is this right?

[William W]

My English is far from good.

True.

Look on the bright side, always the glass half full and be happy that there are quite a few people who are willing to sort through your writing to get to what you mean . . . and then answer it.

My advice is to ignore what you think might be sarcasm:

Ted is your friend: even if neither of you know it.


Bill retires now.

WW

[pschlute]

About 90% of this thread went right over my head too. But if the technical nature of digital photography were to interest no-one, who would be designing the next generation of cameras ?

[george013]

The whole basis of ETTR is as you increase the amount of light (exposure) the noise ratio decreases. Adjusting ISO after exposure amplifies the noise as well as the effective brightness level of the captured charge. It will never match the lower noise achieved by accumulating a bigger charge by correct or prudent (ETTR) over exposure.

I trust someone will state this more eloquently.

Concerning the ETTR we have the same conclusion.
But what happens with a change of ISO? What I learned so far about ISO is that a higher sensitivity is gained either by multiplying the signal or by using a smaller range in the AD process. If nothing else happens why isn't it possible to change the ISO afterwards during the RAW conversion?

George

[pnodrog]

Concerning the ETTR we have the same conclusion.
But what happens with a change of ISO? What I learned so far about ISO is that a higher sensitivity is gained either by multiplying the signal or by using a smaller range in the AD process. If nothing else happens why isn't it possible to change the ISO afterwards during the RAW conversion?

George

The sensitivity of the sensor is constant and the noise has two main components. Sensor noise (newer designs are constantly reducing this) and shot noise. Shot noise reduces quickly as the exposure increases but is a major source of noise for pixels receiving little light (shadows). The only way to reduce shot noise is by increasing exposure. Increasing ISO either by amplifier gain, reducing AD reference voltage thereby changing effective scaling or multiplying value either in camera or post processing does not reduce shot noise but merely multiplies it.

[Manfred M]

Concerning the ETTR we have the same conclusion.
But what happens with a change of ISO? What I learned so far about ISO is that a higher sensitivity is gained either by multiplying the signal or by using a smaller range in the AD process. If nothing else happens why isn't it possible to change the ISO afterwards during the RAW conversion?

George

George - I suspect that language usage plays a part here to. The two examples Paul mentions are data level techniques that are made in camera. Adjustments made in post processing are made to an image that has been created by the raw processor.

There is a reasonable parallel to the film days where ISO sensitivity was generally given to us by the film manufacturer. We could sometimes ignore this and artificially increase ISO by increasing the development time (push processing) for B&W or colour reversal (slide) film. We never spoke of increasing ISO during the printing process when dealing with a “thin”, i.e. underexposed, developed film, but at times we did this to get the print.

ISO was associated with producing the negative or slide. It was never associated with any further steps just as ISO in the digital world is associated with what happens in the camera, not with what we do in post processing.

[george013]

The sensitivity of the sensor is constant and the noise has two main components. Sensor noise (newer designs are constantly reducing this) and shot noise. Shot noise reduces quickly as the exposure increases but is a major source of noise for pixels receiving little light (shadows). The only way to reduce shot noise is by increasing exposure. Increasing ISO either by amplifier gain, reducing AD reference voltage thereby changing effective scaling or multiplying value either in camera or post processing does not reduce shot noise but merely multiplies it.

I introduced the noise argument with ETTR. Not with the higher ISO.
If in the RAW histogram the pixel values are lineair to the exposure as shown by Dave, and double the ISO means half the exposure but still having that same RAW histogram, than the question still remains: can I multiply the RAW data to simulate another ISO?

George

[george013]

George - I suspect that language usage plays a part here to. The two examples Paul mentions are data level techniques that are made in camera. Adjustments made in post processing are made to an image that has been created by the raw processor.

There is a reasonable parallel to the film days where ISO sensitivity was generally given to us by the film manufacturer. We could sometimes ignore this and artificially increase ISO by increasing the development time (push processing) for B&W or colour reversal (slide) film. We never spoke of increasing ISO during the printing process when dealing with a “thin”, i.e. underexposed, developed film, but at times we did this to get the print.

ISO was associated with producing the negative or slide. It was never associated with any further steps just as ISO in the digital world is associated with what happens in the camera, not with what we do in post processing.

I'm not talking about the raster image data. I'm talking about the RAW-data as shown in RawDigger. Adjusting them before the conversion..

George

[xpatUSA]

But what happens with a change of ISO? What I learned so far about ISO is that a higher sensitivity is gained either by multiplying the signal or by using a smaller range in the AD process.

"or by using a smaller range in the AD process" ... probably nobody here will agree with that ... except me!

In most of my Sigma cameras there is no in-camera multiplication for the ISO selection. The selected ISO is passed in the raw file meta-data to the raw converter. The converter does the necessary multiplication. That means that a shot made, for example, with 1600 ISO selected will show raw values in RawDigger with 1/16 of the range of those made at 100 ISO; which, in turn, shows clearly how badly any sensor is under-exposed at high ISOs.

A conventional camera hides that nasty fact by amplifying the sensor signal before it gets to the ADC and the conventional converter buries the consequent noise with large amounts of NR. Ho hum.

If nothing else happens why isn't it possible to change the ISO afterwards during the RAW conversion?

It is possible but there is no slider or buttons marked "ISO" in the average converter that "most of us" use. However, there is a general converter called "DCraw" where you can set the brightness of the conversion during the raw conversion, just as you say. But you actually have to think about that or you'll get blown or too dark conversions. Since "most us" avoid thinking about anything and studiously avoid complications like command-line utilities, DCraw is not real popular.

FastStone Viewer uses DCraw, and that "gain" value is in the RAW settings tab (F12) for you to play with.

[pnodrog]

I introduced the noise argument with ETTR. Not with the higher ISO.
If in the RAW histogram the pixel values are lineair to the exposure as shown by Dave, and double the ISO means half the exposure but still having that same RAW histogram, than the question still remains: can I multiply the RAW data to simulate another ISO?

George

More or less yes. If at ISO 100 you take a photograph that is 50% under exposed you could then correct the exposure by adjusting it to ISO 200 in subsequent processing. It will have a similar noise to one exposed correctly at ISO 200. However it will have more noise than a correctly exposed photograph taken at ISO 100.

[pnodrog]

P.S. The above (and below) assumes using RAW files.

The opposite is more applicable to the benefits of ETTR.

If at ISO 200 a 1 stop over exposure is given (ETTR) and in latter PP it is corrected by adjusting the exposure to an equivalent of ISO 100 the noise will be similar to a photo taken at ISO 100.

Shot noise is solely dependent on the amount of exposure not the ISO.

[dje]

I introduced the noise argument with ETTR. Not with the higher ISO.
If in the RAW histogram the pixel values are lineair to the exposure as shown by Dave, and double the ISO means half the exposure but still having that same RAW histogram, than the question still remains: can I multiply the RAW data to simulate another ISO?

George

George the normal arrangement for increasing the ISO setting is to introduce analogue amplification between the sensor (pixel) output and the ADC. However these days I believe there may be some cameras that do use digital amplification (multiplication), at least for ISO settings above a certain amount (eg ISO1600). Either way, increasing the ISO setting leads to a situation where the Full Well Capacity of the sensor cannot be used and hence there is more noise in the developed image.

Dave

[xpatUSA]

I introduced the noise argument with ETTR. Not with the higher ISO.
If in the RAW histogram the pixel values are lineair to the exposure as shown by Dave, and double the ISO means half the exposure but still having that same RAW histogram, than the question still remains: can I multiply the RAW data to simulate another ISO?

George

No. Not on most cameras. On mine, yes.

Does that answer your question?

[pnodrog]

No. Not on most cameras. On mine, yes.

Does that answer your question?

Only in blindingly bright conditions....

I suggest it would be less confusing for all of us if you got yourself a proper sensor...

[xpatUSA]

Now that George has introduced ISO into the discussion, the plot has thickened, unfortunately. Originally I said:

As my raw histograms fade into obscurity, I recall that the max values were about 3700 presumably out of 4096 levels (12-bit ADC). Half a stop more is a factor of x1.414; therefore the max sensor raw exposure values would have been 3700*1.414 = 5233 meaning clipped highlights because the ADC can not render anything > 4095. In other words, half a stop up from Brian's original setting would have blown some highlights at the sensor.

Since I did not mention ISO at all in that statement, later talk of it introduces a Straw Man (Aunt Sally) argument which has taken this discussion somewhat off-track, IMHO.

As has been confirmed by several posts in this thread, a normal set of raw values can be multiplied by some constant to determine a change of exposure at the sensor - which is the actual claim that George doubts but refuses to disprove with formulae, diagrams or credible references.

Possible causes of obfuscation:

A black level added to the raw data, e.g. some Canon, Sony
A non-linear scaling of the raw data e.g. some Nikon
Exposure causing operation outside of the sensor's linear operating range.

The subjects of ISO, ETTR and discussion of SNR have little or nothing to do with the original post!