Why change ISO at all?

[TrojMacReady]

ISO 160 in the dark. RAW, Boosted (to the limits of the sliders, curves) in Lightroom 4.3. Arms propped.

If I had cranked up ISO or let it float, AND obeyed the meter recommendation I am sure that I would have ended up with a lower exposure (f/ and/or ss) and more noise.

Part of the denial of ISO invariance is continued reliance on the traditional exposure triangle (f/, ss, ISO) instead of realizing that exposure depends on these three factors : f/ (T actually), shutter interval and scene luminance. Experienced ( and even inexperienced ones like me) RAW shooter have to know this to get maximum data (usually light) onto their sensor without blowing/clipping highlights.

Exposure does not depend on gain, ISO.

But when shooting automatically (P, A, S modes) the camera's metering can be affected by changes in ISO and the operator might change exposure as a result. It is important that beginners NOT confuse ISO as directly affecting exposure. Because, it doesn't .....

in spite of various authors' like Bryan Peterson et al, and some photographers' (with JPEG and film backgrounds) insistence that it does.

And this is based on, what????? Then we shouldn’t be surprised to find out there’s a technical letter-writing campaign to photography authors all around the globe, as-well-as to camera manufactures to remove ISO from their publications regarding exposure, and to remove the ISO feature from future camera bodies? :-O

It is important that beginners realize the shooting RAW is a very different proposition than shooting JPEGs and that it is important to understand the correct, traditional, scientific meaning of exposure.

Whether shooting RAW or JPEG, my exposure procedure is always the same, not different, and I continue to use all three variables available to me (ISO, Shutter and Aperture) to set the exposure correctly in camera. I don’t rely on the use of software after-the-fact to make huge adjustments in my shots. And just because I shoot RAW vs JPEG doesn’t mean that I’ll ignore using a crucial variable that's available to me, in camera.

The issue I see with your example is that it’s of a static object, and without any regards whatsoever for DOF. If DOF was of a technical concern/creative issue to your shot (eg f/9), your shutter speed would have been dramatically affected.

So to follow your approach, in my example of a static shot (1/6 sec, f/6.3, ISO 1600), you're saying that I would have been just as well to have taken it at ISO 100, and then compensated the desired exposure adjustments after-the-fact in post processing? Then at the time of this shot, I would have had to make some huge adjustments in my shutter speed and/or aperture to balance the exposure. Bear in mind this shot was handheld at 1/6 sec, and was about as slow as I personally could go without introducing camera shake, and the aperture setting of f/6.3 gave me the DOF I wanted.

Tripods were not permitted inside the caverns, and I wanted a particular DOF, and knowing that I (personally) could only handhold at 1/6 s, the ISO of 1600 was necessary.



Unfortunately, your work does not work for me.

Regards, Mike

I think you misunderstood. DOF has little to do with this, likewise for the shooting parameters you mention above. Because you wouldn't have to adjust them. You retain the same shutterspeed and F stop, hence why on the LCD or on opening the RAW file, the picture will come out underexposed when sticking to a lower ISO. Which is exactly what you correct in post processing.

Here's an example. Full scene:



100% crop from "properly exposed" (according to camera histogram) ISO 6400 RAW file:



100% crop from ISO 200 shot at the same physical exposure (F stop and shutterspeed) settings, corrected (brightened) in postprocessing:



Noise is pretty similar altough I actually prefer the noise structure in shadows from the base ISO shot. But most apparent is the difference in highlight headroom, see the rolloff.

 

[Port Royal Dad]

ISO 160 in the dark. RAW, Boosted (to the limits of the sliders, curves) in Lightroom 4.3. Arms propped.

If I had cranked up ISO or let it float, AND obeyed the meter recommendation I am sure that I would have ended up with a lower exposure (f/ and/or ss) and more noise.

Part of the denial of ISO invariance is continued reliance on the traditional exposure triangle (f/, ss, ISO) instead of realizing that exposure depends on these three factors : f/ (T actually), shutter interval and scene luminance. Experienced ( and even inexperienced ones like me) RAW shooter have to know this to get maximum data (usually light) onto their sensor without blowing/clipping highlights.

Exposure does not depend on gain, ISO.

But when shooting automatically (P, A, S modes) the camera's metering can be affected by changes in ISO and the operator might change exposure as a result. It is important that beginners NOT confuse ISO as directly affecting exposure. Because, it doesn't .....

in spite of various authors' like Bryan Peterson et al, and some photographers' (with JPEG and film backgrounds) insistence that it does.

And this is based on, what????? Then we shouldn’t be surprised to find out there’s a technical letter-writing campaign to photography authors all around the globe, as-well-as to camera manufactures to remove ISO from their publications regarding exposure, and to remove the ISO feature from future camera bodies? :-O

It is important that beginners realize the shooting RAW is a very different proposition than shooting JPEGs and that it is important to understand the correct, traditional, scientific meaning of exposure.

Whether shooting RAW or JPEG, my exposure procedure is always the same, not different, and I continue to use all three variables available to me (ISO, Shutter and Aperture) to set the exposure correctly in camera. I don’t rely on the use of software after-the-fact to make huge adjustments in my shots. And just because I shoot RAW vs JPEG doesn’t mean that I’ll ignore using a crucial variable that's available to me, in camera.

The issue I see with your example is that it’s of a static object, and without any regards whatsoever for DOF. If DOF was of a technical concern/creative issue to your shot (eg f/9), your shutter speed would have been dramatically affected.

So to follow your approach, in my example of a static shot (1/6 sec, f/6.3, ISO 1600), you're saying that I would have been just as well to have taken it at ISO 100, and then compensated the desired exposure adjustments after-the-fact in post processing? Then at the time of this shot, I would have had to make some huge adjustments in my shutter speed and/or aperture to balance the exposure. Bear in mind this shot was handheld at 1/6 sec, and was about as slow as I personally could go without introducing camera shake, and the aperture setting of f/6.3 gave me the DOF I wanted.

Tripods were not permitted inside the caverns, and I wanted a particular DOF, and knowing that I (personally) could only handhold at 1/6 s, the ISO of 1600 was necessary.



Unfortunately, your work does not work for me.

Regards, Mike

I think you misunderstood. DOF has little to do with this, likewise for the shooting parameters you mention above. Because you wouldn't have to adjust them. You retain the same shutterspeed and F stop, hence why on the LCD or on opening the RAW file, the picture will come out underexposed when sticking to a lower ISO. Which is exactly what you correct in post processing.

But when I'm preping for any shot, I'm using the tools available to me in camera to establish the exposure. I'll use software in PP to tweak things like WB, sharpness, and composition. I do not employ a workflow whereby I shoot a scene, any scene regardless of the situation and always at a low ISO (eg 100) and then adjust the final exposure in PP. Obviously, I'll strive for the lowest ISO, but I don't rely on software as the go-to-tool.

Best Regards, Mike





--
B.R.A.S.S. (Breathe, Relax, Aim, Sight, Squeeze)

http://lh5.google.com/PortRoyalDad/Rwu_WzPG87I/AAAAAAAAAnU/fC0kmv_nhY0/s144/IMG_1479.JPG

A Link To - WilbaW's Unofficial Rebel Forum FAQ- http://snipurl.com/RebelFAQ

 

[TrojMacReady]

ISO 160 in the dark. RAW, Boosted (to the limits of the sliders, curves) in Lightroom 4.3. Arms propped.

If I had cranked up ISO or let it float, AND obeyed the meter recommendation I am sure that I would have ended up with a lower exposure (f/ and/or ss) and more noise.

Part of the denial of ISO invariance is continued reliance on the traditional exposure triangle (f/, ss, ISO) instead of realizing that exposure depends on these three factors : f/ (T actually), shutter interval and scene luminance. Experienced ( and even inexperienced ones like me) RAW shooter have to know this to get maximum data (usually light) onto their sensor without blowing/clipping highlights.

Exposure does not depend on gain, ISO.

But when shooting automatically (P, A, S modes) the camera's metering can be affected by changes in ISO and the operator might change exposure as a result. It is important that beginners NOT confuse ISO as directly affecting exposure. Because, it doesn't .....

in spite of various authors' like Bryan Peterson et al, and some photographers' (with JPEG and film backgrounds) insistence that it does.

And this is based on, what????? Then we shouldn’t be surprised to find out there’s a technical letter-writing campaign to photography authors all around the globe, as-well-as to camera manufactures to remove ISO from their publications regarding exposure, and to remove the ISO feature from future camera bodies? :-O

It is important that beginners realize the shooting RAW is a very different proposition than shooting JPEGs and that it is important to understand the correct, traditional, scientific meaning of exposure.

Whether shooting RAW or JPEG, my exposure procedure is always the same, not different, and I continue to use all three variables available to me (ISO, Shutter and Aperture) to set the exposure correctly in camera. I don’t rely on the use of software after-the-fact to make huge adjustments in my shots. And just because I shoot RAW vs JPEG doesn’t mean that I’ll ignore using a crucial variable that's available to me, in camera.

The issue I see with your example is that it’s of a static object, and without any regards whatsoever for DOF. If DOF was of a technical concern/creative issue to your shot (eg f/9), your shutter speed would have been dramatically affected.

So to follow your approach, in my example of a static shot (1/6 sec, f/6.3, ISO 1600), you're saying that I would have been just as well to have taken it at ISO 100, and then compensated the desired exposure adjustments after-the-fact in post processing? Then at the time of this shot, I would have had to make some huge adjustments in my shutter speed and/or aperture to balance the exposure. Bear in mind this shot was handheld at 1/6 sec, and was about as slow as I personally could go without introducing camera shake, and the aperture setting of f/6.3 gave me the DOF I wanted.

Tripods were not permitted inside the caverns, and I wanted a particular DOF, and knowing that I (personally) could only handhold at 1/6 s, the ISO of 1600 was necessary.



Unfortunately, your work does not work for me.

Regards, Mike

I think you misunderstood. DOF has little to do with this, likewise for the shooting parameters you mention above. Because you wouldn't have to adjust them. You retain the same shutterspeed and F stop, hence why on the LCD or on opening the RAW file, the picture will come out underexposed when sticking to a lower ISO. Which is exactly what you correct in post processing.

But when I'm preping for any shot, I'm using the tools available to me in camera to establish the exposure. I'll use software in PP to tweak things like WB, sharpness, and composition. I do not employ a workflow whereby I shoot a scene, any scene regardless of the situation and always at a low ISO (eg 100) and then adjust the final exposure in PP. Obviously, I'll strive for the lowest ISO, but I don't rely on software as the go-to-tool.

Best Regards, Mike

That's a choice and there's nothing wrong with that or having the choice to begin with.

Just realize that sometimes the cameras do in camera (or to a similar extent) what your computer software does or can do afterwards. And in some cases, the software on your computer has more data to work with (highlights for example).

 

[GeorgianBay1939]

ISO 160 in the dark. RAW, Boosted (to the limits of the sliders, curves) in Lightroom 4.3. Arms propped.

If I had cranked up ISO or let it float, AND obeyed the meter recommendation I am sure that I would have ended up with a lower exposure (f/ and/or ss) and more noise.

Part of the denial of ISO invariance is continued reliance on the traditional exposure triangle (f/, ss, ISO) instead of realizing that exposure depends on these three factors : f/ (T actually), shutter interval and scene luminance. Experienced ( and even inexperienced ones like me) RAW shooter have to know this to get maximum data (usually light) onto their sensor without blowing/clipping highlights.

Exposure does not depend on gain, ISO.

But when shooting automatically (P, A, S modes) the camera's metering can be affected by changes in ISO and the operator might change exposure as a result. It is important that beginners NOT confuse ISO as directly affecting exposure. Because, it doesn't .....

in spite of various authors' like Bryan Peterson et al, and some photographers' (with JPEG and film backgrounds) insistence that it does.

And this is based on, what?????

See above Exposure

In photography, exposure is the quantity of light reaching a photographic film, as determined by shutter speed and lens aperture. In digital photography "film" is substituted with "sensor". Exposure is measured in lux seconds, and can be computed from exposure value (EV) and scene luminance in a specified region.

Then we shouldn’t be surprised to find out there’s a technical letter-writing campaign to photography authors all around the globe, as-well-as to camera manufactures to remove ISO from their publications regarding exposure, and to remove the ISO feature from future camera bodies? :-O

Doesn't follow. I change my ISO all the time when shutter interval and f/ are critical to stop motion blur and to get thick DOF.

It is important that beginners realize the shooting RAW is a very different proposition than shooting JPEGs and that it is important to understand the correct, traditional, scientific meaning of exposure.

Whether shooting RAW or JPEG, my exposure procedure is always the same, not different, and I continue to use all three variables available to me (ISO, Shutter and Aperture) to set the exposure correctly in camera. I don’t rely on the use of software after-the-fact to make huge adjustments in my shots. And just because I shoot RAW vs JPEG doesn’t mean that I’ll ignore using a crucial variable that's available to me, in camera.

I guess the issue is what is

to set the exposure correctly in camera

Some folks, usually JPEG shooters, follow the guidance of the metering system to get a "correct" exposure.

Other folks, usually RAW shooters, follow the guidance of the metering system, the Liveview histogram, with a generous amount of chimping of post exposure RGBY channel histograms to load the sensor optimally ... ie to get as much data as possible in the RAW data file.

The issue I see with your example is that it’s of a static object, and without any regards whatsoever for DOF. If DOF was of a technical concern/creative issue to your shot (eg f/9), your shutter speed would have been dramatically affected.

Agree.

So to follow your approach, in my example of a static shot (1/6 sec, f/6.3, ISO 1600), you're saying that I would have been just as well to have taken it at ISO 100, and then compensated the desired exposure adjustments after-the-fact in post processing?

Possibly but nor necessarily ...see below.

Then at the time of this shot, I would have had to make some pretty serious adjustments in my shutter speed and/or aperture to balance the exposure. Bear in mind this shot was handheld at 1/6 sec, and was about as slow as I personally could go without introducing camera shake, and the aperture setting of f/6.3 gave me the DOF I wanted.

If you need to stop down for DOF or to shorten shutter interval you have no choice but to increase ISO if you are following your meter. If you are in one of the auto modes you can use EC, if in M(anual) you can set the f/ and ss to meet your (DOF, Blur) requirements. In either case you can make the decision to crank up ISO, IF you know the degree of ISO (in)variance of your sensor.

As an earlier poster indicated, a significant value of increasing ISO is just to get a viewable image in the EVF. OVF shooters would not be affected by this.

I just checked your "arsenal" :-D and see that you do not use EVF equipped cameras. I am unfamiliar with what the D300 is capable of in post exposure histograms so the discussion above may be moot.

Tripods were not permitted inside the caverns, and I wanted a particular DOF, and knowing that I (personally) could only handhold at 1/6 s, the ISO of 1600 was necessary.

Judging by the slight clipping or blowing of highlights above it looks to me like you tried to ETTR (assuming it was shot in RAW). Without checking the RAW histogram in SW like RawDigger etc I cannot tell if the highlights were (a) recoverable (compressed), (b) blown (oversaturated sensels) or (c) clipped (too much gain (ISO) in the ADU causing overfilling of the data slots in the RAW file.

Probably b) or c) since a) would be recoverable in your RAW processor.

So with my technique I might've shot the above at the same exposure and dropped the ISO down a stop or two. ( Let me be quick to admit that the difference in format would lead to possible noise and DOF issues but lets not get into equivalence issues!)

Unfortunately, your workflow does not work for me. I always have three variables at my disposal, regardless if I'm shooting RAW or JPEG. The last time I looked anyway, the ISO feature on my camera does not shut off when using RAW or JPEG.

LOL. I also have ISO on my dials. Very handy. But I use it a bit differently than I used to when shooting JPEGs.

Regards, Mike

--
B.R.A.S.S. (Breathe, Relax, Aim, Sight, Squeeze)
http://lh5.google.com/PortRoyalDad/Rwu_WzPG87I/AAAAAAAAAnU/fC0kmv_nhY0/s144/IMG_1479.JPG
A Link To - WilbaW's Unofficial Rebel Forum FAQ- http://snipurl.com/RebelFAQ

I just went to your above link http://www.dpreview.com/articles/1905744226/rebel-forum-faq written a couple of years ago.

And read the Glossary

(There's that Gollywop again!)

and Metering and Exposure

"ETTR means setting your metering values so that the histogram curve is to the right hand side of the graph rather than the left, without important highlights "going over the edge" and losing detail. The idea is to maximise the signal-to-noise ratio in the data you capture, but only when the important highlights are safe.

There are two forms of ETTR. "Naive" or "ISO-first" ETTR happens when you assume that ISO is the most stable factor, and therefore compensate the aperture and/or shutter duration to move the histogram curve. (Unfortunately, the user interfaces of Canon DSLRs are designed in keeping with that assumption.) This often results in the capture of less light, which generally means more noise.

"Smart" or "ISO-last" ETTR happens when you –

• set your aperture to the widest that will give you the depth of field you need,
• set your shutter duration to the longest that avoids camera and subject motion blur,
• then set the ISO to put the histogram curve safely TTR.

Doing that means you capture as much light as possible, then optimally process the signal from the image sensor.

Opponents of smart ETTR often claim "there is no noise penalty" to shooting at a lower ISO than one that gives a highlight-preserving curve TTR (e.g. using 1600 instead of the "correct" 6400 and boosting image brightness in post-processing by two steps), but with contemporary Canon DSLRs there is always an image quality penalty in doing that, although it may be unnoticeable in comparisons between carefully processed images viewed at significantly less than 100%."

Since I have no first hand experience with Canon gear I cannot quarrel with the author, (or you for that matter)!

BUT (always a "BUT") ... when you fool with the ISO dial you are fooling with gain, brightening up or dimming down the camera's JPEG ... and moving the (live view or post exposure) histograms to the right or to the left. If your f/ and ss stay the same you are NOT changing exposure when you fool with the ISO dial. You are just "brightening".

BUT (Bigger "BUT") ... What works, works!

Good link, provides good overview with good fundamentals, IMHO. Thanks!

Best!

Tom

 

[GeorgianBay1939]

SNIP

I wish that I'd read your post before writing my lengthy post here!

http://www.dpreview.com/forums/post/52462329

I feel that I am still very much a beginner here and am still trying to sort out the noise from the signal in some of these more controversial topics.

I found Mike's link to WilbaW's Rebel Forum FAQ to be very helpful in this sorting out process. I haven't read it critically yet but it seems helpful.

I think you misunderstood. DOF has little to do with this, likewise for the shooting parameters you mention above. Because you wouldn't have to adjust them. You retain the same shutterspeed and F stop, hence why on the LCD or on opening the RAW file, the picture will come out underexposed when sticking to a lower ISO. Which is exactly what you correct in post processing.

Here's an example. Full scene:



100% crop from "properly exposed" (according to camera histogram) ISO 6400 RAW file:



100% crop from ISO 200 shot at the same physical exposure (F stop and shutterspeed) settings, corrected (brightened) in postprocessing:



Noise is pretty similar altough I actually prefer the noise structure in shadows from the base ISO shot. But most apparent is the difference in highlight headroom, see the rolloff.

"Rolloff" I need to explore that. The brick wall on the left in the cropped image is a good example.

When I look at an image either (ooc JPEG or processed RAW) and see those bright, lost detail, spots I wonder what causes them and come up with three possibilities:

1. Very high "compression" of the tone curve at the high end. ie very steep gradient on the right side of the histogram OR
2. sensels (in an area) are oversaturated, with a loss of spacial tonal information, as a cluster of sensels are oversaturated ... blown OR
3. ISO is so high that the signal from non-saturated sensels gets lost (in the ADU?) when digitized. Ie the (12 bit or 14 bit usually) data field for that sensel "overflows" due to too much gain. When these form an area there is a loss of spatial tonal information.... clipping.

Case 1 can be saved by some form of highlight recovery, ie stretching the histogram in the highlights.

Case 2 and Case 3 are unrecoverable.

Case 2 is caused by overexposure of the sensor ... to much light on those sensels during the interval that the shutter is open ... caused by too much luminance, too long a shutter interval, too open an aperture. I think that "blown sensels/pixels" is the jargon used to describe this.

Case 3 is caused by too much gain, too high an ISO, and the term "clipping" is used to describe this.

IS the above correct?

If so it seems that a significant amount of the compression/blowing/clipping in the ISO 6400 image is caused by an ISO that is too high. ie when the ISO is dropped by FIVE STOPS to ISO 200 the clipping is removed leaving overexposure (and perhaps compression) as the culprit(s).

Right?

Next question:

You say,

100% crop from "properly exposed" (according to camera histogram) ISO 6400 RAW file:

If the camera were set at ISO 6400 and the photographer looked through the EVF (live view) he/she would see a histogram made from a BRIGHTENED JPEG due to his high ISO. In other words the 6400 histogram would be a long ways (5 stops) to the right of the 200 histogram.

I assume that the lens is wide open (at f/2.8) and that you can't expose any longer than 1/2 sec due to motion blur. So that establishes your physical exposure. You move your ISO up until your histogram fills the box and changes colour to indicate a "proper exposure". Really when you dial that ISO up like that aren't you just brightening the in-camera JPEG to give a "proper brightness" in the EVF and in the histogram? When you release the shutter and the JPEG is written to the card, is it the same JPEG as the one you saw in the EVF --- regarding brightness at least?

Am I on the right track above?

Can you give me some links to reliable information re compression, oversaturation, clipping?

Many thanks,

Tom

 

[TrojMacReady]

SNIP

I wish that I'd read your post before writing my lengthy post here!

http://www.dpreview.com/forums/post/52462329

I feel that I am still very much a beginner here and am still trying to sort out the noise from the signal in some of these more controversial topics.

I found Mike's link to WilbaW's Rebel Forum FAQ to be very helpful in this sorting out process. I haven't read it critically yet but it seems helpful.

I think you misunderstood. DOF has little to do with this, likewise for the shooting parameters you mention above. Because you wouldn't have to adjust them. You retain the same shutterspeed and F stop, hence why on the LCD or on opening the RAW file, the picture will come out underexposed when sticking to a lower ISO. Which is exactly what you correct in post processing.

Here's an example. Full scene:



100% crop from "properly exposed" (according to camera histogram) ISO 6400 RAW file:



100% crop from ISO 200 shot at the same physical exposure (F stop and shutterspeed) settings, corrected (brightened) in postprocessing:



Noise is pretty similar altough I actually prefer the noise structure in shadows from the base ISO shot. But most apparent is the difference in highlight headroom, see the rolloff.

"Rolloff" I need to explore that. The brick wall on the left in the cropped image is a good example.

When I look at an image either (ooc JPEG or processed RAW) and see those bright, lost detail, spots I wonder what causes them and come up with three possibilities:

1. Very high "compression" of the tone curve at the high end. ie very steep gradient on the right side of the histogram OR
2. sensels (in an area) are oversaturated, with a loss of spacial tonal information, as a cluster of sensels are oversaturated ... blown OR
3. ISO is so high that the signal from non-saturated sensels gets lost (in the ADU?) when digitized. Ie the (12 bit or 14 bit usually) data field for that sensel "overflows" due to too much gain. When these form an area there is a loss of spatial tonal information.... clipping.

Case 1 can be saved by some form of highlight recovery, ie stretching the histogram in the highlights.

Case 2 and Case 3 are unrecoverable.

Case 2 is caused by overexposure of the sensor ... to much light on those sensels during the interval that the shutter is open ... caused by too much luminance, too long a shutter interval, too open an aperture. I think that "blown sensels/pixels" is the jargon used to describe this.

Case 3 is caused by too much gain, too high an ISO, and the term "clipping" is used to describe this.

IS the above correct?

I think it's indeed a combination of the first 2 points in the ISO 200 shot, since the light sources were obviously much brighter than the surrounding areas, a problem seen in many "nightscapes". Basically, even using all the DR of the sensor available by sticking to base ISO and exposing largely for the highlights (didn't want to underexpose further as motion blur would become an issue and shadows and noise in general would become more compromised too, thus some highlight information in the original scene fell off the right side of the histogram still), I couldn't capture the full DR of the original scene with a single exposure. So some pixels were oversaturated, mostly those where the lightbulbs were. But what I did catch, I compressed to fit into a single jpeg, including the use of the recovery tools in Lightroom.

For the ISO 6400 file on the other hand, the camera uses more analogue gain (until about ISO 1600 IIRC), point 3 enters the equation too. Plus the last two stops, which are done digitally in camera (like most APS-C/FF cameras carrying Sony sensors, Pentax and Nikon included), where it simply multiplies values digitally (representing the extra 2 stops from ISO 1600 to ISO 6400) before writing the RAW file. This means that still some of the highlight information gets cut off before writing the RAW file. Some of that is discussed here too.

If so it seems that a significant amount of the compression/blowing/clipping in the ISO 6400 image is caused by an ISO that is too high. ie when the ISO is dropped by FIVE STOPS to ISO 200 the clipping is removed leaving overexposure (and perhaps compression) as the culprit(s).

Right?

Agreed.

Next question:

You say,

100% crop from "properly exposed" (according to camera histogram) ISO 6400 RAW file:

If the camera were set at ISO 6400 and the photographer looked through the EVF (live view) he/she would see a histogram made from a BRIGHTENED JPEG due to his high ISO. In other words the 6400 histogram would be a long ways (5 stops) to the right of the 200 histogram.

Correct. Basically for the ISO 6400 shot, the EVF will show you an image that roughly reflects the ISO 6400 shot posted, where as for the ISO 200 shot it will mostly be... dark. That being said, many EVF cameras now offer the option to turn live exposure visualizations off, meaning that the EVF will no longer approximate the brightness of the resulting (in camera) jpeg, but rather average the brightness of the scene in the EVF, regardless of physical exposure settings. Which is really helpful for people who underexpose in RAW for example.

I assume that the lens is wide open (at f/2.8) and that you can't expose any longer than 1/2 sec due to motion blur. So that establishes your physical exposure.

Not wide open (which is f/1.4) but a compromise between depth of field, sharpness, motion blur and (shadow) noise.

You move your ISO up until your histogram fills the box and changes colour to indicate a "proper exposure". Really when you dial that ISO up like that aren't you just brightening the in-camera JPEG to give a "proper brightness" in the EVF and in the histogram? When you release the shutter and the JPEG is written to the card, is it the same JPEG as the one you saw in the EVF --- regarding brightness at least?

If the EVF is set to represent the jpeg brightness (and histogram), yes. But see comment above about the option to turn that off.

Am I on the right track above?

Pretty much spot on.

Can you give me some links to reliable information re compression, oversaturation, clipping?

Many thanks,

Tom

I'm not on my own computer right now, but will get back to this topic later this week and post some links.

Regards,

Jort

 

[TTMartin]

I think you misunderstood. DOF has little to do with this, likewise for the shooting parameters you mention above. Because you wouldn't have to adjust them. You retain the same shutterspeed and F stop, hence why on the LCD or on opening the RAW file, the picture will come out underexposed when sticking to a lower ISO. Which is exactly what you correct in post processing.

Here's an example. Full scene:



100% crop from "properly exposed" (according to camera histogram) ISO 6400 RAW file:



100% crop from ISO 200 shot at the same physical exposure (F stop and shutterspeed) settings, corrected (brightened) in postprocessing:



Noise is pretty similar altough I actually prefer the noise structure in shadows from the base ISO shot. But most apparent is the difference in highlight headroom, see the rolloff.

Please post the full scene ISO 200 photo. Showing 100% crops of a relatively well lit area, doesn't really prove anything.

How much detail in the trees above the buildings was retained in the ISO 200 shot?

 

[GeorgianBay1939]

SNIP

I wish that I'd read your post before writing my lengthy post here!

http://www.dpreview.com/forums/post/52462329

I feel that I am still very much a beginner here and am still trying to sort out the noise from the signal in some of these more controversial topics.

I found Mike's link to WilbaW's Rebel Forum FAQ to be very helpful in this sorting out process. I haven't read it critically yet but it seems helpful.

I think you misunderstood. DOF has little to do with this, likewise for the shooting parameters you mention above. Because you wouldn't have to adjust them. You retain the same shutterspeed and F stop, hence why on the LCD or on opening the RAW file, the picture will come out underexposed when sticking to a lower ISO. Which is exactly what you correct in post processing.

Here's an example. Full scene:



100% crop from "properly exposed" (according to camera histogram) ISO 6400 RAW file:



100% crop from ISO 200 shot at the same physical exposure (F stop and shutterspeed) settings, corrected (brightened) in postprocessing:



Noise is pretty similar altough I actually prefer the noise structure in shadows from the base ISO shot. But most apparent is the difference in highlight headroom, see the rolloff.

"Rolloff" I need to explore that. The brick wall on the left in the cropped image is a good example.

When I look at an image either (ooc JPEG or processed RAW) and see those bright, lost detail, spots I wonder what causes them and come up with three possibilities:

1. Very high "compression" of the tone curve at the high end. ie very steep gradient on the right side of the histogram OR
2. sensels (in an area) are oversaturated, with a loss of spacial tonal information, as a cluster of sensels are oversaturated ... blown OR
3. ISO is so high that the signal from non-saturated sensels gets lost (in the ADU?) when digitized. Ie the (12 bit or 14 bit usually) data field for that sensel "overflows" due to too much gain. When these form an area there is a loss of spatial tonal information.... clipping.

Case 1 can be saved by some form of highlight recovery, ie stretching the histogram in the highlights.

Case 2 and Case 3 are unrecoverable.

Case 2 is caused by overexposure of the sensor ... to much light on those sensels during the interval that the shutter is open ... caused by too much luminance, too long a shutter interval, too open an aperture. I think that "blown sensels/pixels" is the jargon used to describe this.

Case 3 is caused by too much gain, too high an ISO, and the term "clipping" is used to describe this.

IS the above correct?

I think it's indeed a combination of the first 2 points in the ISO 200 shot, since the light sources were obviously much brighter than the surrounding areas, a problem seen in many "nightscapes". Basically, even using all the DR of the sensor available by sticking to base ISO and exposing largely for the highlights (didn't want to underexpose further as motion blur would become an issue and shadows and noise in general would become more compromised too, thus some highlight information in the original scene fell off the right side of the histogram still), I couldn't capture the full DR of the original scene with a single exposure. So some pixels were oversaturated, mostly those where the lightbulbs were. But what I did catch, I compressed to fit into a single jpeg, including the use of the recovery tools in Lightroom.

For the ISO 6400 file on the other hand, the camera uses more analogue gain (until about ISO 1600 IIRC), point 3 enters the equation too. Plus the last two stops, which are done digitally in camera (like most APS-C/FF cameras carrying Sony sensors, Pentax and Nikon included), where it simply multiplies values digitally (representing the extra 2 stops from ISO 1600 to ISO 6400) before writing the RAW file. This means that still some of the highlight information gets cut off before writing the RAW file. Some of that is discussed here too.

That is a useful link. I didn't realize that increasing ISO at the upper end can be purely digital, with no real reference to the gain in the Analogue to Digital Converter. So in the Fuji case highlights would not be further blown when going past ISO 1600.

If so it seems that a significant amount of the compression/blowing/clipping in the ISO 6400 image is caused by an ISO that is too high. ie when the ISO is dropped by FIVE STOPS to ISO 200 the clipping is removed leaving overexposure (and perhaps compression) as the culprit(s).

Right?

Agreed.

Next question:

You say,

100% crop from "properly exposed" (according to camera histogram) ISO 6400 RAW file:

If the camera were set at ISO 6400 and the photographer looked through the EVF (live view) he/she would see a histogram made from a BRIGHTENED JPEG due to his high ISO. In other words the 6400 histogram would be a long ways (5 stops) to the right of the 200 histogram.

Correct. Basically for the ISO 6400 shot, the EVF will show you an image that roughly reflects the ISO 6400 shot posted, where as for the ISO 200 shot it will mostly be... dark. That being said, many EVF cameras now offer the option to turn live exposure visualizations off, meaning that the EVF will no longer approximate the brightness of the resulting (in camera) jpeg, but rather average the brightness of the scene in the EVF, regardless of physical exposure settings. Which is really helpful for people who underexpose in RAW for example.

OK

I assume that the lens is wide open (at f/2.8) and that you can't expose any longer than 1/2 sec due to motion blur. So that establishes your physical exposure.

Not wide open (which is f/1.4) but a compromise between depth of field, sharpness, motion blur and (shadow) noise.

OK

You move your ISO up until your histogram fills the box and changes colour to indicate a "proper exposure". Really when you dial that ISO up like that aren't you just brightening the in-camera JPEG to give a "proper brightness" in the EVF and in the histogram? When you release the shutter and the JPEG is written to the card, is it the same JPEG as the one you saw in the EVF --- regarding brightness at least?

If the EVF is set to represent the jpeg brightness (and histogram), yes. But see comment above about the option to turn that off.

OK

Am I on the right track above?

Pretty much spot on.

Always relieved to hear that! Merci!

Can you give me some links to reliable information re compression, oversaturation, clipping?

Many thanks,

Tom

I'm not on my own computer right now, but will get back to this topic later this week and post some links.

I just re-found an excellent summary of how imagery is developed off of a sensor. It is pretty concentrated but is readable.

http://theory.uchicago.edu/~ejm/pix/20d/tests/noise/index.html

In it I found that I'd misused the term "ADU" in my above posts. I should've used ADC ...sorry!

Analogue to Digital Converter: A piece of circuitry that converts the voltage generated by the number of photons collected by each sensel (sensor pixel) to a digital number, "raw value".

Raw value: The digital number output by the ADC; the units in which raw values are measured are called ADU (analog-to-digital units) or DN (data numbers).

This is a more readable article showing how how a sensor / ADC works:

http://www.astropix.com/HTML/I_ASTROP/HOW.HTM

Lotsa fun, this photography stuff! Although I think that I'd rather be out doing this sort of thing:


Evening on Black Bay, Georgian Bay, Ontario, Canada

:-D

tom

Regards,

Jort

 

[TrojMacReady]

SNIP

I wish that I'd read your post before writing my lengthy post here!

http://www.dpreview.com/forums/post/52462329

I feel that I am still very much a beginner here and am still trying to sort out the noise from the signal in some of these more controversial topics.

I found Mike's link to WilbaW's Rebel Forum FAQ to be very helpful in this sorting out process. I haven't read it critically yet but it seems helpful.

I think you misunderstood. DOF has little to do with this, likewise for the shooting parameters you mention above. Because you wouldn't have to adjust them. You retain the same shutterspeed and F stop, hence why on the LCD or on opening the RAW file, the picture will come out underexposed when sticking to a lower ISO. Which is exactly what you correct in post processing.

Here's an example. Full scene:



100% crop from "properly exposed" (according to camera histogram) ISO 6400 RAW file:



100% crop from ISO 200 shot at the same physical exposure (F stop and shutterspeed) settings, corrected (brightened) in postprocessing:



Noise is pretty similar altough I actually prefer the noise structure in shadows from the base ISO shot. But most apparent is the difference in highlight headroom, see the rolloff.

"Rolloff" I need to explore that. The brick wall on the left in the cropped image is a good example.

When I look at an image either (ooc JPEG or processed RAW) and see those bright, lost detail, spots I wonder what causes them and come up with three possibilities:

1. Very high "compression" of the tone curve at the high end. ie very steep gradient on the right side of the histogram OR
2. sensels (in an area) are oversaturated, with a loss of spacial tonal information, as a cluster of sensels are oversaturated ... blown OR
3. ISO is so high that the signal from non-saturated sensels gets lost (in the ADU?) when digitized. Ie the (12 bit or 14 bit usually) data field for that sensel "overflows" due to too much gain. When these form an area there is a loss of spatial tonal information.... clipping.

Case 1 can be saved by some form of highlight recovery, ie stretching the histogram in the highlights.

Case 2 and Case 3 are unrecoverable.

Case 2 is caused by overexposure of the sensor ... to much light on those sensels during the interval that the shutter is open ... caused by too much luminance, too long a shutter interval, too open an aperture. I think that "blown sensels/pixels" is the jargon used to describe this.

Case 3 is caused by too much gain, too high an ISO, and the term "clipping" is used to describe this.

IS the above correct?

I think it's indeed a combination of the first 2 points in the ISO 200 shot, since the light sources were obviously much brighter than the surrounding areas, a problem seen in many "nightscapes". Basically, even using all the DR of the sensor available by sticking to base ISO and exposing largely for the highlights (didn't want to underexpose further as shadows and noise in general would become more compromised too, thus some highlight information in the original scene fell off the right side of the histogram still), I couldn't capture the full DR of the original scene with a single exposure. So some pixels were oversaturated, mostly those where the lightbulbs were. But what I did catch, I compressed to fit into a single jpeg, including the use of the recovery tools in Lightroom.

For the ISO 6400 file on the other hand, the camera uses more analogue gain (until about ISO 1600 IIRC), point 3 enters the equation too. Plus the last two stops, which are done digitally in camera (like most APS-C/FF cameras carrying Sony sensors, Pentax and Nikon included), where it simply multiplies values digitally (representing the extra 2 stops from ISO 1600 to ISO 6400) before writing the RAW file. This means that still some of the highlight information gets cut off before writing the RAW file. Some of that is discussed here too.

That is a useful link. I didn't realize that increasing ISO at the upper end can be purely digital, with no real reference to the gain in the Analogue to Digital Converter. So in the Fuji case highlights would not be further blown when going past ISO 1600.

If so it seems that a significant amount of the compression/blowing/clipping in the ISO 6400 image is caused by an ISO that is too high. ie when the ISO is dropped by FIVE STOPS to ISO 200 the clipping is removed leaving overexposure (and perhaps compression) as the culprit(s).

Right?

Agreed.

Next question:

You say,

100% crop from "properly exposed" (according to camera histogram) ISO 6400 RAW file:

If the camera were set at ISO 6400 and the photographer looked through the EVF (live view) he/she would see a histogram made from a BRIGHTENED JPEG due to his high ISO. In other words the 6400 histogram would be a long ways (5 stops) to the right of the 200 histogram.

Correct. Basically for the ISO 6400 shot, the EVF will show you an image that roughly reflects the ISO 6400 shot posted, where as for the ISO 200 shot it will mostly be... dark. That being said, many EVF cameras now offer the option to turn live exposure visualizations off, meaning that the EVF will no longer approximate the brightness of the resulting (in camera) jpeg, but rather average the brightness of the scene in the EVF, regardless of physical exposure settings. Which is really helpful for people who underexpose in RAW for example.

OK

I assume that the lens is wide open (at f/2.8) and that you can't expose any longer than 1/2 sec due to motion blur. So that establishes your physical exposure.

Not wide open (which is f/1.4) but a compromise between depth of field, sharpness, motion blur and (shadow) noise.

OK

You move your ISO up until your histogram fills the box and changes colour to indicate a "proper exposure". Really when you dial that ISO up like that aren't you just brightening the in-camera JPEG to give a "proper brightness" in the EVF and in the histogram? When you release the shutter and the JPEG is written to the card, is it the same JPEG as the one you saw in the EVF --- regarding brightness at least?

If the EVF is set to represent the jpeg brightness (and histogram), yes. But see comment above about the option to turn that off.

OK

Am I on the right track above?

Pretty much spot on.

Always relieved to hear that! Merci!

Can you give me some links to reliable information re compression, oversaturation, clipping?

Many thanks,

Tom

I'm not on my own computer right now, but will get back to this topic later this week and post some links.

I just re-found an excellent summary of how imagery is developed off of a sensor. It is pretty concentrated but is readable.

http://theory.uchicago.edu/~ejm/pix/20d/tests/noise/index.html

That was one of the websites that I was about to link! Emil has written many excellent articles on the subject, even if most of them rely more heavily on Canon sensor technology (as that's what he shoots and tests with), which has a lot more problems when underexposing in RAW (pattern noise, most notably banding).

In it I found that I'd misused the term "ADU" in my above posts. I should've used ADC ...sorry!

Analogue to Digital Converter: A piece of circuitry that converts the voltage generated by the number of photons collected by each sensel (sensor pixel) to a digital number, "raw value".

Raw value: The digital number output by the ADC; the units in which raw values are measured are called ADU (analog-to-digital units) or DN (data numbers).

This is a more readable article showing how how a sensor / ADC works:

http://www.astropix.com/HTML/I_ASTROP/HOW.HTM

Lotsa fun, this photography stuff! Although I think that I'd rather be out doing this sort of thing:


Evening on Black Bay, Georgian Bay, Ontario, Canada

Nice shot of an interesting area. Canada could see my next trip...

Regards,

Jort

 

[TrojMacReady]

I think you misunderstood. DOF has little to do with this, likewise for the shooting parameters you mention above. Because you wouldn't have to adjust them. You retain the same shutterspeed and F stop, hence why on the LCD or on opening the RAW file, the picture will come out underexposed when sticking to a lower ISO. Which is exactly what you correct in post processing.

Here's an example. Full scene:



100% crop from "properly exposed" (according to camera histogram) ISO 6400 RAW file:



100% crop from ISO 200 shot at the same physical exposure (F stop and shutterspeed) settings, corrected (brightened) in postprocessing:



Noise is pretty similar altough I actually prefer the noise structure in shadows from the base ISO shot. But most apparent is the difference in highlight headroom, see the rolloff.

Please post the full scene ISO 200 photo. Showing 100% crops of a relatively well lit area, doesn't really prove anything.

How much detail in the trees above the buildings was retained in the ISO 200 shot?

This is a test I did quite some time ago and I no longer have the full files. But the darker areas in the crops are anything but well lit, since they hit pure black too, even after a push of wait for it.... 5 stops. That's not well lit in my book and there isn't and wasn't more detail in the native ISO 6400 shot.

 

[TTMartin]

I think you misunderstood. DOF has little to do with this, likewise for the shooting parameters you mention above. Because you wouldn't have to adjust them. You retain the same shutterspeed and F stop, hence why on the LCD or on opening the RAW file, the picture will come out underexposed when sticking to a lower ISO. Which is exactly what you correct in post processing.

Here's an example. Full scene:



100% crop from "properly exposed" (according to camera histogram) ISO 6400 RAW file:



100% crop from ISO 200 shot at the same physical exposure (F stop and shutterspeed) settings, corrected (brightened) in postprocessing:



Noise is pretty similar altough I actually prefer the noise structure in shadows from the base ISO shot. But most apparent is the difference in highlight headroom, see the rolloff.

Please post the full scene ISO 200 photo. Showing 100% crops of a relatively well lit area, doesn't really prove anything.

How much detail in the trees above the buildings was retained in the ISO 200 shot?

This is a test I did quite some time ago and I no longer have the full files. But the darker areas in the crops are anything but well lit, since they hit pure black too, even after a push of wait for it.... 5 stops. That's not well lit in my book and there isn't and wasn't more detail in the native ISO 6400 shot.

We'll have to take your word for it.

But, my gut tells me that even shooting RAW, that taking that forested area above the buildings down 5 stops would have resulted in little to no recoverable data and that there would have been a loss of details in the trees.

As a matter of fact upon close inspection of the crops you posted there is a loss of roof line details towards the back of the buildings, near the trees as it is.

I'm throwing the BS flag on this one.

 

[TrojMacReady]

I think you misunderstood. DOF has little to do with this, likewise for the shooting parameters you mention above. Because you wouldn't have to adjust them. You retain the same shutterspeed and F stop, hence why on the LCD or on opening the RAW file, the picture will come out underexposed when sticking to a lower ISO. Which is exactly what you correct in post processing.

Here's an example. Full scene:



100% crop from "properly exposed" (according to camera histogram) ISO 6400 RAW file:



100% crop from ISO 200 shot at the same physical exposure (F stop and shutterspeed) settings, corrected (brightened) in postprocessing:



Noise is pretty similar altough I actually prefer the noise structure in shadows from the base ISO shot. But most apparent is the difference in highlight headroom, see the rolloff.

Please post the full scene ISO 200 photo. Showing 100% crops of a relatively well lit area, doesn't really prove anything.

How much detail in the trees above the buildings was retained in the ISO 200 shot?

This is a test I did quite some time ago and I no longer have the full files. But the darker areas in the crops are anything but well lit, since they hit pure black too, even after a push of wait for it.... 5 stops. That's not well lit in my book and there isn't and wasn't more detail in the native ISO 6400 shot.

We'll have to take your word for it.

But, my gut tells me that even shooting RAW, that taking that forested area above the buildings down 5 stops would have resulted in little to no recoverable data and that there would have been a loss of details in the trees.

As a matter of fact upon close inspection of the crops you posted there is a loss of roof line details towards the back of the buildings, near the trees as it is.

I'm throwing the BS flag on this one.

Of course there's a loss of details in shadows. We're talking about the aging 12 MP APS-C Exmor sensor here and shadows at the equivalent of ISO 6400 in both exposures. What were you expecting?

The only differences you might pick up in the shadows, are differences in blackpoint rolloff (tonecurve), which has a lot to do with how Adobe converters behave when pushing RAW files to extreme extends as done here (contrary to linear lifts in converters based on DCRAW).

This example isn't exactly news either, there are plenty APS-C Exmor examples from others out there (K-5 does best with the later 16MP, D7000 a close second with its own version of the 16 MP sensor).

 

[Spectrum2007]

I have no idea if my camera is ISO variant or invariant (Canon G15). Is it possible those of us who still believe the ISO should be set in the camera in order to get a fast shutter speed have one of those ISO variant models while Gollywop and Great Bustard (amongst others) have the invariant versions?

If I were concerned about whether my G15 was ISOless, I would take a series of pictures at different exposures and ISO settings and attempt to see if I got the same thing by brightening a dark low-ISO image as by raising the ISO.

If I have absolutely nothing to do, I may run this test.

You and me both!

 

[Spectrum2007]

I think you misunderstood. DOF has little to do with this, likewise for the shooting parameters you mention above. Because you wouldn't have to adjust them. You retain the same shutterspeed and F stop, hence why on the LCD or on opening the RAW file, the picture will come out underexposed when sticking to a lower ISO. Which is exactly what you correct in post processing.

Here's an example. Full scene:



100% crop from "properly exposed" (according to camera histogram) ISO 6400 RAW file:



100% crop from ISO 200 shot at the same physical exposure (F stop and shutterspeed) settings, corrected (brightened) in postprocessing:



Noise is pretty similar altough I actually prefer the noise structure in shadows from the base ISO shot. But most apparent is the difference in highlight headroom, see the rolloff.

WOW TrojMac. I was ready to dismiss this as an interesting but not so practical concept until I saw your examples.

I'm probably not going to practice this method full time - as I do like to see what I just shot - but when I have more time to take a couple at differing ISOs like you did, I will definately give it a go.

 

[Spectrum2007]

Thanks everyone for your comments and examples.

I've been shooting jpegs for about five years so while I'm not a beginner beginner, I am new at RAW and this idea about ignoring ISO sounded interesting.

I just wasnt quite ready for the advanced level demanded in the science forum so I appreciate all the effort you guys put into explaining this in conceptual terms that a beginner can understand.

 

[GeorgianBay1939]

SNIP

I wish that I'd read your post before writing my lengthy post here!

http://www.dpreview.com/forums/post/52462329

I feel that I am still very much a beginner here and am still trying to sort out the noise from the signal in some of these more controversial topics.

I found Mike's link to WilbaW's Rebel Forum FAQ to be very helpful in this sorting out process. I haven't read it critically yet but it seems helpful.

I think you misunderstood. DOF has little to do with this, likewise for the shooting parameters you mention above. Because you wouldn't have to adjust them. You retain the same shutterspeed and F stop, hence why on the LCD or on opening the RAW file, the picture will come out underexposed when sticking to a lower ISO. Which is exactly what you correct in post processing.

Here's an example. Full scene:



100% crop from "properly exposed" (according to camera histogram) ISO 6400 RAW file:



100% crop from ISO 200 shot at the same physical exposure (F stop and shutterspeed) settings, corrected (brightened) in postprocessing:



Noise is pretty similar altough I actually prefer the noise structure in shadows from the base ISO shot. But most apparent is the difference in highlight headroom, see the rolloff.

"Rolloff" I need to explore that. The brick wall on the left in the cropped image is a good example.

When I look at an image either (ooc JPEG or processed RAW) and see those bright, lost detail, spots I wonder what causes them and come up with three possibilities:

1. Very high "compression" of the tone curve at the high end. ie very steep gradient on the right side of the histogram OR
2. sensels (in an area) are oversaturated, with a loss of spacial tonal information, as a cluster of sensels are oversaturated ... blown OR
3. ISO is so high that the signal from non-saturated sensels gets lost (in the ADU?) when digitized. Ie the (12 bit or 14 bit usually) data field for that sensel "overflows" due to too much gain. When these form an area there is a loss of spatial tonal information.... clipping.

Case 1 can be saved by some form of highlight recovery, ie stretching the histogram in the highlights.

Case 2 and Case 3 are unrecoverable.

Case 2 is caused by overexposure of the sensor ... to much light on those sensels during the interval that the shutter is open ... caused by too much luminance, too long a shutter interval, too open an aperture. I think that "blown sensels/pixels" is the jargon used to describe this.

Case 3 is caused by too much gain, too high an ISO, and the term "clipping" is used to describe this.

IS the above correct?

I think it's indeed a combination of the first 2 points in the ISO 200 shot, since the light sources were obviously much brighter than the surrounding areas, a problem seen in many "nightscapes". Basically, even using all the DR of the sensor available by sticking to base ISO and exposing largely for the highlights (didn't want to underexpose further as shadows and noise in general would become more compromised too, thus some highlight information in the original scene fell off the right side of the histogram still), I couldn't capture the full DR of the original scene with a single exposure. So some pixels were oversaturated, mostly those where the lightbulbs were. But what I did catch, I compressed to fit into a single jpeg, including the use of the recovery tools in Lightroom.

For the ISO 6400 file on the other hand, the camera uses more analogue gain (until about ISO 1600 IIRC), point 3 enters the equation too. Plus the last two stops, which are done digitally in camera (like most APS-C/FF cameras carrying Sony sensors, Pentax and Nikon included), where it simply multiplies values digitally (representing the extra 2 stops from ISO 1600 to ISO 6400) before writing the RAW file. This means that still some of the highlight information gets cut off before writing the RAW file. Some of that is discussed here too.

That is a useful link. I didn't realize that increasing ISO at the upper end can be purely digital, with no real reference to the gain in the Analogue to Digital Converter. So in the Fuji case highlights would not be further blown when going past ISO 1600.

If so it seems that a significant amount of the compression/blowing/clipping in the ISO 6400 image is caused by an ISO that is too high. ie when the ISO is dropped by FIVE STOPS to ISO 200 the clipping is removed leaving overexposure (and perhaps compression) as the culprit(s).

Right?

Agreed.

Next question:

You say,

100% crop from "properly exposed" (according to camera histogram) ISO 6400 RAW file:

If the camera were set at ISO 6400 and the photographer looked through the EVF (live view) he/she would see a histogram made from a BRIGHTENED JPEG due to his high ISO. In other words the 6400 histogram would be a long ways (5 stops) to the right of the 200 histogram.

Correct. Basically for the ISO 6400 shot, the EVF will show you an image that roughly reflects the ISO 6400 shot posted, where as for the ISO 200 shot it will mostly be... dark. That being said, many EVF cameras now offer the option to turn live exposure visualizations off, meaning that the EVF will no longer approximate the brightness of the resulting (in camera) jpeg, but rather average the brightness of the scene in the EVF, regardless of physical exposure settings. Which is really helpful for people who underexpose in RAW for example.

OK

I assume that the lens is wide open (at f/2.8) and that you can't expose any longer than 1/2 sec due to motion blur. So that establishes your physical exposure.

Not wide open (which is f/1.4) but a compromise between depth of field, sharpness, motion blur and (shadow) noise.

OK

You move your ISO up until your histogram fills the box and changes colour to indicate a "proper exposure". Really when you dial that ISO up like that aren't you just brightening the in-camera JPEG to give a "proper brightness" in the EVF and in the histogram? When you release the shutter and the JPEG is written to the card, is it the same JPEG as the one you saw in the EVF --- regarding brightness at least?

If the EVF is set to represent the jpeg brightness (and histogram), yes. But see comment above about the option to turn that off.

OK

Am I on the right track above?

Pretty much spot on.

Always relieved to hear that! Merci!

Can you give me some links to reliable information re compression, oversaturation, clipping?

Many thanks,

Tom

I'm not on my own computer right now, but will get back to this topic later this week and post some links.

I just re-found an excellent summary of how imagery is developed off of a sensor. It is pretty concentrated but is readable.

http://theory.uchicago.edu/~ejm/pix/20d/tests/noise/index.html

That was one of the websites that I was about to link! Emil has written many excellent articles on the subject, even if most of them rely more heavily on Canon sensor technology (as that's what he shoots and tests with), which has a lot more problems when underexposing in RAW (pattern noise, most notably banding).

That might explain some of the resistance to the fundamental understanding of exposure and the effects of ISO on brightening by some Canon shooters. ie, Canon shooters, because of their experience with their sensors see a lot of noise/banding when shooting with low ISO.

This is only speculation on my part, as I respect photographers who conduct their own experiments to determine the actual performance of their gear. Of course my utmost respect goes to those artists/technicians who really KNOW the science/technology fundamentals, and who put them to good use when making photographs.

In it I found that I'd misused the term "ADU" in my above posts. I should've used ADC ...sorry!

Analogue to Digital Converter: A piece of circuitry that converts the voltage generated by the number of photons collected by each sensel (sensor pixel) to a digital number, "raw value".

Raw value: The digital number output by the ADC; the units in which raw values are measured are called ADU (analog-to-digital units) or DN (data numbers).

This is a more readable article showing how how a sensor / ADC works:

http://www.astropix.com/HTML/I_ASTROP/HOW.HTM

Lotsa fun, this photography stuff! Although I think that I'd rather be out doing this sort of thing:


Evening on Black Bay, Georgian Bay, Ontario, Canada

Nice shot of an interesting area. Canada could see my next trip...

Thanks, yes, it is a nice place!

I found another link that might be helpful to some folks who are having difficulty with this discussion:

http://www.sensorgen.info/

Tom

Regards,

Jort

 

[GeorgianBay1939]

Thanks everyone for your comments and examples.

I've been shooting jpegs for about five years so while I'm not a beginner beginner, I am new at RAW and this idea about ignoring ISO sounded interesting.

I just wasnt quite ready for the advanced level demanded in the science forum so I appreciate all the effort you guys put into explaining this in conceptual terms that a beginner can understand.

I shot jpegs for a couple of years and started RAW shooting a year ago. I was really hampered by a false understanding of exposure caused by the "Exposure Triangle". It was not until I read Gollywop's Exposure vs Brightening that the "light turned on".

This is also a more advanced but helpful post:

http://www.dpreview.com/forums/post/51785668

bobn2 is the (somewhat controversial) author of the sensorgen site which lists the sensor data for many cameras.

Hope that this helps,

Tom

 

[TrojMacReady]

I think you misunderstood. DOF has little to do with this, likewise for the shooting parameters you mention above. Because you wouldn't have to adjust them. You retain the same shutterspeed and F stop, hence why on the LCD or on opening the RAW file, the picture will come out underexposed when sticking to a lower ISO. Which is exactly what you correct in post processing.

Here's an example. Full scene:



100% crop from "properly exposed" (according to camera histogram) ISO 6400 RAW file:



100% crop from ISO 200 shot at the same physical exposure (F stop and shutterspeed) settings, corrected (brightened) in postprocessing:



Noise is pretty similar altough I actually prefer the noise structure in shadows from the base ISO shot. But most apparent is the difference in highlight headroom, see the rolloff.

WOW TrojMac. I was ready to dismiss this as an interesting but not so practical concept until I saw your examples.

I'm probably not going to practice this method full time - as I do like to see what I just shot - but when I have more time to take a couple at differing ISOs like you did, I will definately give it a go.

Thanks. It pays off to give your own camera a test run as every camera behaves differently, even some from the same exact model.

While objective measurements such as done by DXOmark and used on Sensorgen can given an idea (look for the read noise curves and values and see at which ISO's read noise is lowest, as that could mean those are good values to start from when pushing RAW files), they are no guarantee for success as there are many other factors at play that usually aren't quantified but could have an effect on outcome too. Such as banding, colour shifts, type of noise patterns in general, sample variance, heat effects from prolonged use, hot pixels, etc.