When to push it

[Stephen G]

For a given aperture/shutter combo, is it better to-
A) Shoot @ base iso 160 and apply a push in Lightroom (+1 stop)
B) Shoot @ iso 320
(this example is simple, 1stop, now extend this to iso 640, etc)

Theoretically I could see why these should be the same.
Either way, you are taking the RAW data off sensor, and applying gain to it, either in-camera, or in-postprocessing.

There are lots of small details that can throw it either way though.
Is the M8 applying noise reduction according to ISO used? So then you are not getting the same RAW data as had you shot @ base ISO.
Does LR have a better "gain algorithm" (if there is such a thing)?
Etc..

Does this impact resolution & noise differently?

I plan to try and test this, though the results will likely be open to interpretation.

My bias has been towards thinking these are either the same, in-computer was better, or that the difference was small enough not to matter. I shoot at base ISO and push later as needed. This has the added benefit of not accidently shooting daylight @ ISO 1600 after forgetting to change ISO back.

Or is the simple solution that I need a Noctilux?

 

[dof]

For a given aperture/shutter combo, is it better to-
A) Shoot @ base iso 160 and apply a push in Lightroom (+1 stop)
B) Shoot @ iso 320
(this example is simple, 1stop, now extend this to iso 640, etc)

Speaking strictly re RAW files, the base sensitivity of the sensor is fixed. When shooting at higher ISO settings there is nearly always a push involved, unless for creative reasons some underexposure is desired. You are effectively underexposing for the scene and some amount of the full range of the image can be made up for in post-processing. Due to greater underexposure, less dynamic range is available at the higher ISO settings.

Theoretically I could see why these should be the same.
Either way, you are taking the RAW data off sensor, and applying gain to it, either in-camera, or in-postprocessing.

There are lots of small details that can throw it either way though.
Is the M8 applying noise reduction according to ISO used? So then you are not getting the same RAW data as had you shot @ base ISO.
Does LR have a better "gain algorithm" (if there is such a thing)?
Etc..

Does this impact resolution & noise differently?

Absolutely, yes. The many small details are part of the fun IMO and the best "Gain Algorithm" is a skilled human operator. Lightroom has many tools to increase overall and selective gain in your images. A combination Exposure, Contrast, and Curves alone provides a wealth of different "looks".

I plan to try and test this, though the results will likely be open to interpretation.

My bias has been towards thinking these are either the same, in-computer was better, or that the difference was small enough not to matter. I shoot at base ISO and push later as needed. This has the added benefit of not accidently shooting daylight @ ISO 1600 after forgetting to change ISO back.

Or is the simple solution that I need a Noctilux?

The simple solution if that you need the Noctilux only if your given aperture/shutter combo includes f1.0. While this can be used to great advantage, the thin depth of field at this aperture is not always optimal.

Best,

 

[kzphoto]

The way digital sensors are designed, they capture more detail in bright areas as opposed to dark areas.

If you're able to capture more highlight range with iso 320, and then scale the file back down, I'd do it. I'm interested to see how the tests come out.

 

[Stephen G]

OK so I took some test shots.
Here is one set.
I won't tell you which is which, to avoid bias for time being.
Process-
Fixed aperture&shutter, shoot @ ISO 160 & ISO 640. On tripod.
In LR2
Take ISO 640 image as-is, export
Take ISO 160 image, up exposure 2 stops, export.
Only other changes were to align color temperature & tint (but they were within 1% of each other anyway).

One thing I noticed-
The images do not look like they have had the same exposure. Possibly the way Leica M8 applies gain, and LR2 applies gain is a bit different, and one has some tone curve.

A

Fullsize download
Crop


B

Fullsize download
Crop

 

[tele_player]

Slim - I don't know which is which - but I think I prefer B, it looks sharper, more focused.

 

[mikenic]

Looks like "B" has more contrast. Crushed at the bottom end almost.

"A" looks to me like a better overall exposure for this subject (on my monitor anyhow).

FWIW

 

[Stephen G]

Contrary to what I expected-
A seems to have the higher dynamic range, and is in fact the ISO 640 shot.

B, when pushed 2 stops, came out looking a bit darker, but higher contrast. Essentially the highlights were about as bright as A, but the shadows and midtones were not. Attempting to push it further, or manipulate the curve did not quite get me there either. Basically it seemed as though B did not have the same dynamic range to work with.

Now, beyond what people think of the two pictures above.. what have you actually seen in your own work? How do you handle this? Any interesting articles you can link to on this?

 

[Stephen G]

And to make matters more interesting, I did a more proper post processing on B.
+9 brightness (now @ 59)
-2 blacks (now @ 3)
And all of the sudden it looks damn near identical to A..

See for yourself-
C


C FULLSIZE

cropped-

C, followed by A again


Differences here are far more subtle I think.
So maybe simple answer is - learn to use lightroom properly! (and then get LR3 for noise reduction as per the other recent thread).

 

[semilog]

Very, very interesting.

In a simple implementation of a CCD or a CMOS device it shouldn't matter. As some here have said, the chip has a base sensitivity. That said, there are games that a mfr. can play (for example, by changing the clock rate of the read circuitry) that could make a substantial difference at different ISO settings.

In the case of the class-leading Pentax K-x, this is in fact the case: the decrease in SNR for the RAW file as ISO setting is increased is less than predicted by theory. In other words, useful games are played before the RAW file is written.

 

[Stephen G]

Two more interesting ones..

Only posting the crops as I cut them down to the interesting parts
100% crops

ISO640 vs ISO160 + 1.83exp +6 brightness -3 blacks
Note i watched the tone curves and this is what it took me to get identical curves. Interesting that the answers was not "+2exp".

Which is which?

A)

link to full-res crop

B)

link to full-res crop

 

[semilog]

I'll bite. Is the second one shot at 640?

The first one shows pattern noise in the leaf, and less shadow detail.

Did you say what camera you're using for these experiments?

 

[Stephen G]

A is 640 / B is 160
M8 original

Also note-
Had a question from someone re: apparent vertical bands 1/3 from right of image.
These are not image artifacts, but the actual lighting situation. Light coming in through blinds, and shadows being cast by the dead plant branches.

I will also try disabling all NR in LR2.
Occurred to me that I had left it at whatever the default is.
Would be curious to see without.

I also notice that I may have left A very, very slightly darker, now that I see these images from another monitor...