Imagining an ISO-less, AutoETTR camera

I've spent the first days of 2017 in self-enlightenment, reading some excellent articles & discussions published in DPReview (eg. #1, 2 & 3) about how to capture the best underlying information (in a RAW file) from which to base a final image.

Which got me to thinking (as a layperson !) about how an ISO-less, Auto-Expose-to-the-Right/ETTR camera might work.

I'm imagining the ability to set AutoETTR as being either ON or OFF, via a camera-menu option.

When in the ON status, the camera would operate ONLY at its base/natural ISO (#2) ... with no ability to change ISO and no consideration of ISO when setting exposure (‘cos, as explained in #3, ISO plays no part in the proper definition of exposure).

Perhaps there’d be a setting for a minimum shutter-speed (as ETTRminSS, for camera-shake consideration) and minimum aperture (as ETTRminAp) to be used in the camera’s calculations for AutoETTR ... Perhaps these could be a min/max range(?)

With mode-dial = Shutter Priority, the user selected shutter speed would be fixed and the camera’s ETTR exposure logic would select the Aperture value (limited by ETTRminAp) that pushed the histogram furthest to the right without causing any of the sensor’s pixels to be exposed beyond their capacity.

With mode-dial = Aperture Priority, the user selected aperture would be fixed and the camera’s ETTR exposure logic would select the shutter speed (limited by ETTRminSS) that pushed the histogram furthest to the right without causing any of the sensor’s pixels to be exposed beyond their capacity.

With mode-dial = Program, the camera’s ETTR exposure logic might suggest a combination of shutter speed (limited by ETTRminSS) and Aperture (limited by ETTRminAp) that pushed the histogram furthest to the right without causing any of the sensor’s pixels to be exposed beyond their capacity ... with ability for the user to then apply Program Shift to select another equivalent combination of shutter-speed/aperture (in considerations for depth-of-field, reducing camera-shake, etc).

Perhaps the EVF (where I'm assuming AutoETTR would be implemented in a mirror-less camera) could receive some appropriate brightening (as defined in #3) – just to assist in composition-framing, focusing, etc – without this being passed on to the RAW file.

Perhaps an ETTR Zebra pattern could be selected that would indicate the area(s) of the image causing the sensor’s pixels to be exposed beyond their capacity.

Would any of my ponderings here be technically possible ???

In the Sony world (wherein I live), perhaps this might be implemented as a downloadable PlayMemories camera-app (eg. Sony camera-apps ) ... If so, I’d most certainly buy it for my Sony RX10m2.

#1 - See; Shedding some light on the sources of noise - by Richard Butler
#2 – See; Exposure vs brightening (Notes)
#3 – See; Exposure vs brightening (Exposure defined)

I know you spend a lot of type typing that, so think its a great idea. But...

It all seems like a lot of words to encapsulate exposure bracketing and d-lighting. So turn on exposure bracketing, set ISO to 100, and you are done. Also the highlight and shadow compensation mechanisms do pretty much the same thing, they under expose and lift the levels by N-stops. You can set these to max for highlites and it will be doing pretty much what you want.

I guess that you are suggesting is really a "compensate for bad photographer" mode. Oddly enough, Pentax actually has that. There is a AE Comp mode advanced setting which overrides whatever stupid thing the photographer wants to do and gets a properly exposed shot anyway.

The thing to observe is that brights stars, planets, specular highlights - they are essentially infinitely bright, so no degree of under exposure adequately attenuates the burnt areas. At least so much so that underexposing the whole frame to capture a few pixels on their margins in a HUGE sacrifice to make.

-- Bob
http://bob-o-rama.smugmug.com -- Photos
http://www.vimeo.com/boborama/videos -- Videos
http://blog.trafficshaper.com -- Blog

BobORama said:

I know you spend a lot of type typing that, so think its a great idea. But...

It all seems like a lot of words to encapsulate exposure bracketing and d-lighting. So turn on exposure bracketing, set ISO to 100, and you are done. Also the highlight and shadow compensation mechanisms do pretty much the same thing, they under expose and lift the levels by N-stops. You can set these to max for highlites and it will be doing pretty much what you want.

I guess that you are suggesting is really a "compensate for bad photographer" mode. Oddly enough, Pentax actually has that. There is a AE Comp mode advanced setting which overrides whatever stupid thing the photographer wants to do and gets a properly exposed shot anyway.

The thing to observe is that brights stars, planets, specular highlights - they are essentially infinitely bright, so no degree of under exposure adequately attenuates the burnt areas. At least so much so that underexposing the whole frame to capture a few pixels on their margins in a HUGE sacrifice to make.

-- Bob

Click to expand...

I appreciate your reply, Bob ... esp. given you obviously waded thru my rather long post (sorry 'bout that ... its length seems to have discouraged every one else !)

However, your response suggests you may(?) not be aware of benefits of ISO-invariant sensors, which can be exploited with ETTR practice.

See here: Exposure vs Brightening ... & ISO-invariance

BobORama said:

I know you spend a lot of type typing that, so think its a great idea. But...

It all seems like a lot of words to encapsulate exposure bracketing and d-lighting. So turn on exposure bracketing, set ISO to 100, and you are done. Also the highlight and shadow compensation mechanisms do pretty much the same thing, they under expose and lift the levels by N-stops. You can set these to max for highlites and it will be doing pretty much what you want.

I guess that you are suggesting is really a "compensate for bad photographer" mode. Oddly enough, Pentax actually has that. There is a AE Comp mode advanced setting which overrides whatever stupid thing the photographer wants to do and gets a properly exposed shot anyway.

The thing to observe is that brights stars, planets, specular highlights - they are essentially infinitely bright, so no degree of under exposure adequately attenuates the burnt areas. At least so much so that underexposing the whole frame to capture a few pixels on their margins in a HUGE sacrifice to make.

Click to expand...

Planets are certainly not infinitely bright. Nor is the Moon.

BobORama said:

I know you spend a lot of type typing that, so think its a great idea. But...

It all seems like a lot of words to encapsulate exposure bracketing and d-lighting. So turn on exposure bracketing, set ISO to 100, and you are done. Also the highlight and shadow compensation mechanisms do pretty much the same thing, they under expose and lift the levels by N-stops. You can set these to max for highlites and it will be doing pretty much what you want.

I guess that you are suggesting is really a "compensate for bad photographer" mode. Oddly enough, Pentax actually has that. There is a AE Comp mode advanced setting which overrides whatever stupid thing the photographer wants to do and gets a properly exposed shot anyway.

The thing to observe is that brights stars, planets, specular highlights - they are essentially infinitely bright, so no degree of under exposure adequately attenuates the burnt areas. At least so much so that underexposing the whole frame to capture a few pixels on their margins in a HUGE sacrifice to make.

-- Bob
http://bob-o-rama.smugmug.com -- Photos
http://www.vimeo.com/boborama/videos -- Videos
http://blog.trafficshaper.com -- Blog

Click to expand...

I seem to recall (from ???) that the Light Value [ EV (@at ISO speed=100)] for the sun (as seen from Earth) is around 34 stops and the Light Value [EV(@ISO Speed=100)] of sunlit snow is around 16 stops . So I figure that a specular sparkle of sunlight in the snow is about 18 stops brighter than the surround, not quite 'infinite' but practically or "essentially" so --- as seen from the point of view of either a meter or a sensor.

Every winter I obsess over trying to capture "sparkles" in the snow. Here is a typical example from last winter:


From: https://www.dpreview.com/galleries/0796805747/albums/20160103-snow-texture-sparkles-fz1000

The situation is a bit easier when shooting sparkly snow in the light of a full moon: (This experiment was shot hand-held is not meant to be a "work of art"!! Details of my (wishful) thinking and experiments are here: Technical: Snow sparkles under a full moon. )


From: https://www.dpreview.com/galleries/0796805747/albums/20160125-sparkles-in-moonlight

Although I am not a fan of automating ETTR I am a fan of gollywop and Martinec in their treatments of Exposure and Noise.

I haven't used uniWB in my cameras and won't. Instead I use RawDigger to (sort of) calibrate the puny jpeg histograms in my MFT and FZ1000 cameras.

Instead of AutoETTR I would really like to see LiveView histograms and Zebra Stripings (blinkies) informed by sensor data. (But I am not going to switch to Canon's Magic Lantern to do so.)

This winter I am going to try to get those sparkles (and stars) to be diffracted by the blades of the iris in a stopped down Voigtlander Rokon 10.5 mm prime. Unfortunately it doesn't look like the weather will cooperate over the next few days or so. :-( Maybe February!

--
Tom
The best part of growing old is having the opportunity to do so.

Thanks to HumanTarget, BobORama, D.Cox & GeorgianBay for all your responses ...

I'm reading everything carefully, in attempt to understand this stuff as well as I can, (from my background as a novice in this area - tho, not without a "technical bent" !)

In my opinion, you will get realistic sparkles only by shooting in stereo. What gives the sparkly effect is that the crystals are aligned to reflect the sun into one of your eyes but not both.

When the brain receives conflicting information from the two eyes, you perceive "sparkly" or "silvery".

GeorgianBay1939 said:

Instead of AutoETTR I would really like to see LiveView histograms and Zebra Stripings (blinkies) informed by sensor data. (But I am not going to switch to Canon's Magic Lantern to do so.)

Click to expand...

Never seen it live, but I gather ML histograms are too narrow to be of good use. I'd be glad to see good wide off-line (image review) JPG histograms with raw clipping points marks of the three raw channels, with wide-DR rendering with least highlights headroom.

I think an additional auto meter option that would approximate raw-ETTR with proper adjustment options would be useful often. Nikon did add the HWM, but presumably it is more like JPG-ETTR.

Proper raw-ETTR would take two exposures (at least), and a true raw histogram computation in between, which is rather inconvenient. DSLR exposure sensors don't have enough resolution; Live-View and mirrorless measure exposure via the main sensor, but they surely don't compute the exposure from the whole raw data.

I think we are likely to see progress here soon if wide-DR displays become mainstream.

D Cox said:

In my opinion, you will get realistic sparkles only by shooting in stereo. What gives the sparkly effect is that the crystals are aligned to reflect the sun into one of your eyes but not both.

Click to expand...

Yes.

Yesterday morning I looked at some snow sparkles from alternate eyes and noticed that the number and location of the sparkles depended on which eye was open. So the distance between the eyes is sufficiently wide for each eye to catch the specular reflections off of different sets of snow crystals. The distance between my pupils is a typical 6.5 cm .

If I keep both eyes open my ey/brain system can see both sets of reflections quite easily, especially if I alternately blink a few times to sort out which signal comes from each eye.

If I move my head side to side a few inches with a closed eye the sparkles appear and disappear as the eye moves through the area where the individual glints are reflected.

With both eyes open, side to side movement enhances the sense of sparkle as it is dynamic--- always changing. That effect was enhanced later in the day when I drove through a few upsun snow fields to experiment with movement. Slow movement seems to give a more enhanced sense of sparkling snow.

D Cox said:

When the brain receives conflicting information from the two eyes, you perceive "sparkly" or "silvery".

Click to expand...

That is very interesting. Do you have a reference for this. I would like to learn more about the subject.

Many thanks.

_sem_ said:

GeorgianBay1939 said:

Instead of AutoETTR I would really like to see LiveView histograms and Zebra Stripings (blinkies) informed by sensor data. (But I am not going to switch to Canon's Magic Lantern to do so.)

Click to expand...

Never seen it live, but I gather ML histograms are too narrow to be of good use. I'd be glad to see good wide off-line (image review) JPG histograms with raw clipping points marks of the three raw channels, with wide-DR rendering with least highlights headroom.

Click to expand...

I haven't seen it either.

I agree with your point of wide histograms. It would make sense to able to toggle to a big histogram in the EVF. However toggling to zoomed in zebra striping would also be informative. On my panny cameras with zebra striping, I set the onset of zebra striping at "105%" which gives me about 1/3 to 2/3 stop of headroom. [Where I use headroom to indicate the distance between clipping on the camera and blown highlights in Lightroom 5.6. This is confirmed by some checking using RawDigger.] So with critical scenes I usually bracket exposure ( -2/3, 0, +2/3)

I have mirrorless cameras so I'd really like to see that taken advantage of to give sensor data in the Liveview EVF. That would put very accurate (and precise) Exposure information onto the EVF at the time of shutter release.

_sem_ said:

I think an additional auto meter option that would approximate raw-ETTR with proper adjustment options would be useful often. Nikon did add the HWM, but presumably it is more like JPG-ETTR.

Click to expand...

You are way beyond me with this. I am still learning to get accurate and precise Exposures by ETTR. The critical point (for me) is getting enough good information at the high end to make a good judgement call on Exposure. I have lots of fun shooting sparkles and textures in snow .... a good challenge. Please see: https://brtthome.com/2017/01/03/20170102-more-snow-texture-imagery/

_sem_ said:

Proper raw-ETTR would take two exposures (at least), and a true raw histogram computation in between, which is rather inconvenient. DSLR exposure sensors don't have enough resolution;

Click to expand...

I don't have the background to comment on the above.

_sem_ said:

Live-View and mirrorless measure exposure via the main sensor, but they surely don't compute the exposure from the whole raw data.

Click to expand...

I really don't need the whole histogram of sensor Exposure. But I really do need to know where the clipping is taking place (zebra striping) and by how much (histogram)

_sem_ said:

I think we are likely to see progress here soon if wide-DR displays become mainstream.

Click to expand...

I hope so. But I also hope that the whole thing doesn't become algorithmic which dumb-down the actual processes, thereby reducing camera control by the operator.

I use RawDigger regularly to check my raw exposure.

While all of my Sigma cameras are truly ISO-less, raw ETTR does not guarantee that highlights in the converted image will not be blown (unless an individual Foveon layer is being extracted to create a monochrome image).

The reason being that the matrices needed to get decent color from the sensor layers' raw data are pretty drastic with quite heavy coefficients.

Here's an early Foveon cam-to-XYZ matrix:


RGB = bottom, middle, top sensor layers.

--
"What we've got hyah is Failyah to Communicate": 'Cool Hand Luke' 1967.
Ted

D Cox said:

BobORama said:

The thing to observe is that brights stars, planets, specular highlights - they are essentially infinitely bright, so no degree of under exposure adequately attenuates the burnt areas. At least so much so that underexposing the whole frame to capture a few pixels on their margins in a HUGE sacrifice to make.

Click to expand...

Planets are certainly not infinitely bright. Nor is the Moon.

Click to expand...

If its burnt out, its infinitely bright. Infinity + 1 = Infinity. and a burnt sopt + 1 more photon is still burnt out.

Anything brighter, all the way to a supernova, is the same measured brightness.

Note also qualifier "bright" e.g. Venus, or Vega

The moon, for example, is 8 EV brighter than the next brightest object in the night sky, which is to say, its so bright that the difference is beyond the DR of the camera.

Back to the point, which is when you need to underexpose by 7 stops to avoid burning out some spot, its not a practical solution, at least not with sensors available today or i n the foreseeable future.

-- Bob
http://bob-o-rama.smugmug.com -- Photos
http://www.vimeo.com/boborama/videos -- Videos
http://blog.trafficshaper.com -- Blog

Good job with the sparkles!


-- Bob
http://bob-o-rama.smugmug.com -- Photos
http://www.vimeo.com/boborama/videos -- Videos
http://blog.trafficshaper.com -- Blog

GeorgianBay1939 said:

Instead of AutoETTR I would really like to see LiveView histograms and Zebra Stripings (blinkies) informed by sensor data. (But I am not going to switch to Canon's Magic Lantern to do so.)

Click to expand...

The Pentax offers this, so I guess I am spoiled. As mentioned, I tend to use bracketing rather than torture myself trying to do something in a single shot.

GeorgianBay1939 said:

This winter I am going to try to get those sparkles (and stars) to be diffracted by the blades of the iris in a stopped down Voigtlander Rokon 10.5 mm prime. Unfortunately it doesn't look like the weather will cooperate over the next few days or so. :-( Maybe February!

Click to expand...

Well, it was 60 degrees last week, 10 degrees this week, so who knows. I have been wanting to do more snowflake macros, but no snowflakes.


-- Bob
http://bob-o-rama.smugmug.com -- Photos
http://www.vimeo.com/boborama/videos -- Videos
http://blog.trafficshaper.com -- Blog

JohnTheKeenAmateur said:

However, your response suggests you may(?) not be aware of benefits of ISO-invariant sensors, which can be exploited with ETTR practice.

See here: Exposure vs Brightening ... & ISO-invariance

Click to expand...

I bristle whenever something people have been doing for a century is given a new name and blogged about endlessly.

-- Bob
http://bob-o-rama.smugmug.com -- Photos
http://www.vimeo.com/boborama/videos -- Videos
http://blog.trafficshaper.com -- Blog

BobORama said:

D Cox said:

BobORama said:

The thing to observe is that brights stars, planets, specular highlights - they are essentially infinitely bright, so no degree of under exposure adequately attenuates the burnt areas. At least so much so that underexposing the whole frame to capture a few pixels on their margins in a HUGE sacrifice to make.

Click to expand...

Planets are certainly not infinitely bright. Nor is the Moon.

Click to expand...

If its burnt out, its infinitely bright. Infinity + 1 = Infinity. and a burnt sopt + 1 more photon is still burnt out.

Click to expand...

If a planet or the moon is burned out in your image file, then you are over-exposing.

BobORama said:

Anything brighter, all the way to a supernova, is the same measured brightness.

Note also qualifier "bright" e.g. Venus, or Vega

The moon, for example, is 8 EV brighter than the next brightest object in the night sky, which is to say, its so bright that the difference is beyond the DR of the camera.

Back to the point, which is when you need to underexpose by 7 stops to avoid burning out some spot, its not a practical solution, at least not with sensors available today or i n the foreseeable future.

-- Bob
http://bob-o-rama.smugmug.com -- Photos
http://www.vimeo.com/boborama/videos -- Videos
http://blog.trafficshaper.com -- Blog

Click to expand...

BobORama said:

GeorgianBay1939 said:

Instead of AutoETTR I would really like to see LiveView histograms and Zebra Stripings (blinkies) informed by sensor data. (But I am not going to switch to Canon's Magic Lantern to do so.)

Click to expand...

The Pentax offers this, so I guess I am spoiled. As mentioned, I tend to use bracketing rather than torture myself trying to do something in a single shot.

Click to expand...

Yeah, same here. I usually shoot -2/3, 0, +2/3 EV in critical situations, like shooting snow textures.

BobORama said:

BobORama said:

This winter I am going to try to get those sparkles (and stars) to be diffracted by the blades of the iris in a stopped down Voigtlander Rokon 10.5 mm prime. Unfortunately it doesn't look like the weather will cooperate over the next few days or so. :-( Maybe February!

Click to expand...

Well, it was 60 degrees last week, 10 degrees this week, so who knows. I have been wanting to do more snowflake macros, but no snowflakes.

Click to expand...

It was -20ºC last night with a partial moon, but unfortunately high cloud obscured enough moonlight to minimize any sparkles even to the moving naked eye. I detected a bit of sparkle in the result but not enough to make it interesting so I didn't set up a tripod.

Here is a strongly brightened (LR 5.6) hand-held shot (Voigtlander Nokton 10.5 mm f/0.95) from last night about an hour before the snow started. A few stars are visible but the moon is too heavily obscured to give any visible specular reflections ...





So I'll be looking (in vain, according to the forecast) for a clearer night over the next few days or next month.

BobORama said:

-- Bob
http://bob-o-rama.smugmug.com -- Photos
http://www.vimeo.com/boborama/videos -- Videos
http://blog.trafficshaper.com -- Blog

Click to expand...

By the way, I very much enjoyed your smugmug imagery.... very creative stuff.

And I enjoyed your blogs about Dead Text, Finland #2, and "Whoever dealt it..."

--
Tom
The best part of growing old is having the opportunity to do so.

BobORama said:

JohnTheKeenAmateur said:

However, your response suggests you may(?) not be aware of benefits of ISO-invariant sensors, which can be exploited with ETTR practice.

See here: Exposure vs Brightening ... & ISO-invariance

Click to expand...

I bristle whenever something people have been doing for a century is given a new name and blogged about endlessly.

Click to expand...

It's never been like digital RAW with film. Before digital, there is no medium that allows you to increase exposure for higher S&N at all tones involved, without sacrifice to increased grain contrast relative to signal contrast. With digital, it can just get better, better, better, then "clip". With film, contrast will deteriorate and grain will dominate. Film needs to be centered for overall optimum results.

John Sheehy said:

BobORama said:

JohnTheKeenAmateur said:

However, your response suggests you may(?) not be aware of benefits of ISO-invariant sensors, which can be exploited with ETTR practice.

See here: Exposure vs Brightening ... & ISO-invariance

Click to expand...

I bristle whenever something people have been doing for a century is given a new name and blogged about endlessly.

Click to expand...

It's never been like digital RAW with film. Before digital, there is no medium that allows you to increase exposure for higher S&N at all tones involved, without sacrifice to increased grain contrast relative to signal contrast. With digital, it can just get better, better, better, then "clip". With film, contrast will deteriorate and grain will dominate. Film needs to be centered for overall optimum results.

Click to expand...

Not specifically directed at you, but in general...

That's like saying all mammals are elephants, and by nomenclature declaring a cat is a "small elephant with fur, no trunk, and pointy ears and that purrs."

Your assertion about centering exposure for film as being ideal is not true. The non-linear response of film allows you to under and over expose and manipulate in post, some times by large margins. And to put more fidelity underneath one or the other end of the histogram. Playing to the shadows or highlights is just one of the usual tricks we used.

The thing many modern photographers fail to understand, with very few exceptions, film is a digital medium. let that marinade.

A "simple" challenge to the ETTRvangellicals:

Prove it works, not just for specific or contrived use cases, but in a systematic and algorithmic way. That would require an algorithmic means to measure image "goodness" and how a specific ETTR strength modifies that "goodness" that so that the amount of ETTR to apply is an optimizable variable.

Then you need to look at the results and see if it is any better than the scene aware AE program modes available, e.g. Green Mode.

-- Bob

http://bob-o-rama.smugmug.com -- Photos
http://www.vimeo.com/boborama/videos -- Videos
http://blog.trafficshaper.com -- Blog

GeorgianBay1939 said:

...Yeah, same here. I usually shoot -2/3, 0, +2/3 EV in critical situations, like shooting snow textures.

Click to expand...

...

When I expose in 'iSO-less mode', I always do this by auto-bracketing aperture at base ISO. I use the exposure that overexposes only the unimportant highlight regions. I switch between 1/3 and 2/3 stop steps depending on the scene. I estimate at least 75% of the time I use 1/3 steps .