Re: [OM] Raw is ???? & ISO invariance [was GX85 Clarity Slider]

Subject: Re: [OM] Raw is ???? & ISO invariance [was GX85 Clarity Slider]
From: Wayne Shumaker <om3ti@xxxxxxxx>
Date: Sun, 01 Sep 2019 06:48:53 -0700
The sensors are ISO invariant to some extent but at some point the sensor or 
ADC will saturate. I read an article comparing Nikon to Sony to Canon 
(https://theslantedlens.com/2018/mirrorless-camera-comparison/) and in their 
test they shot at one ISO and pushed and pulled the exposure knob. Nikon tended 
to preserve the shadows and blow out the highlights while Sony preserved the 
highlights but lost in the shadows. If one is to opt for the one ISO setting, 
best to experiment with your specific camera. At some point you run out of 
photons or fill the bucket too full.

The tradeoff with pushing ISO reminds me of shooting at the golden hours with 
film. Depending on the scene, some scenes will work better with ISO + exposure 

Some interesting Sony sensors, one with quad Bayer pattern, to increase dynamic 
range with short and long pixel exposures. Maybe such a sensor will be even 
more ISO invariant.

counting photons - WayneS

At 8/31/2019 10:18 PM, Moose wrote:
>On 8/30/2019 3:23 PM, Mike Gordon via olympus wrote:
>><<<If one uses ACR (LR or PS), distortion correction is not optional.
>>Yes, no option using ACR to turn off use of metadata corrections.  I haven't 
>>reviewed this in a couple years.  Some corrections in metadata also include 
>>CA and vignetting.  Starting 2014 or slightly before, Panny lenses on Panny 
>>bodies began including  the CA correction metadata.  DXO does NR on the RAW 
>>data pre-conversion, but that is fully user controlled.  Not clear how 
>>clarity could be applied pre-conversion though some other aspects of the 
>>image are baked into the RAW files:
>While this essay is a great resource, the title is a little misleading. Raw is 
>indeed Raw, but useless until converted into something else. The point is, I 
>think, that the process is more interpretation than conversion. This is rather 
>obvious if one uses different converters on the same Raw file. I can see 
>people wanting to know which conversion is the CORRECT one, but the answer is 
>both none, as there is no inherent converted result in the Raw file, and all.
>The interesting thing that's happened since this essay is improvements in 
>sensor systems such that some cameras may be said to be ISO Invariant (over 
>some range of ISOs). That is to say, the result of shooting at ISO 1600 is 
>identical in shadows/noise to shooting at ISO 200 and increasing luminosity by 
>three stops in processing.
>Quite remarkable is that this is true of some 1" sensor systems, such as the 
>Sony RX100 and Panny ZS200:
>"The ZS200's sensor is essentially ISO invariant, so you can (in most 
>instances) shoot at base ISO and increase the brightness several stops while 
>processing the Raw image, with a minimal noise penalty. By keeping the ISO low 
>the camera captures additional highlight data instead of 'throwing it away' at 
>higher sensitivities by amplifying the signal. You can see similar results 
>from the Sony RX100 IV, which has a more modern CMOS sensor. In turn, this 
>gives some scope for underexposing a low ISO setting to protect highlights, 
>then brightening later."
>They don't mention it, but this also means it will generally be possible to 
>underexpose to maintain usable shutter speeds.
>This change makes some of the conclusions of Ctein's tests rather moot, as one 
>may choose exposure arbitrarily and adjust "exposure point" and curve later, 
>with no IQ penalty.
>Invariably Moose
>What if the Hokey Pokey *IS* what it's all about?
>Options: http://lists.thomasclausen.net/mailman/listinfo/olympus
>Archives: http://lists.thomasclausen.net/mailman/private/olympus/
>Themed Olympus Photo Exhibition: http://www.tope.nl/
Options: http://lists.thomasclausen.net/mailman/listinfo/olympus
Archives: http://lists.thomasclausen.net/mailman/private/olympus/
Themed Olympus Photo Exhibition: http://www.tope.nl/

<Prev in Thread] Current Thread [Next in Thread>