Adobe Reports 16 bit on 14 bit GFX 100 Shots

[GregJ]

Lightroom Version Number

Latest Version of Classic via Adobe Cloud

Operating System

1. Windows 10

I'm a GFX medium format shooter but always shoot 14, not 16 bit. Yet LR, Bridge and PS all report 16 bits on the metadata - EXIF readouts when all of my shots are 14 bit. This has caused a lot of discussion on the Medium Format Forums in DPR and other places. Can anyone explain it?

Thanks, Greg

 

[clee01l]

I'm a GFX medium format shooter but always shoot 14, not 16 bit. Yet LR, Bridge and PS all report 16 bits on the metadata - EXIF readouts when all of my shots are 14 bit. This has caused a lot of discussion on the Medium Format Forums in DPR and other places. Can anyone explain it?

Thanks, Greg

What you are seeing at 16GB is a demosaic’d RGB file. While the photosite are recorded only to 14 bits, when converted to color pixels, the standard 16 bit length is used for the pixels. If not filled to the full 16 bits, the last two bits get padded with zeros.


Sent from my iPad using Tapatalk

 

[johnrellis]

I saw the thread in dpreview.com -- not very illuminating about the metadata issue.

In LR, I think you're examining DNGs produced from the RAFs? The only place that LR reports Bits Per Sample is in Metadata > DNG > Bits Per Sample. LR won't report Bits Per Sample for any other file type.

When I examine a Sony ILCE-6500 DNG and a GFX 50S II DNG in LR, they both report DNG > Bits Per Sample = 14, while a GFX 100S DNG (shot at 16 bits) reports DNG > Bits Per Sample = 16.

I couldn't find a sample GFX 100S RAF shot at 14 bits -- can you upload one to Dropbox or similar and post the sharing link here?

 

[GregJ]

I saw the thread in dpreview.com -- not very illuminating about the metadata issue.

In LR, I think you're examining DNGs produced from the RAFs? The only place that LR reports Bits Per Sample is in Metadata > DNG > Bits Per Sample. LR won't report Bits Per Sample for any other file type.

When I examine a Sony ILCE-6500 DNG and a GFX 50S II DNG in LR, they both report DNG > Bits Per Sample = 14, while a GFX 100S DNG (shot at 16 bits) reports DNG > Bits Per Sample = 16.

I couldn't find a sample GFX 100S RAF shot at 14 bits -- can you upload one to Dropbox or similar and post the sharing link here?

That thread in DPR is lame, but there were some better ones a couple of years ago. In Bridge, it reports 16 on every one of my raf raw files and has for years. Its not a big deal, I was just curious. I have shot tens of thousands of GFX files the past three years and all of them are 14 since we all agree there is no current need to shoot at 16. Raw Digger of course reports it correctly. Adobe simply does not (I don't know why). They should fix that.

 

[johnrellis]

I'm not expert in Bridge. But if you upload a sample 14-bit RAF or DNG from the camera, I can verify that LR is (or is not) reporting the correct bits per sample and file a bug report if necessary.

 

[GregJ]

I'm not expert in Bridge. But if you upload a sample 14-bit RAF or DNG from the camera, I can verify that LR is (or is not) reporting the correct bits per sample and file a bug report if necessary.

John, I will do that when I get a minute tonight, but trust me.... Adobe reports 16 bit on all 14 bit GFX RAF raw files. It probably has something to do with what Cletus said. That was speculated before on several Medium Format Forums and it has something to do with Adobe reporting off of the demosaiced RBG file. Besides, we all know if we want to get a really good readout of the raw histo and what is really going on on the pure raw file, we need to look at raw digger. But still ... Adobe should report on a 16 vs 14 bit raw files since more people are shooting 16 now (probably unnecessarily).

 

[johnrellis]

I'd like to see the precise details of what's going on, both within LR and inside the file. I'm expert at photo metadata and LR metadata in particular, and without an actual example file, it's near impossible to determine what's happening.

"Adobe reports 16 bit on all 14 bit GFX RAF raw files."

Where is LR reporting that? In the Metadata > DNG panel or somewhere else?

 

[johnrellis]

That was speculated before on several Medium Format Forums and it has something to do with Adobe reporting off of the demosaiced RBG file.

That speculation might be right, though the fact that LR correctly reports 14 bits per sample for Sony 6500 and GFX 50S II DNGs argues against that. But without putting actual files under the microscope, it remains speculation.

 

[GregJ]

I'd like to see the precise details of what's going on, both within LR and inside the file. I'm expert at photo metadata and LR metadata in particular, and without an actual example file, it's near impossible to determine what's happening.

"Adobe reports 16 bit on all 14 bit GFX RAF raw files."

Where is LR reporting that? In the Metadata > DNG panel or somewhere else?

Hey John, I will do when I get back home tonight. I am no expert on metadata but know as much as most general photographers and I do like to play around with raw digger. There are several scientists on the DPR MF Board and they know that Adobe reports 16 on 14 bit shots. There are various theories as to why. Anyway, I don't care because I know I shoot 14 and not 16 bit. I just checked again and Bridge reports 16 bit depth on all GFX 100 raf files - right on the main metadata panel. Besides the GFX 100, I have an IR-converted (720nm) GFX 50r. Bridge reports 16 bit on the raf raw files, and 8 bit on the full-size exported 90% quality jpegs.

I also have a Leica Q2. Bridge reports 14 bit depth on all Leica Q2 DNG raw files and 8 bit on all Q2 exported full-size 90% quality jpegs.

Whoa! I just got curious and went back and looked at old Fuji APSC (X) raf raw files and Adobe reports 16 bit on Fuji XT-1, XT-2, XT-3, XT-4, XH-1 and X100 f&v raf raw files. I used to have all of those cameras and don't anymore. So I just called up some olf shots and checked.

So, it looks like Adobe reports 16 bit depth on all Fuji cameras (I have had them all except the new GFX 100s because I already have the GFX 100).

 

[johnrellis]

LR, Bridge and PS all report 16 bits on the metadata - EXIF readouts when all of my shots are 14 bit. ... Can anyone explain it?

I investigated this in depth, and the actual behavior of LR, Bridge, and Camera Raw is different from what you described:

- LR doesn't report bits per sample for manufacturers' proprietary raws.

- LR correctly reports DNGs' bits per sample in Metadata > DNG > Bits Per Sample.

- Bridge always reports 16 for manufacturers' proprietary raws in the Metadata > File Properties > Bit Depth, regardless of the actual bits per sample recorded in the raw.

- Bridge correctly reports DNGs' bits per sample in Metadata > File Properties > Bit Depth.

- Camera Raw doesn't report the raw or DNG's bits per sample. Camera Raw does display "8 bit" or "16 bit" at the bottom center, which is just the Color Space > Depth setting in Camera Raw Preferences. This specifies the bits/channel of the RGB image that will be created by Camera Raw when you open the edited image in PS.

Here are the cameras I tested:

 

[GregJ]

Thanks for doing this. I sent this to Jim Kasson on the MF Forum. I don't understand the nuances of this, except that Adobe needs to fix this and report the correct 14 vs 16 bit for GFX files, since shooting 14 vs 16 is such a big deal these days with GFX files. No one cares about DNG, since no one I know converts GFX raw RAF files to DNG unless they are going to focus stack with Helicon. Adobe needs to fix this and get up to speed. Ignoring proprietary raw data to me seems a bit archaic and lazy. I'm a big Adobe user and fan and happily pay them my money, but is anyone else of the opinion that they need to fix this or am I missing something?

 

[johnrellis]

It's pretty straightforward for Bridge to report the bits per sample of raw files. Most raw file formats are based on TIFF or (for Cannon) on the industry standard ISO Base Media File Format, and extracting the bits per sample is the same independent of the camera or manufacturer.

Adobe wants bugs in Bridge reported here:
Report Bugs: Hot (763 ideas) – Adobe Bridge Feedback

Looking at the forum, it looks like Bridge developers don't engage with the bug reports, so it's hard to tell how much Adobe pays attention to them. (With LR Classic, Adobe frequently engages with many bug reports and actively maintains their status.)

With a google search, I've found just one report of this bad Bridge behavior:
https://community.adobe.com/t5/phot...converter-13-1-gets-14-bit-depth/m-p/11985882

 

[johnrellis]

Being obsessive about metadata, I learned some more, which I'm recording for future reference:

For DNGs, you can't rely on what Exiftool and Rawdigger (which uses Exiftool) report as "bits per sample" to be the number of bits captured by the camera.

What those programs report as "bits per sample" for a DNG is really the number of bits used to store samples in the file (16). It is not the number of bits captured by the camera (usually 12 or 14). In my test set of DNGs from 26 different cameras with 12- or 14-bit sensors, I've found many that use 16 bits to store samples.

Some experimentation shows that LR and Bridge use the formula ceiling (log2 (WhiteLevel)) to compute DNG Bits Per Sample / Bit Depth.

There may be manufacturers' raw files that have the same issue (using 16 bits to store samples for camera sensors that capture just 12 or 14 bits), but I didn't find any among my random test set of raws from 41 camera models.