Pages: [1]   Go Down

Author Topic: Sigma SD Quattro H - DNG Foveon RAW format available  (Read 2154 times)

Guillermo Luijk

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 1651
    • http://www.guillermoluijk.com
Sigma SD Quattro H - DNG Foveon RAW format available
« on: April 12, 2017, 02:12:44 PM »

12-bit DNG is available for Sigma RAW files:

https://www.dpreview.com/sample-galleries/4388344015/sigma-sd-quattro-h-real-world-samples-gallery



The RAW histogram is as follows:


Looking at it I one can think of:

  • Kind of pre-white balacing has been applied since all three RGB channels are aligned, and it's hard to believe that all three sensor layers have such equal sensitivity
  • The statistically abnormal peaks suggest some kind of RAW preprocessing, otherwise the histogram would be a soft curve
  • The fact that the histogram is completely filled having no empty holes, suggests that all pre-white balancing and preprocessing (perhaps just rounding to 12 bits) could be internally performed at a higher bit resolution (14 bits?)

Does anyone have any knowledge of this very particular sensor arquitecture from Sigma?.

Regards

Logged

Jack Hogan

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 631
    • Hikes -more than strolls- with my dog
Re: Sigma SD Quattro H - DNG Foveon RAW format available
« Reply #1 on: April 13, 2017, 05:00:35 AM »

Yes, lots of pre-processing there.  For instance the DNG shows three channels at full resolution but we know that the sensor produces only the B channel at full resolution, the other two at half the linear resolution.  Here are the per channel MTF curves of a Sigma dp2 Quattro, which I believe utilizes similar sensor technology to the sdQH, obtained from the native X3F raw file.  MTF50 is 0.338 c/p for the Blue channel and about 0.169 c/p for the other two.



Before being written to the DNG file the original raw data is therefore obviously partly upscaled/interpolated and the three channels mixed, I would assume both linearly and non-linearly.  In another context we call such an algorithm demosaicing.  The resulting 'raw' resolution is therefore a combination of that of the three underlying planes, one probably wouldn't be far off averaging them.  We'll see when we get a DNG file of a slanted edge captured with good technique.

Jack
« Last Edit: April 13, 2017, 05:05:35 AM by Jack Hogan »
Logged

Bruce Cox

  • Sr. Member
  • ****
  • Online Online
  • Posts: 1045
    • pbase galleries
QQ
« Reply #2 on: April 15, 2017, 08:55:29 PM »

Thanks for your explanations and your graphs.

DPReview posted one X3F file paired with a DNG.

I attached, at 100%, my processing of the X3F and DPReviews's processing of the DNG.

I think it would be more helpful to me if the Quattro H put out both DNG and X3F for the same file, but I wish a lot of things.

This sample of one suggests to me, that if I had a Quattro H, I would still want to process out of X3F files rather than DNGs.  My first two tries from DNG were even less good than DPReview's jpg.  Of course I might learn.

Bruce
Logged

Jack Hogan

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 631
    • Hikes -more than strolls- with my dog
Re: Sigma SD Quattro H - DNG Foveon RAW format available
« Reply #3 on: April 16, 2017, 05:30:53 AM »

I am not sure that shows much Bruce, just possibly the result of different starting defaults/profiles in the two software packages.  I'd be interested in seeing identical X3F and DNG captures by a Quattro of a Standard ColorChecker chart if anyone can provide a link.  My feeling is that they do a good portion of the SSF->XYZ heavy lifting in-camera before writing data to the DNG file.  Then let Adobe finish off the job from there.

Jack
Logged

Bruce Cox

  • Sr. Member
  • ****
  • Online Online
  • Posts: 1045
    • pbase galleries
Re: Sigma SD Quattro H - DNG Foveon RAW format available
« Reply #4 on: April 16, 2017, 09:23:21 AM »

I should have said I do not intend to show anything about the structure or potential of the DNG files, but rather that even an amateur can readily process images from X3F files which appear much less gelatinous than the sample from X3F provided by DPReview.

Most of your readers may already know this, but I hoped it might be of aid to the most innocent.

Bruce
Logged

Bruce Cox

  • Sr. Member
  • ****
  • Online Online
  • Posts: 1045
    • pbase galleries
Re: Sigma SD Quattro H - DNG Foveon RAW format available
« Reply #5 on: May 01, 2017, 02:40:10 PM »

I updated the firm ware on my DPQ2 with an oil spot in the lens. 

Yesterday, I blew out the high-lights in this shot saved to DNG.

Is it just wishful thinking on my part or does DNG allow a smother roll-off to the blown out areas?

Bruce

switched to color version in the evening
« Last Edit: May 01, 2017, 08:33:39 PM by Bruce Cox »
Logged

capital

  • Full Member
  • ***
  • Offline Offline
  • Posts: 194
    • Website
Re: Sigma SD Quattro H - DNG Foveon RAW format available
« Reply #6 on: May 01, 2017, 09:49:02 PM »

Hi Bruce, you can actually try out this comparison of DNG vs X3F files, though it is not an in camera generated DNG.

Find the conversion tool in this thread:

https://www.dpreview.com/forums/post/56168263

Would actually be interesting to compare an in camera generated DNG, with both a X3F and converted DNG using the above tool, to see if the Sigma DNG is sub-optimal in comparison.
Logged
Pages: [1]   Go Up