Jump to content

Monochrom M246 DNG technical analysis


sandymc

Recommended Posts

Advertisement (gone after registration)

For those interested, I took a look inside of a M246 DNG:

 

Version number shows as 1.0.0.4 - I'm told that this is production firmware

 

DNG version shows as 1.3, and "backward version" as 1.1 - this is as expected. There are no DNG opcodes, etc. This is just a basic vanilla DNG that should be compatible with most software. It works "out of the box" with current versions of all of AccuRaw, AccuRaw Monochrome and PhotoRaw.

 

There are only two slight surprises:

 

Firstly, the file is only 12 bit. Actually, with a white level of 3750, true bit depth is slightly less than 12-bits. This is lower than the 14 bits of the original M Monochrom, although in practice its very unlikely that anyone will notice a difference.

 

Secondly, there are a few EXIF string fields that aren't properly terminated. This isn't usually a problem as most software will still be able to correctly read the fields, but it's kind of sloppy work by Leica.

 

Sandy

  • Like 5
Link to post
Share on other sites

A white level of 3750 decimal just does not make any sense as the level for white for a 14-bit ADC! A white level of '3750'x after some internal processing- maybe. I wonder if white level changes with lens type. The camera could be doing a non-uniformity correction based on the lens-code, and this could bring the white level down to match the edges?

 

 

Is this an uncompressed file?  [Compressed NEF files ("lossless"- HAH!) bin high intensity values using a table]

 

Are there any uncompressed DNG files available for download? I would like to compare the data stored in the IFD and sub-IFD with that of my M Monochrom.

 

and just to add... I would notice the difference between 3750 decimal and '3FFF'x used by my M Monochrom!

Edited by Lenshacker
Link to post
Share on other sites

Why would anyone destroy perfectly good 14-bit data by only using 12-bits to store it!

 

Are the data values used directly for intensity, or do they get converted via some sort of function back to 14-bits? The DCS760m used 12-bit data. That was almost 15 years ago. I cannot believe that Leica would use 12 bits to store the image after giving us a full 14-bits in the M Monochrom. Something else has got to be going on.

 

Do you have a link to download the uncompressed DNG? I wrote a code for parsing the IFD and subIFD, want to try it on the M246 files.

Edited by Lenshacker
Link to post
Share on other sites

Looking at the sub-IFD in the DNG file, what value is listed for Tag ID '0102'x? This is the 4th Tag in the table in the M9 and M Monochrom, tags are in ascending order.

 

FOUND 23 ENTRIES.

TAG 4= 0102 0003 00000001 0010 0000 BITS PER SAMPLE

 

M9 and M Monochrom give it as 16 bits, or course the two upper bits are 0. The M8 also lists the bits per sample as 16, even though it goes through a look up table.

 

from the Leica Blog: "DNG™: Compressed 20-30MB, uncompressed 34.5MB,"

 

34.5MBytes for an uncompressed image works out to be 12 bits per sample 24MPixels. Storing 14-bit data using the traditional 16-bits would work out to be 48MBytes per uncompressed image. 

 

 

The M240 :"File Size:                                           

DNG approx. 20 Mbyte – approx. 30 Mbyte (compressed)/ 48,2 MByte (uncompressed), JPEG: Resolution and image content "

 

Now I'm depressed. 

Edited by Lenshacker
  • Like 1
Link to post
Share on other sites

 

There are only two slight surprises:

 

Firstly, the file is only 12 bit. Actually, with a white level of 3750, true bit depth is slightly less than 12-bits. This is lower than the 14 bits of the original M Monochrom, 

 

Oh dear .......

 

I have cancelled my order

 

mortified with disappointment  :(

  • Like 3
Link to post
Share on other sites

I think of bits-per-pixel as a dimension in the image, the M246 is (about) Image( 6000, 4000, 4096) and the M Monochrom is Image( 5212, 3468, 16384). The latter stores more information.

 

The high-ISO images from this camera look good. What I cannot understand: the sensor has a high dynamic range. Use 14-bit values for base/low ISO and truncate the low-order bits as sensor artifacts creep in at higher ISO. I'm going to guess that the camera is applying a lot of processing to the internal 14-bit image and storing 12-bits of it. I would be very surprised if CMOSIS is using a 12-bit ADC for the image. But- they might be, their other monochrome CMOS sensors use 12-bit ADC's.

Edited by Lenshacker
Link to post
Share on other sites

Other CMOS sensors in the CMOSIS lineup use 12-bit A/D convertors and can be programmed with either a linear conversion or a non-linear (changes slope) for different parts of the range. Essentially pixel values get more coarse for part of the range, but cover a higher total linear dynamic range of intensity. This method would work well for color images where the values are interpolated and you have 3 bands of data.If this is what the M240 is using, the firmware might be taking the 12-bit non-linear A/D output and store it as a 14-bit linear intensity value. This would not work so well with a monochrome image where the changes in slope would be distracting on the image. In that case a linear scale used for the A/D would be better, but you are limited to the number of bits coming out of the hardware.

 

I cannot believe that the camera would just throw away grey levels and take output from a 14-bit a/d and use 12-bits to store it. There has to be a good reason for a decision like that, would be nice to know what it is.

  • Like 2
Link to post
Share on other sites

The M246 does indeed use 12-bit per sample pixels.

 

Has anyone published the Button Dance? M246RAW2DNG - I bet we can get true 14-bit pixels out of it in no time. There is a 14-bit image on a PNY card in there somewhere.

 

Does the M240 do the same thing?  I would be very surprised if there was a change anywhere other than sensor filter and firmware.

 

-K

Link to post
Share on other sites

You guys know more about this than I ever will. If due to eyesight problems one most often needs EVF, then I am stuck using the M246 if I want a Monochrom camera.

 

My downside is 12 versus 14 bit files from what I can tell. In someones schooled opinion, will that be notciable in 30x40 B&W prints? 

  • Like 1
Link to post
Share on other sites

Well, theoretically that would only be possible to say if you compare prints. However, from what I have seen from the M246 output up to now, it would seem to be (nearly?-only slightly?) irrelevant as the level obtained  is incredibly high.

Link to post
Share on other sites

My downside is 12 versus 14 bit files from what I can tell. In someones schooled opinion, will that be notciable in 30x40 B&W prints? 

 

For me to reliably notice a difference you'd have to do some major post-processing, and give me a (big) magnifying glass. But there are many that would claim to be able to see the difference without any of that.....

 

Sandy

  • Like 1
Link to post
Share on other sites

The forum is full of old posts debating the bit depth of the M8, with simultaneous threads showing 30" by 40" prints using it.  As always, though, results vary based on overall workflow and technique.

 

Personally, I wouldn't need magnifying glasses; rather I'd need my distance glasses to effectively view that size print.

 

Jeff

Edited by Jeff S
  • Like 3
Link to post
Share on other sites

What worries me is the combination of increased dynamic range combined with reduced bit depth.

On the surface, this would imply that the mid tones are represented by far fewer bits - i.e. fine differentiation will be lost.

 

The only way Leica can dig their way out of this is if the sensor and ADC combine to form a non linear response similar to a classic film S curve - with compression of the highlights and shadows. 

 

The use of such a tone map should show up in the DNG files - perhaps this tag:

"ProfileToneCurve Tag 50940 (C6FC.H) Type FLOAT Count Samples * 2 Value See below Default None Usage IFD 0 or Camera Profile IFD Description This tag contains a default tone curve that can be applied while processing the image as a starting point for user adjustments. The curve is specified as a list of 32-bit IEEE floatingpoint value pairs in linear gamma. Each sample has an input value in the range of 0.0 to 1.0, and an output value in the range of 0.0 to 1.0. The first sample is required to be (0.0, 0.0), and the last sample is required to be (1.0, 1.0). Interpolated the curve using a cubic spline."

Link to post
Share on other sites

The CMOS sensor produced by CMOSIS uses 12 bit ADC which is expected as this sensor technology is most affordable way of creating a camera that will meet the needs of most enthusiasts. 

 

Well, when I last looked the CMOSIS sensor in the M240 was 14 bits, so I'm a bit confused by the "expected" part of that. Also confused by the "affordable", but that's philosophical point, not a technical one.  :wacko:

 

Sandy

Link to post
Share on other sites

What worries me is the combination of increased dynamic range combined with reduced bit depth.

On the surface, this would imply that the mid tones are represented by far fewer bits - i.e. fine differentiation will be lost.

 

The only way Leica can dig their way out of this is if the sensor and ADC combine to form a non linear response similar to a classic film S curve - with compression of the highlights and shadows. 

 

The use of such a tone map should show up in the DNG files - perhaps this tag:

"ProfileToneCurve Tag 50940 (C6FC.H) Type FLOAT Count Samples * 2 Value See below Default None Usage IFD 0 or Camera Profile IFD Description This tag contains a default tone curve that can be applied while processing the image as a starting point for user adjustments. The curve is specified as a list of 32-bit IEEE floatingpoint value pairs in linear gamma. Each sample has an input value in the range of 0.0 to 1.0, and an output value in the range of 0.0 to 1.0. The first sample is required to be (0.0, 0.0), and the last sample is required to be (1.0, 1.0). Interpolated the curve using a cubic spline."

 

There is no tone curve.

 

Sandy

  • Like 1
Link to post
Share on other sites

Join the conversation

You can post now and register later. If you have an account, sign in now to post with your account.
Note: Your post will require moderator approval before it will be visible.

Guest
Reply to this topic...

×   Pasted as rich text.   Paste as plain text instead

  Only 75 emoji are allowed.

×   Your link has been automatically embedded.   Display as a link instead

×   Your previous content has been restored.   Clear editor

×   You cannot paste images directly. Upload or insert images from URL.

×
×
  • Create New...