens is stopped down to the indicated f/stop.  The CMOS chip is emptied of charge and made light sensitive.  

3) Then the exposure starts.  The first curtain of the shutter starts to open up and the CMOS chip starts to detect light.  Since the SL has a focal plane shutter, not all portions of the chip receive light at the exact same moment, but it's pretty quick--something like 1/300s for the first curtain to get out of the way.  The entire CMOS chip is in "light gathering" mode for the entire exposure--only the shutter is keeping the chip from receiving light.

4) At a time determined by your selected shutter speed, the second curtain starts to close the shutter.  This may or may not be before the first curtain has completed it's trip--doesn't really matter.  If you have selected a fast shutter speed, like 1/500s or faster, the first curtain won't yet have completed its journey and you effectively have a slit of light moving across the focal plane to expose the CMOS chip, but all parts of the chip are exposed for the same total duration.

5) Once the second curtain has completed its journey, the camera starts the process of reading out the image data from the chip.  The designers of the camera must decide how fast to read out the data--and this is critical.  If you read it out very quickly, you actually increase the amount of noise in your images.  If you read it out very slowly, you get lots of blackout and a low frame rate for your camera.  It's a balance.  I don't know how long it takes to read out all 4,000 lines on the SL, but it's probably something between 1/15s and 1/30s.  

6) The shutter is again opened, and the camera goes back into electronic shutter mode in order to provide an image to the EVF.

 

The important point for this thread is that the read out speed can be (comparatively) slow.  The camera doesn't need to read out the chip very quickly while the shutter is closed since no light can get to the chip, so a compromise between camera performance and read noise can be reached.

 

Many manufacturers of mirrorless cameras have implemented a feature called EFCS or Electronic First Curtain Sync.  Here is the idea behind it:

 

When the first curtain reaches the end of its path in step 3 above, it can introduce some vibration into the camera.  This vibration is WHILE the camera is busy taking a picture, so it has the potential to introduce motion blur.  All shutters are damped to minimize this effect, but it is still visible with some cameras.  This is particularly an issue as camera resolution gets higher and higher; you don't have much room for vibration if you are using a 40+ megapixel chip as on the Sony A7RII cameras!  

 

If you instead change your process, release the first curtain early and wait for vibrations to end before actually emptying your chip of charge and starting your exposure, you can use the electronic shutter as the first curtain--turning the rows "on" one at a time at the proper interval ahead of the second curtain--then you don't need to worry about shutter vibration.  The second curtain will still cause vibrations when it slams home, but that is only after light has been cut off from reaching the chip so it doesn't matter.

 

This is called EFCS, and it seems to work quite well.  Not all cameras seem to need it.  There is a wide variety of shutters out there, and a wide variety of camera designs, and there is therefore a wide variety of vibration levels induced by the shutter.  Obviously, the effects are most noticeable when shooting a very high resolution camera such as the A7RII.  Query whether the SL could benefit from an EFCS?  No way to tell since you would need empirical data.  In any event, EFCS is now a common feature in higher megapixel cameras and does, in fact, work.  The one down side I am aware of is some additional shutter lag.

 

What about purely electronic shutters?  When the mechanical shutter is not used at all?  In "Silent Mode" (on Sony cameras)?  Why use a mechanical shutter at all if you can just turn rows off and on whenever you like?  You could be silent all the time if you wanted!  Well, it turns out there are a couple downsides to electronic shutters, and they go back to step 5 above.  The only way to stop a chip from continuing to collect light once it has started is to read out the row.  And if you try to read it out too fast you will degrade image quality (by adding read noise).  If you read it out at a more "optimal" rate you get what's called rolling shutter.  

 

Imagine you are turning the rows off and on one at a time so they each collect 1/500s exposure, for example.  And you are reading them out at that rate I mentioned earlier--the reasonable compromise I mentioned earlier where it takes you something like 1/15s to read out the entire chip.  Any given row received 1/500s of exposure time, but not every row received light at the exact SAME time.  You can get some weird effects.  Imagine you are panning on a moving car and there is a lamp post in your frame.  Any given part of the image receives 1/500s exposure, but the top part of the image might be exposed 1/15s before the bottom part.  The lamp post will be in an entirely different position at the beginning and end of the 1/15s due to your panning.  It will appear to bend!  This is called rolling shutter.

 

So how can you avoid rolling shutter?  Just read out the chip faster.  Instead of 1/15s to read the whole thing out, imagine you only take 1/60s or 1/125s.  That would minimize rolling shutter, but at what cost?  Read noise goes up!  Image quality degrades.

 

For a purely electronic shutter, the designer can choose either approach--read out really fast  in order to minimize rolling shutter, or read out at the "optimal" rate in order to maintain the best possible signal-to-noise ratio.  Neither decision is "right" or "wrong"--they are just different compromises.  I was curious which decision Leica had made.  Were they reading out faster than normal when you choose 1/16,000s or 1/12,500s?  Or reading out at the normal speed--same as they do with the mechanical shutter--and risking rolling shutter?

 

Here is what I did:

 

1) Took multiple exposures of a brightly lit, neutral object with 1/16,000s shutter selected and the aperture wide open

2) Took multiple exposures of the same brightly lit, neutral object with 1/8,000s shutter selected and the lens stopped down one stop, same ISO

3) Brought these images into Photoshop and measured the SNR (signal / std deviation of signal) of each frame

4) Averaged the SNR for the various frames shot at 1/16,000s (electronic shutter)

5) Averaged the SNR for the various frames shot at 1/8,000s (mechanical shutter)

6) Compared the averages for the two types of frames

 

Here is what I found:

 

In all cases, the SNR was measurably higher with the mechanical shutter than with the electronic shutter.  It appears that in the case of the electronic shutter, Leica is choosing to read out the chip faster than with the mechanical shutter, likely to minimize rolling shutter effects.  

 

Is it enough to matter?  Would you be able to tell the difference in a real image?  Should one shoot with a neutral density filter and avoid using the electronic shutter?  Well, I haven't done any tests with real photographs, but based on the numbers I was getting I'd say it's highly unlikely you could tell the difference. 

 

You've got to look at why you would use a fast shutter speed like 1/16,000s.  There are two practical possibilities.  Either you need that shutter speed to stop some insanely fast action, or you are trying to shoot a fast lens wide open in bright lighting conditions.  I think the second scenario is much more likely.  In the second scenario, by definition you have plenty of light coming in so shot noise just isn't going to be a problem.  You're going to be shooting at base ISO, so you don't need to worry about gain hurting your dynamic range.  The only question, really, is just how much is your noise floor creeping up in this sort of an image, and would you therefore notice any loss in dynamic range.

 

Basically, I was getting a SNR of around 21.5db with the mechanical shutter and around 20.3db with the electronic shutter.  That's about 1/3 of a stop of range.  Basically, you'd see the same difference if you moved from shooting at a base ISO of 100 to shooting at 125.  Can you measure the difference in a computer?  Absolutely--easy to do.  I was able to do it faster than I could write up this post.  Are your images going to be noticeably inferior by having your noise floor move up 1/3 of a stop?  I don't think you could see it in the real world.  

 

If anyone has questions on my technique or want to offer a correction, I'd love to hear from you.  Again, I see this as a purely intellectual exercise--I am convinced you'd never be able to tell the difference in the real world, and whatever approach you take it isn't going to affect the artistic merit of your picture.  

 

Now back to actual photography...

 

- Jared

[rpittal]

Thank you Jared for the great explanation and the tests you conducted. I am still a bit confused why the SL doesn't have a completely silent mode (camera very quiet anyways) for all shutter speeds, like the options on the Sonys. From your explanations, it seems that the SNR would likely increase with lower amounts of light for the slower shutter speeds in many cases? It would be interesting to repeat your tests using a Sony (or other camera) with silent mode for all shutter speeds.

Bob

[Advertisement]

The Leica Q3 is available for pre-order at B&H Photo Video and Adorama

[FlashGordonPhotography]

When shooting above 1/1000 with the A7 EFCS the deterioration in image quality compared to a full mechanical shutter are dramatic and obvious. It's not just some noise, although there's plenty of that. There's also severe vignetting and colour shifts as well. The sensor simply couldn't read out the information quickly enough before the mechanical shutter caught up and started to reduce the exposure as the curtain moved across the chip. I experienced this myself.

 

As each generation of sensor is released it gets a bit better and the current sensors can read at higher speeds, like the SL where the sensor and shutter are better implemented. The true next step will be a global shutter sensor big enough for a camera like ours.

 

Gordon

[Jared]

Thank you Jared for the great explanation and the tests you conducted. I am still a bit confused why the SL doesn't have a completely silent mode (camera very quiet anyways) for all shutter speeds, like the options on the Sonys. From your explanations, it seems that the SNR would likely increase with lower amounts of light for the slower shutter speeds in many cases? It would be interesting to repeat your tests using a Sony (or other camera) with silent mode for all shutter speeds.

Bob

 

 

If you look for it, I believe people have already published on the web some fairly extensive testing of the SNR's for the Sony A7RII with various combinations of mechanical shutter, EFCS, and silent shutter.  Sony is quite aggressive in how the handle the readout when in full "silent" mode--they read the chip really quickly and even lower the bit depth on the file.  I assume this is in an effort to reduce the effects of rolling shutter--it's certainly intentional, not an oversight.  The difference in SNR is quite a bit larger with the Sony under these circumstances--the silent mode takes a bigger bite out of the image quality.  There are enough people out there who have run tests on the Sony's that I expect you could make a good determination whether silent mode is for you when shooting Sony cameras.  I expect it would be a very welcome option for a wedding photographer, for example.  Doesn't make much of a difference to me.  It doesn't appear to me that Leica was as aggressive in their approach to the electronic shutter--some effort to speed up read speed, yes, but not so much that it degrades image quality significantly.  Sounds like the typical Leica approach, for better or for worse.

 

- Jared

[Jared]

When shooting above 1/1000 with the A7 EFCS the deterioration in image quality compared to a full mechanical shutter are dramatic and obvious. It's not just some noise, although there's plenty of that. There's also severe vignetting and colour shifts as well. The sensor simply couldn't read out the information quickly enough before the mechanical shutter caught up and started to reduce the exposure as the curtain moved across the chip. I experienced this myself.

 

As each generation of sensor is released it gets a bit better and the current sensors can read at higher speeds, like the SL where the sensor and shutter are better implemented. The true next step will be a global shutter sensor big enough for a camera like ours.

 

Gordon

 

Yes, I know there were/are issues with color shifts and banding under certain lighting conditions (LED's and fluorescent lights).  Did you see issues on your cameras under natural light as well?  I hadn't read that, but I don't have a Sony camera so I don't tend to keep up with their quirks and problems.  Each brand has its own, of course.

 

Not getting the timing right on the EFCS kind of surprises me--you don't need to have a super fast read speed to stay ahead of the second curtain. You just need to START the exposure of each row at the correct time.  If you are slow to read everything out, it shouldn't matter since the shutter will keep the chip from accumulating more charge while you take your time getting around to the reads.  Strange that they didn't do a better job implementing that feature.  I know the original A7R had significant problems with shutter shock, degrading much of the benefits of that high resolution chip.  The EFCS feature was supposed to address the issue and bring back all the quality the chip and lenses were capable of.

 

- Jared

[rpittal]

Thanks to Jared and FlashGordon for further explanation. I leave EFCS on for my SL and it is great to be able to use fast lenses wide open in the sun; just surprised when it is completely silent (usually check that exposure was actually made). Sounds like I should not use silent mode on my Sonys, but just EFCS.

Bob

[FlashGordonPhotography]

Advertisement (gone after registration)

Yes, I know there were/are issues with color shifts and banding under certain lighting conditions (LED's and fluorescent lights).  Did you see issues on your cameras under natural light as well?  I hadn't read that, but I don't have a Sony camera so I don't tend to keep up with their quirks and problems.  Each brand has its own, of course.

 

 

- Jared

 

Yep. tested extensively on the original A7. The original A7R doesn't offer EFCS for this reason. &2 and 7R2 both have ECFS and there is less issues but some changes. Shots at speeds between 1/250 and 1/8000. Same lens same time. ECFS on and then OFF. natural light. The difference was huge. And I mean huge. Worse the higher the SS got.

 

Gordon

[Jared]

Thanks to Jared and FlashGordon for further explanation. I leave EFCS on for my SL and it is great to be able to use fast lenses wide open in the sun; just surprised when it is completely silent (usually check that exposure was actually made). Sounds like I should not use silent mode on my Sonys, but just EFCS.

Bob

 

 

Just a minor correction... The SL has electronic shutter at higher shutter speeds, but it isn't EFCS.  It's just a pure electronic shutter.

 

- Jared

[steppenw0lf]

The electronic shutter of the SL is there for 1/16000. Probably also for the middle between 1/8000 and 1/16000. (3/32000, about equal to 1/12500 ?)

What if ISO is changed in thirds ? Are there two more speeds ? (between 1/8000 and 1/16000).

With some Zeiss ZM lenses you could produce third stops - how precise would the SL react to that ?

 

(Not important in practice, but just trying to understand what the camera offers.)

Edited February 12, 2017 by steppenw0lf

[ramarren]

The electronic shutter of the SL is there for 1/16000. Probably also for the middle between 1/8000 and 1/16000. (3/32000, about equal to 1/12500 ?)

What if ISO is changed in thirds ? Are there two more speeds ? (between 1/8000 and 1/16000).

With some Zeiss ZM lenses you could produce third stops - how precise would the SL react to that ?

 

(Not important in practice, but just trying to understand what the camera offers.)

 

 

It should be easy for you test that out, Stefan. Why not do that and report back your findings?

Edited February 13, 2017 by ramarren

[Jared]

The electronic shutter of the SL is there for 1/16000. Probably also for the middle between 1/8000 and 1/16000. (3/32000, about equal to 1/12500 ?)

What if ISO is changed in thirds ? Are there two more speeds ? (between 1/8000 and 1/16000).

With some Zeiss ZM lenses you could produce third stops - how precise would the SL react to that ?

 

(Not important in practice, but just trying to understand what the camera offers.)

 

 

As far as I know, there are no third stops in shutter speeds on the SL, just half stops.  So 1/12,500s and 1/16,000 are electronic shutter settings but that's it.

 

- Jared

[scott kirkpatrick]

Is there a spec somewhere for the SL's readout rate in either mechanical or electronic shutter mode?  Assuming that it is at least faster than the viewfinder's pretty good refresh rate it should be somewhere better than 1/100 sec.  Panning swiftly across a telephone pole to show a rolling shutter effect might make this measurable, if there is some way to calibrate the panning speed.  Perhaps having a friend drive their Porsche (or Mercedes SLK) across in front of the pole at a high but known speed...

 

scott

[ramarren]

As far as I know, there are no third stops in shutter speeds on the SL, just half stops.  So 1/12,500s and 1/16,000 are electronic shutter settings but that's it.

 

 

I don't know for a fact whether the shutter can operate only at discrete settings or continuously variably (in A and P modes), but there are a few more options for the electronic shutter speeds possible. 

 

I checked my library of ALL SL exposures recorded in LR and found the following shutter settings there: 

1/8000 sec Mechanical

1/10000 sec

1/12000 sec

1/16000 sec

 

I then switched the camera to M mode and found I could make these settings: 

1/8000 sec Mechanical

1/10000 sec

1/12500 sec

1/16000 sec

 

In the recorded settings from actual exposures, there are more than just whole, half, and third stop shutter settings recorded. From this I conjecture that the camera can set any eShutter speed from 1/8000 sec up to the limit at 1/16000 in A or P mode. What it does, in practical terms, I can't say ... I only had five exposures at > 1/8000 second in my library. :-)

[Jared]

So I poked around a little bit in the menus... The SL will allow you to set exposure in ⅓ stop increments, which is how you got the 1/10,000s intermediate value. Mine is configured to half stops (to match Leica's lens click stops), so only electronic values are 1/12,500s and 1/16,000s.

 

- Jared

[Jared]

Is there a spec somewhere for the SL's readout rate in either mechanical or electronic shutter mode? Assuming that it is at least faster than the viewfinder's pretty good refresh rate it should be somewhere better than 1/100 sec. Panning swiftly across a telephone pole to show a rolling shutter effect might make this measurable, if there is some way to calibrate the panning speed. Perhaps having a friend drive their Porsche (or Mercedes SLK) across in front of the pole at a high but known speed...

 

scott

Just because the readout is quick when running the EVF does not necessarily mean the readout is at the same rate for actual exposures. In fact, it is likely different. Keep in mind, the EVF has around 1,000 rows per picture not the 4,000 rows for the imaging chip, so they may only be reading out every 4th line or so for the EVF. Hard to know.

 

Also, while there is a reason to pick a slower read rate when actually capturing an image--to minimize read noise--no such pressure exists when just running the EVF, so you might well choose a higher read rate to avoid lag in the EVF.

 

I don't think we should assume just because the EVF can handle 60fps or more as a refresh rate that the readout during a real exposure is around 1/100s. Leica will have chosen whatever they think is the best compromise of image quality vs. camera performance.

 

To pick an extreme example... I have an 11 megapixel astronomy camera that reads out over about 30s or so (in an effort to minimize read noise). That is so slow that if you look at 1/10s exposure taken with the lens cap on, you actually see a gradient across the mostly black image just from the thermal noise that accumulates during the 30s!

 

Leica doesn't offer a lot of insight into their approach for technical issues like this. They seem to want us to trust that they know what's best from a technical image quality perspective. Generally, I am pleased with their choices. It makes for a very "simple" photographic experience since they don't expect their cameras to compete on every imagined feature. I personally love the (comparatively) simplified menus, buttons, and controls. Let's me focus on the rest of the picture.

 

- Jared

[ramarren]

So I poked around a little bit in the menus... The SL will allow you to set exposure in ⅓ stop increments, which is how you got the 1/10,000s intermediate value. Mine is configured to half stops (to match Leica's lens click stops), so only electronic values are 1/12,500s and 1/16,000s.

 

- Jared

 

 

That makes sense ... I always set the shutter and EV adjustment to third stop increments regardless of what the lenses offer.

 

But the camera recorded both 1/10,000 and 1/12,000, and I can set 1/12,500 as well. So I think in A and P modes the SL can set more differentiated exposure times than the manual settings allow. It makes sense: it's in control of the shutter time in those modes. 

[Jared]

That makes sense ... I always set the shutter and EV adjustment to third stop increments regardless of what the lenses offer.

 

But the camera recorded both 1/10,000 and 1/12,000, and I can set 1/12,500 as well. So I think in A and P modes the SL can set more differentiated exposure times than the manual settings allow. It makes sense: it's in control of the shutter time in those modes. 

 

 

Yes, it seems like a reasonable explanation.  No reason to think the automatic modes couldn't bee free floating with regard to shutter speed.

 

- Jared

[scott kirkpatrick]

Yes, it seems like a reasonable explanation.  No reason to think the automatic modes couldn't bee free floating with regard to shutter speed.

 

- Jared

I suspect even the automatic modes are restricted to a limited set of digital values.  I set f/1.4, ISO 1600 (it's grey here this morning), and only got speeds of 8000, 10000, 12500, and 16000, while spot metering on different parts of a scene.  I have EV increment set to 1/3 and mode is A.  I didn't see 12000 in half a dozen random shots.

 

I'm still hoping someone will shoot a fast moving car in front of a light pole with electronic shutter.

 

scott

[Sye Williams]

Jared looks like a very informative piece that went right over my head HA.  I just purchased the SL after owning the M8 and M9. I was told by someone at Leica that future firmwares will eventually make the camera silent below the shutter speed of 1/10,000. Based on your technical knowledge will that happen, and happen relatively soon? I shoot 1 or 2 movies a year and I'm over using a sound blimp. What are your thoughts? )) Thank you

[Advertisement]

Hallo,

Look here: Leica L-Mount

[ramarren]

Jared looks like a very informative piece that went right over my head HA.  I just purchased the SL after owning the M8 and M9. I was told by someone at Leica that future firmwares will eventually make the camera silent below the shutter speed of 1/10,000. Based on your technical knowledge will that happen, and happen relatively soon? I shoot 1 or 2 movies a year and I'm over using a sound blimp. What are your thoughts? )) Thank you

 

 

Video doesn't use a mechanical shutter at all, so the SL is already 100% silent in motion capture mode.