Why 'more or less stopped using ISO'?

[l_d_allan]

In a recent thread where the OP was about base ISO
What determines a sensors Base ISO?

I was unclear about the following posts ...

Bob, even with film ISO rating is not very useful, it is just some approximate guideline. It is even less useful with digital. You know that...

--
http://www.libraw.org/

I do know that, Iliah. Personally, I've more or less stopped using ISO.

I'm in the category of "not knowing that" ... about digital ISO being less than not very useful. Please clarify.

• Is this specifically about "iso-less sensors" like many Sony and Nikon cameras?
• So for such cameras ... just leave ISO at the camera's base ISO (typically 100)?
• ...
• Or does this also applies to Canon DSLR's (which I don't think can be described as iso-less)
• Or because modern DSLR's have so little noise that use of Auto-ISO is appropriate?
• So bobn2 and Iliah aren't concerned about what ISO is picked by Auto-ISO?

 

[Mark S Abeln]

I'm in the category of "not knowing that" ... about digital ISO being less than not very useful. Please clarify.

• Is this specifically about "iso-less sensors" like many Sony and Nikon cameras?
• So for such cameras ... just leave ISO at the camera's base ISO (typically 100)?
• ...
• Or does this also applies to Canon DSLR's (which I don't think can be described as iso-less)
• Or because modern DSLR's have so little noise that use of Auto-ISO is appropriate?
• So bobn2 and Iliah aren't concerned about what ISO is picked by Auto-ISO?

Boosting ISO usually decreases the dynamic range of the image, for all brands of cameras, although this is greatly influenced by the characteristics of your camera. Choosing exposure to preserve highlights, at base ISO, often leaves enough dynamic range in the shadows, which can be selectively brightened in post processing.

I don’t always use this process for two reasons:

• Laziness.
• Inability, at the time of capture, to see precisely what tones have been preserved well.

 

[Luke Kaven]

bobn2 wrote:

Bob, even with film ISO rating is not very useful, it is just some approximate guideline. It is even less useful with digital. You know that...

I do know that, Iliah. Personally, I've more or less stopped using ISO.

I'm in the category of "not knowing that" ... about digital ISO being less than not very useful. Please clarify.

• Is this specifically about "iso-less sensors" like many Sony and Nikon cameras?
• So for such cameras ... just leave ISO at the camera's base ISO (typically 100)?
• ...
• Or does this also applies to Canon DSLR's (which I don't think can be described as iso-less)
• Or because modern DSLR's have so little noise that use of Auto-ISO is appropriate?
• So bobn2 and Iliah aren't concerned about what ISO is picked by Auto-ISO?

The Sony Exmor family of sensors (used in, for example, the D3x/D800/D7000, Sony cams, Pentax K-5), by doing all processing on the sensor, much of it right around the photosite, reduces read noise such that read noise is near-optimal at base ISO. Near optimal.

Among these, in the D800, there is apparently a slight measured noise advantage to using analog gain up to about ISO 1200. But it is not a big advantage, and you could shoot at ISO 100 all day and all night if you choose. Everywhere from ISO 1600 and up is done strictly using digital multiplication and no analog gain, and potential highlight bits are thrown away needlessly. If you stay at ISO 1600, you can always defer on the decision to use gain for ... years if you wish.

The Canon cameras (at least the ones I know) are similar to the D3/D700, D3s. These sensors route low-level sensor signals to outboard chips for amplification and A-to-D. They are noise limited at low ISO settings. I would use full-stop ISO settings in-camera for everything up to ISO 1600. But everywhere from ISO 1600 and up, these cameras are basically linear, and you can top out your auto-ISO settings at ISO 1600 with some confidence.

 

[bobn2]

In a recent thread where the OP was about base ISO
What determines a sensors Base ISO?

I was unclear about the following posts ...

Bob, even with film ISO rating is not very useful, it is just some approximate guideline. It is even less useful with digital. You know that...

 

[xpatUSA]

The whole concept of 'ISO' is based around a methodology which says 'choose a processing regime' and then 'choose an exposure to match'. Since ISO is all about mapping from exposure to output image brightness, it isn't much use when you want to maximise capture information - then you need to use other methodologies for exposure control, which have become loosely wrapped up as 'ETTR' or 'ISOless'.

Quite so.

With my camera, if shoot a scene at ISO 100 and 1/100 sec and then at ISO 400 and 1/100 sec, the raw data is the same in each image. I guess that makes it a) an 'ISO-less' camera and b) a waste of time messing with the ISO 'knob'.

cheers . . .

 

[edhannon]

Which cameras do not change the content of the raw file when you change ISO without changing exposure (f-stop & shutter speed)? All that I am familiar with change the brightness.

All of the discusions about using bzse ISO and brightening in PP only talk about the impact to noise. The arguement is that, since SNR does not get better with increased ISO with the ISO-less cameras, you a re better off doing brightening in PP.

This ignores the potential for posterization when you push in PP. The curve of ADU wrt exposure in stops is exponential. The curve is very flat in the shadows and steep in the highlights. When you push by more thsn a couple of stops you loose a lot of bit re solution.

IMHO the best approach is to expose with as much light as you can and use ISO brightening to bring the highlights up to just short of saturation.

 

[bobn2]

Which cameras do not change the content of the raw file when you change ISO without changing exposure (f-stop & shutter speed)? All that I am familiar with change the brightness.

All of the discusions about using bzse ISO and brightening in PP only talk about the impact to noise. The arguement is that, since SNR does not get better with increased ISO with the ISO-less cameras, you a re better off doing brightening in PP.

This ignores the potential for posterization when you push in PP. The curve of ADU wrt exposure in stops is exponential. The curve is very flat in the shadows and steep in the highlights. When you push by more thsn a couple of stops you loose a lot of bit re solution.

IMHO the best approach is to expose with as much light as you can and use ISO brightening to bring the highlights up to just short of saturation.

 

[xpatUSA]

Which cameras do not change the content of the raw file when you change ISO without changing exposure (f-stop & shutter speed)?

Sigma SD9, SD10, SD 14 and maybe the other Sigmas. The camera firmware creates the raw file from the sensor analog outputs (all three of them). The ISO setting does not change the 'gain' between the sensor outputs of (SD9) 550mV and the three 12-bit ADC converters. The firmware does some scaling, noise reduction, linearization but only passes the user's ISO selection as metadata in the raw file.

All that I am familiar with change the brightness.

Then you are only familiar with cameras that mess around with the gain between the sensor and the ADC, if that is what is meant by 'change the brightness'?

--
A Statement passed without Proof carries little Weight
Ted

 

[Iliah Borg]

Which cameras do not change the content of the raw file when you change ISO without changing exposure (f-stop & shutter speed)?

Sigma SD9, SD10, SD 14 and maybe the other Sigmas.

All my "Merrill" Sigmas. Plus nearly all MF cameras, some Panasonic cameras for high ISO settings, some early Pentax cameras, etc. Maybe a good name for the method is "ISO by MakerNote tag".

 

[edhannon]

There are three places that brightening can be applied:

• With amplification between the sensor andq analog to digital conversion (e.g. Pentax *ist line)
• By using more bits in the analosg to digital conversion than are used in raw output file then bit shifting in firmware (e.g. Pentax K10)
• By processing the raw data either in camera or in the raw converter.

The last one uses cheaper hardware but results in loss of effective bit depth.

 

[Iliah Borg]

Brightening can be applied only to an image. Raw is not an image.

 

[D Cox]

The whole concept of 'ISO' is based around a methodology which says 'choose a processing regime' and then 'choose an exposure to match'. The implication is that the purpose of controlling exposure is to control output image brightness. If you are one who controls the processing after capture, for instance by processing from raw, then you don't want to control exposure to conrol brightness, you need to control it in order to capture the maximum scene information, then you process to get the required brightness. Since ISO is all about mapping from exposure to output image brightness, it isn't much use when you want to maximise capture information - then you need to use other methodologies for exposure control, which have become loosely wrapped up as 'ETTR' or 'ISOless'.

In any case, what you're trying to do is use all of the sensor's capture range for whatever is the luminosity range of the subject.

The problem I see with ETTR is the very small and low resolution display of the histogram on an LCD screen. Even in Photoshop's Levels control, it is poor (they haven't updated the display since screens were all 640x480).

This makes it difficult to find the highlights.

A "zebra" display might help (I haven't tried one), but would it show clipping on small specular highlights ?

(By the way, why do you write "methodologies" when you mean "methods" ? )

 

[D Cox]

The whole concept of 'ISO' is based around a methodology which says 'choose a processing regime' and then 'choose an exposure to match'. Since ISO is all about mapping from exposure to output image brightness, it isn't much use when you want to maximise capture information - then you need to use other methodologies for exposure control, which have become loosely wrapped up as 'ETTR' or 'ISOless'.

Quite so.

With my camera, if shoot a scene at ISO 100 and 1/100 sec and then at ISO 400 and 1/100 sec, the raw data is the same in each image. I guess that makes it a) an 'ISO-less' camera and b) a waste of time messing with the ISO 'knob'.

Is it completely identical, including the metadata ? I would expect the file header to include some indication for the raw converter telling it by how much to multiply the numbers.

 

[Iliah Borg]

The problem I see with ETTR is the very small and low resolution display of the histogram on an LCD screen.

Histogram is not telling much. Part of the image is blown, ok, but what part? Is it important part or not? Histogram does not tell it.

--

About LibRaw | LibRaw

www.libraw.org

 

[D Cox]

Brightening can be applied only to an image. Raw is not an image.

A digital image (as opposed to a film or a print) is an array of numbers with X, Y coordinates representing brightness levels at each point. Usually there are three values for each point, for the three primary colours.

A raw image is also an array of numbers stating the values read by the sensor at each X, Y point.

It is just as much an image file as is a TIFF or a GIF. JPGs could be regarded as corrupted image files.

 

[Iliah Borg]

Brightening can be applied only to an image. Raw is not an image.

A digital image (as opposed to a film or a print) is an array of numbers with X, Y coordinates representing brightness levels at each point.

No, not brightness.

 

[edhannon]

The way I use ETTR is:

• Determine through testing how many stops above the meter reading saturation of the first channel occurs. I do this by taking images of a gray card at various exposures and plotting the raw values of the three channels.
• With this knowledge I spot meter the important highlights - the ones I need detail in.
• I use Zone System to "place" important highlights below saturation.

Spectral highlights do not need to be placed below saturation - only those that I need to retain detail.

Note that I only do ETTR when I have a difficult subject brightness range and need the precision. Most of the time Av or Tv modes are adequate.

 

[xpatUSA]

With my camera, if shoot a scene at ISO 100 and 1/100 sec and then at ISO 400 and 1/100 sec, the raw data is the same in each image. I guess that makes it a) an 'ISO-less' camera and b) a waste of time messing with the ISO 'knob'.

Is it completely identical, including the metadata ? I would expect the file header to include some indication for the raw converter telling it by how much to multiply the numbers.

To paraphrase Bill Clinton "it all depends what the meaning of 'it' is" :-D

Myself, I said 'raw data', not meta-data and not 'the file'.

 

[jonas ar]

The way I use ETTR is:

• Determine through testing how many stops above the meter reading saturation of the first channel occurs. I do this by taking images of a gray card at various exposures and plotting the raw values of the three channels.
• With this knowledge I spot meter the important highlights - the ones I need detail in.
• I use Zone System to "place" important highlights below saturation.

Spectral highlights do not need to be placed below saturation - only those that I need to retain detail.

Note that I only do ETTR when I have a difficult subject brightness range and need the precision. Most of the time Av or Tv modes are adequate.

Would it not be much simpler to bracket the exposure and pick the appropriate one retrospectively?

 

[edhannon]

Even with bracketing you need to determine a starting point and how many stops to bracket.

Starting point needs to be close to where highlights are just below saturation.

In difficult exposure situations I find it best to spot meter several areas and then plan the exposure.

I have not found bracketing that useful.