[Canada_Rockies]

ETTR refers to the position of the image in the histogram. Keep the histogram to the right as far as you can without clipping the bright areas where you want detail. This provides the most generous exposure for the shadows where any amplification will worsen the noise. It does not refer to willy-nilly exposing at some predetermined number of stops over exposed.

I find that I can get ETTR without chimping histograms by turning on the "blinkies" and setting my exposure to have at least one red blinky on a specular highlight in the image preview. I find this easier to do, and gives me the same effect of retaining as much detail as possible in the shadow areas.

[PentaxPoke]

ETTR refers to the position of the image in the histogram. Keep the histogram to the right as far as you can without clipping the bright areas where you want detail. This provides the most generous exposure for the shadows where any amplification will worsen the noise. It does not refer to willy-nilly exposing at some predetermined number of stops over exposed.

I find that I can get ETTR without chimping histograms by turning on the "blinkies" and setting my exposure to have at least one red blinky on a specular highlight in the image preview. I find this easier to do, and gives me the same effect of retaining as much detail as possible in the shadow areas.

Exactly! People are making this a bigger deal than it is. The +1 -1EV are demonstrations, not a technique.

[pingflood]

Exactly! People are making this a bigger deal than it is. The +1 -1EV are demonstrations, not a technique.

I don't think anyone here thinks otherwise; the +1 EV was just used as an example in the discussion. Obviously the goal is to maximize the exposure without blowing the highlights, whether that's 0 EV compensation, +1 or +3 or +2.4 for that matter...

[bymy141]

Actually I think you have more shadow detail in the 1600 shot.

"Better" is very subjective though!

You are right on the detail level at 1600, however it clearly shows more noise.

And of course you are right on the subjectivity!
However, that is somewhat beside the point I was trying to make in the original post.
The point I was trying to make is that people up their ISO in order to be able to handle a certain shutterspeed / aperture combination.
Applying ETTR to lower the noise effectively brings you back to the ISO 800 values, right?
So, you are back where you started from, and I'm sure many people don't realise that fact.

However, I you want more detail out of an ISO 800 shot, you have the option to use 1600 and ETTR. But isn't that a different discussion?

- Bert

[pingflood]

The point I was trying to make is that people up their ISO in order to be able to handle a certain shutterspeed / aperture combination.
Applying ETTR to lower the noise effectively brings you back to the ISO 800 values, right?
So, you are back where you started from, and I'm sure many people don't realise that fact.

THAT we can agree on. Where things went off the track is people thinking ISO 1600 +1 EV gives the same resulting image as ISO 800.

Been educational either way!

It'd be really interesting to run both the ISO 800 and 1600 shot above through something like NeatImage and see if there really is more _useful_ detail in the 1600 shot once the noise has been reduced a bit.

[jeffkrol]

Applying ETTR to lower the noise effectively brings you back to the ISO 800 values, right?
So, you are back where you started from, and I'm sure many people don't realize that fact.

Bert

Yes but I'm going to try to explain myself better (and perhaps gain my own clarity on this).
At any f stop/ speed combination, regardless of iso, you get the exact same amount of light hitting the sensor. A sensor that cannot be made more or less sensitive.
All the iso is doing (in general up to 1600, AND assuming any in-camera trickery on the part of the manufacturer, and assuming equal post gain processing parameters) is applying an analog gain to the signal to feed to the ADC for processing. My problem of understanding is that analog gain would be a tad better than a simple adjustment of the signal that was not amplified higher (higher iso) by simply manipulating the digital data. Up to iso 1600 this does not seem to be the case though. After 1600 they are generally equivalent since the camera will only amplify the signal to 1600,, any further boost is just digital math which can be done inside or outside the camera for equal gain.
ETTR only mainly applies to an individual iso when you are able to increase the physical signal to the sensor (slower shutter/larger aperture).
It appears (I'm still not totally convinced) to have a value if you also just increase the gain (same f stop/shutter) but at a higher iso (not true ETTR since you are NOT changing exposure, just post capture gain)...
As Mr. Martinec pointed out to me:
You are correct that ISO would be meaningless (ie equally well done post-capture in RAW conversion software), *if* post-amplification noise in the camera electronics could be made negligible.
But since it is not currently "negligible" the boost in gain in camera seems to hold some benefit for the shadows (and shadows only it seems). Of course "overexposing" at higher iso vs "not over exposing" at low iso has it's own drawbacks. The loss of highlight detail due to missing the recovery window, and the loss of DR due to the analog boost in the gain amplifier.. but you may gain a slightly less noisy image. I still think noise ninja or the like would generally clean this up as good as an analog boost... but??
Feel free to fix anything in error here....
Reference for later:
http://www.dgrin.com/showthread.php?t=65352&page=2

[kristoffon]

No, the A/D converter (I guess technically the amplifier prior to it) is fed a gain depending on the ISO selected (up until the point where the ISO is faked). I don't consider the A/D conversion step part of the "processor". The exposure and ISO(gain) affects the contents of the RAW data.

edit: I found a post from the guy who wrote all the information linked above on the topic, and you are incorrect, this is not something they "did 10 years ago" -- this refers to the rather current 1D3 and explains WHY.



So the reason why is that the sensor has a higher dynamic range than the circuitry following it, so they use the VGA to 'fit' the signal into a suitable range for the circuitry...

This is pretty cool stuff to learn about!

What this guy writes doesn't make the least bit of sense. The graphs he shows don't support his assertions. He makes an HDR pic of ISO 100 and ISO 1600 taken with the same exposure (DUH) what's the point? And he concludes that ISO 100 has less noise than ISO 1600. DUH. What a total waste of time.

Look at this quote from his ranting: For the 1D3, ISO 100 read noise, limited by the downstream components, is about 24 electrons. WTH is he talking about? Because if he really means 24 electrons (as opposed to, say, 24 x 10^10 electrons) then he really has absolutely no idea what he's talking about.

[pingflood]

What this guy writes doesn't make the least bit of sense. The graphs he shows don't support his assertions. He makes an HDR pic of ISO 100 and ISO 1600 taken with the same exposure (DUH) what's the point? And he concludes that ISO 100 has less noise than ISO 1600. DUH. What a total waste of time.

Look at this quote from his ranting: For the 1D3, ISO 100 read noise, limited by the downstream components, is about 24 electrons. WTH is he talking about? Because if he really means 24 electrons (as opposed to, say, 24 x 10^10 electrons) then he really has absolutely no idea what he's talking about.

The "read noise in electrons" is referenced in a ton of different articles about camera sensors, e.g. Clarkvision.com: Canon 1D Mark II Sensor Noise, Dynamic Range, and Full Well Analysis (FWIW, I think it refers to electrons per _photo site_, but then again, I'm no EE...)

Also, he states ISO 100 has *more* noise and ISO 1600 gives more useful detail in the lower areas, and having two parallel readouts from the sensor, one at 100 and one at 1600, would be sufficient to produce a much higher quality image... rather than a series of ISO settings.

I'm starting to wonder who has "no idea what he's talking about" What about that variable gain amplifier that you thought isn't used anymore? People just making it up whenever they reference it?

[Jodokast96]

The point I was trying to make is that people up their ISO in order to be able to handle a certain shutterspeed / aperture combination.
Applying ETTR to lower the noise effectively brings you back to the ISO 800 values, right?
So, you are back where you started from, and I'm sure many people don't realise that fact.

However, I you want more detail out of an ISO 800 shot, you have the option to use 1600 and ETTR. But isn't that a different discussion?

- Bert

No, you aren't quite getting it. Go back to Canada Rockies last post. He explains it perfectly. By exposing to the right, that doesn't necessarily mean you are pushing it up by any Ev value. It's easy to think that it would, and in some cases it may be true, but it's not an absolute. It's all about what part of your shot you are exposing for. All exposing to the right pretty much means is to expose for the shadows, and maybe push it a little bit more. This will reduce shadow noise when brought back down, instead of exposing for the entire scene, and then pushing the shadows if needed. Are you getting what I'm saying?

[bymy141]

Yes but I'm going to try to explain myself better (and perhaps gain my own clarity on this).
At any f stop/ speed combination, regardless of iso, you get the exact same amount of light hitting the sensor. A sensor that cannot be made more or less sensitive.
All the iso is doing (in general up to 1600, AND assuming any in-camera trickery on the part of the manufacturer, and assuming equal post gain processing parameters) is applying an analog gain to the signal to feed to the ADC for processing. My problem of understanding is that analog gain would be a tad better than a simple adjustment of the signal that was not amplified higher (higher iso) by simply manipulating the digital data. Up to iso 1600 this does not seem to be the case though. After 1600 they are generally equivalent since the camera will only amplify the signal to 1600,, any further boost is just digital math which can be done inside or outside the camera for equal gain.
ETTR only mainly applies to an individual iso when you are able to increase the physical signal to the sensor (slower shutter/larger aperture).
It appears (I'm still not totally convinced) to have a value if you also just increase the gain (same f stop/shutter) but at a higher iso (not true ETTR since you are NOT changing exposure, just post capture gain)...
As Mr. Martinec pointed out to me:
You are correct that ISO would be meaningless (ie equally well done post-capture in RAW conversion software), *if* post-amplification noise in the camera electronics could be made negligible.
But since it is not currently "negligible" the boost in gain in camera seems to hold some benefit for the shadows (and shadows only it seems). Of course "overexposing" at higher iso vs "not over exposing" at low iso has it's own drawbacks. The loss of highlight detail due to missing the recovery window, and the loss of DR due to the analog boost in the gain amplifier.. but you may gain a slightly less noisy image. I still think noise ninja or the like would generally clean this up as good as an analog boost... but??
Feel free to fix anything in error here....
Reference for later:
RAW exposure vs. High ISO - Page 2 - Digital Grin Photography Forum

Jeff,

I agree with your reasoning, I could not have said it any better.
This is also my understanding of the electronical setup of a camera.

- Bert

[bymy141]

No, you aren't quite getting it. Go back to Canada Rockies last post. He explains it perfectly. By exposing to the right, that doesn't necessarily mean you are pushing it up by any Ev value. It's easy to think that it would, and in some cases it may be true, but it's not an absolute. It's all about what part of your shot you are exposing for. All exposing to the right pretty much means is to expose for the shadows, and maybe push it a little bit more. This will reduce shadow noise when brought back down, instead of exposing for the entire scene, and then pushing the shadows if needed. Are you getting what I'm saying?

Then I still fail to understand that argument, I believe however that you do not get the point I'm trying to make in the original post.

Let me try again.
Since there no other means available to get any higher Ev than shutterspeed and aperture, getting your histogram to the right as canadian_rocky says, you'd either need to slow your shutterspeed or to open your aperture, or both.
However, that is not what you want (see original post). That is why you got yourself at that high ISO in the first place. There is no other means than those two, except for increasing ISO even further...

I agree with Pingflood, that if you are at any ISO (except for your lowest perhaps) you may get better results (may it be shadow detail [what I have seen] / or less noise [what I have not seen]), but choosing a higher ISO value and applying ETTR.
There is also in camera noise generation besides the sensor (cross talk being one of them, the one that is the main source of "banding"). So in both settings the result in detail and noise output may differ a little.

I understand very well the idea of ETTR and I can follow the logic, but it is not complete. Still I believe that you are misled.
Jeff Kroll has given an excellent explanation of the mechanics of a camera system.
ISO gain change will not change the amount of light captured by the sensor, it will only increase signal gain before it is digitized.
So, in order to increase the sensors s/n ratio (worst in low signal / dark areas), the only thing you can do is to increase the light captured.
But that is not what you wanted because your lens is 200mm, it is a handheld shot and you want to maintain some DOF!
And of course in a high ISO setting adding more light to the sensor will create a right positioned histogram. But then effectively nothing has changed. Except for the highlight clipping loss due to digital multiplication of the signal.

- Bert

[Jodokast96]

No. Think of it this way. It's about how you choose a "correct" shutter and aperture, regardless of ISO. The point of it is, that what you would normally choose as "correct" is in fact wrong when using ETTR. If shutter and aperture need to be set at very specific settings for a certain shot, then ETTR is not something to be used. And if you are at the absolute max at all settings, then it can't be used. But in most situations, you have some latitude with regard to all of the settings. So with that latitude, you push for as much over-exposure as you can get, within the chosen ISO. If you can push so far that a lower ISO can be used, then obviously you go with the lower ISO. But if not, you take that latitude and make the best of it.

[Canada_Rockies]

No. Think of it this way. It's about how you choose a "correct" shutter and aperture, regardless of ISO. The point of it is, that what you would normally choose as "correct" is in fact wrong when using ETTR. If shutter and aperture need to be set at very specific settings for a certain shot, then ETTR is not something to be used. And if you are at the absolute max at all settings, then it can't be used. But in most situations, you have some latitude with regard to all of the settings. So with that latitude, you push for as much over-exposure as you can get, within the chosen ISO. If you can push so far that a lower ISO can be used, then obviously you go with the lower ISO. But if not, you take that latitude and make the best of it.

Well put. Pentax designers have chosen to virtually eliminate blown highlights by biasing the metering system slightly toward under exposure. For most of my photography, I use the Pentax metering settings. They are as good as any other manufacturer's choice, and better than most.

When I have an image to take that has a high dynamic range and cannot use HDR to assemble images, then ETTR takes over. The closer my exposure can get to clipping only the specular highlights while clipping ALL the specular highlights, the better my shadow areas will be exposed. That's what it's all about, folks.

[bymy141]

If you can push so far that a lower ISO can be used, then obviously you go with the lower ISO. But if not, you take that latitude and make the best of it.

Ok, I think I've got it.
Still, to me that, sounds like the world upside down.

Because it would lead to a situation where you start with your highest ISO and work your way lower finding the minimum ISO without overexposing the picture?
Or, put it in an other way, what procedure do you suggest you'd follow?

Also, in the tests I did, I cannot see any noise reduction advantages, ETTR with my K10D gives a little better detail but worse noise results.

- Bert

[jeffkrol]

Ok, I think I've got it.
Still, to me that, sounds like the world upside down.

Because it would lead to a situation where you start with your highest ISO and work your way lower finding the minimum ISO without overexposing the picture?
Or, put it in an other way, what procedure do you suggest you'd follow?

Also, in the tests I did, I cannot see any noise reduction advantages, ETTR with my K10D gives a little better detail but worse noise results.

- Bert

2 things.. do what works for you regardless if it seems right or not.
Also shoot at the lowest iso possible and the slowest aperature/speed possible. It will always be better than higher iso's unless you really blow the exposure.
What is this image you say... well it's an image I shot outside at iso 1600 w/ +2EV.
The other one is at 200iso w/ the exact same f stop/shutter.. effectively 4 stops underexposed from the 1600iso one....
This is just a real rough comparison but I encourage people to try for themselves keeping just one thing in mind.. not all processors are equal.....n


Opp's I probably used the iso 800 as compared to the 1600.
200 below: