Dante_Stella
Rex canum cattorumque
Looking at the interactive charts in the dpreview article, it looks like the original X100 pretty consistently exceeds the new version in dynamic range (sometimes by a little, sometimes by a lot), and this is the case in DR modes as well. Walk through the charts, ISO for ISO.
What gives? The dpreview/Fuji "explanation" does not seem to make a lot of sense. DR is simply the distance in EVs or dB between a noise floor and a highlight clip. Everything else is exposure (placement along that range) or curves (input-output translation).
To be sure, the real-world DR difference between an X100 and an X-Trans (take the X-Pro1) is hardly noticeable (and probably from X100 to -s too), but from model to model, tested on a principled basis, these numbers do not seem to be moving in the right direction.
Thoughts? Am I reading these charts incorrectly?
Dante
What gives? The dpreview/Fuji "explanation" does not seem to make a lot of sense. DR is simply the distance in EVs or dB between a noise floor and a highlight clip. Everything else is exposure (placement along that range) or curves (input-output translation).
To be sure, the real-world DR difference between an X100 and an X-Trans (take the X-Pro1) is hardly noticeable (and probably from X100 to -s too), but from model to model, tested on a principled basis, these numbers do not seem to be moving in the right direction.
Thoughts? Am I reading these charts incorrectly?
Dante
gavinlg
Veteran
They are JPEG tone curves, not sensor dynamic range. The x-trans is more contrasty in the JPEGs than the bayer x100.
Dante_Stella
Rex canum cattorumque
JPG is definitely what they are measuring, but in some ways, when the best conversion is in camera (although Iridient is up there), that result effectively *is* the sensor DR.
One thing that seems pretty apparent in the high ISO RAW tests is just how heavy-handed the in-camera NR is at ISO 3200. I can't imagine that ACR is doing something markedly different when processing identically laid out files from almost exactly the same sensor on the X-Pro, XE-1, and X100s.
Dante
One thing that seems pretty apparent in the high ISO RAW tests is just how heavy-handed the in-camera NR is at ISO 3200. I can't imagine that ACR is doing something markedly different when processing identically laid out files from almost exactly the same sensor on the X-Pro, XE-1, and X100s.
Dante
I found this thread about the DPR findings to be helpful.
Dante_Stella
Rex canum cattorumque
I love the predictable reaction in the thread above that the difference only applies to dpreview's particular testing method (which just happens to be the same testing method used to compare everything). Easy to dismiss their testing - but to get around \it, you would both need to show that another method exists, that it reaches a more favorable result, AND that it is more appropriate. That would be quite a stretch, and reality is that dynamic range is tied to a system not to a subpart like the hardware. Dxomark should have its numbers soon, which may or may not agree with dpreview.
Dante
Jamie Pillers
Skeptic
Dante, first regarding DXOmark testing: it doesn't appear that they are going to test cameras with the X-Trans sensor (or Foveon sensored cameras either). So I wouldn't wait around for those results in the near future.
Second, regarding performance of these new cameras: it seems to me that we've arrived at a point in digital camera development where most everything out there is 'good enough'. I have the X100s (upgraded from X100) I upgraded because I wanted 16Mp to use all my printer's 13x19 capabilities, and quicker focusing for street photos. When I make prints with the X100S output, I get results that are excellent. The X100 did the same. I know that manufacturers are going to continue developing 'bells and whistles' that they hope will entice me to upgrade. But unless I buy a much bigger printer, I'm safe from GAS. The bit of dynamic range difference in current non-DSLR cameras is either not noticeable IMO or can be managed just like when we managed the limitations of film.
willie_901
Veteran
The sensor dynamic range is an analog phenomenon. The sensor stores a charge proportional to the light amplitude that entered sensor well. The sensor full-well capacity is a measure of the maximum charge. The sensor read noise also generates charge. Less noise means more full-wel capacity is available to record signal. These are purely analog processes. The reason analog dynamic range decreases as ISO increases is because the sensor is underexposed. The signal is well below the sensor full-well capacity but the read noise remains essentially constant.
Eventually the voltage is digitized. If the digitizer is appropriately designed to match the analog dynamic range, then all the analog dynamic range is modeled in the raw file. This is not always the case. Some contemporary Canon cameras do not have ADCs capable of modeling their sesnors' analog dynamic range. This has led to firmware upgrades by third-party company (Magic Lantern).
What happens during jpeg rendering is entirely different. You can not create more dynamic range than the sensor's analog range. But you can throw dynamic range away during the rendering process. The DR enhancing menu options for in-camera jpeg rendering just attempt to manipulate the data to make full use of the raw file's dynamic range. Tis is a difficult task as the people who design the DR algorithms will never see the images they are trying to improve.
Eventually the voltage is digitized. If the digitizer is appropriately designed to match the analog dynamic range, then all the analog dynamic range is modeled in the raw file. This is not always the case. Some contemporary Canon cameras do not have ADCs capable of modeling their sesnors' analog dynamic range. This has led to firmware upgrades by third-party company (Magic Lantern).
What happens during jpeg rendering is entirely different. You can not create more dynamic range than the sensor's analog range. But you can throw dynamic range away during the rendering process. The DR enhancing menu options for in-camera jpeg rendering just attempt to manipulate the data to make full use of the raw file's dynamic range. Tis is a difficult task as the people who design the DR algorithms will never see the images they are trying to improve.
romi.gilles
i'm a wanderer
ahh.. read it wrong.
N.delaRua
Well-known
According to Fuji via David Hobby:
12.54 Stops
8.23 below medium gray
4.31 Above medium gray.
http://www.flickr.com/photos/davidhobby/9674078375/
Sounds like more than enough.
12.54 Stops
8.23 below medium gray
4.31 Above medium gray.
http://www.flickr.com/photos/davidhobby/9674078375/
Sounds like more than enough.
willie_901
Veteran
According to Fuji via David Hobby:
12.54 Stops
8.23 below medium gray
4.31 Above medium gray.
http://www.flickr.com/photos/davidhobby/9674078375/
I read this as well. It seems Fuji verbally gave Hobby theses numbers. There is no other context.
N.delaRua
Well-known
That's exactly right. That is why I included that smiley face. Take it at face value.