[swanlefitte]

So there was a thread about the K1 mii that said with PEF files it gained 2 stops over the K1. I was intrigued. Was this just what they used or was it inherent in the PEF files. I have always used DNG because my camera raw 7.1 didn't support PEF for 1 reason. I also figured using photoshop and DNG was more compatible. Well today I updated to camera raw 9.1.1 the last one that works with CS6.
First thing I noticed was that the profile is slightly different. I used a k-s2 with a A 50 1.4 at f8 ISO at 3200 and 1/15th. I do like the noise on the PEF files a little better but not by much. I then used topaz labs denoise 6 to do a light raw clean up. my crops are not exact but sized to 3cm x 3cm at 500/cm so that could have an effect. ( I was sloppy, I know but to lazy to redo now.)
Anyone have any observations on ISO and file type?
dng crop


pef crop


topaz on dng crop


topaz on pef crop

[Adam]

So there was a thread about the K1 mii that said with PEF files it gained 2 stops over the K1. I was intrigued.

It was false information. PEF and DNG files contain the same data. It's up to the raw converter to correctly decode that data and output a good result based on your settings. If the converter doesn't support a particular aspect of one format, it could output worse results compared to the other, though.

[jbinpg]

There is one difference between the K1 Mark1 PEF and DNG files that I can pinpoint: image output size for PEF is 7392 x 4059 whereas DNG is 7376 x 4932. Pentax probably has specific uses for the extra pixels that DNG does not.

[stevebrot]

First thing I noticed was that the profile is slightly different.

Which profile is this?

I then used topaz labs denoise 6 to do a light raw clean up.

Topaz does not work against the RAW, only against rendered images.

Anyone have any observations on ISO and file type?

FWIW, the claim on the linked article on the other thread reads that PEF from the K-1II have 2 stops better DR than DNG from (presumably) the same camera. The claim was made without any particular support or example.

Translation (from Spanish, original HERE)

In the first tests we have done, there are visible differences between the Pentax K-1 Mark II and its predecessor when pulling high sensitivities, but do not forget that the K-1 already has a great response to noise. The best performance is found when shooting in the Pentax PEF format, which offers more than two ISO steps with respect to the Adobe DNG file.

Conventional wisdom and dogma is that PEF and DNG are data identical and probably also metadata identical. I suppose someone could ping Dave Coffin of dcraw fame to confirm.

Lacking specific examples from the new model, it would be hard to know for sure.


Steve

[stevebrot]

There is one difference between the K1 Mark1 PEF and DNG files that I can pinpoint: image output size for PEF is 7392 x 4059 whereas DNG is 7376 x 4932. Pentax probably has specific uses for the extra pixels that DNG does not.

Did you slip a digit?

What are you using for RAW processing? Those I have worked with using LR and PDCU render to the published 7360 x 4912.


Steve

[swanlefitte]

Which profile is this?

Steve

I took both photos and loaded them in camera raw 9.1.1 they are different color temperature so I assume a different profile. Never shot in pef before this. The pef is closer, the flower is white under incandescent light. that is all I can say. any other ideas or what to look for is welcome of course.

Its a shame that I can't get a higher usable ISO with pef. I was hoping. I think the slight difference I see is because the pef flower has less color. oh well.

That article probably meant pef/dng was used and not jpg.

And my last comment is that the pef and dng were loaded into pscs6 and topaz was applied. to see what differences would come up.

[joergens.mi]

I had a look on my PEF's an DNG with irfanview and the newest PS CC ACR 10.3 both show me 7360 x 4912.

I know there was an update to the development process from 3 to 4. I think your "camera raw 7.1" version is using the Process 3 from 2012
and the "camera raw 9.1.1" is using new Process 4. In the 9.1.1 you can select the older process.


The DNG is a little bit more colorfull, is this one of your settings, or is this caused by the sw?

[stevebrot]

I took both photos and loaded them in camera raw 9.1.1 they are different color temperature so I assume a different profile. Never shot in pef before this.

Hard to say on this one. The color temperature is provided as a "hint" in the metadata for both file types based on the in-camera setting. (RAW captures don't have color temperature, per se.) I don't know about ACR by itself, but LR seldom uses the value from the hint without some modification. PDCU uses the hint without modification.


Steve

[pschlute]

7360*4912 here irrespective of the software used.

[jbinpg]

Did you slip a digit?

What are you using for RAW processing? Those I have worked with using LR and PDCU render to the published 7360 x 4912.


Steve

Yes, slip of the fingers earlier. Here is what exiftool reports for a K-1 Mk1 file:

$ exiftool IMGP0158.PEF | grep "Image Size"
Preview Image Size : 720x480
Image Size : 7392x4950

$ exiftool K1A_0182.DNG | grep "Image Size"
Preview Image Size : 720x480
Image Size : 7392x4950

So they are the same according to exiftool.

But the astro program PixInsight which uses dcraw as its raw converter states that PEF is 7392x4950 while DNG is 7376x4932. I know this is a fact for at least this one program since one cannot use a DNG calibration frame with PEF light frames (or vice versa) in PixInsight due to an image size mismatch.

lt seems that different converters render different file sizes.

- Jack

[Adam]

I took both photos and loaded them in camera raw 9.1.1 they are different color temperature so I assume a different profile. Never shot in pef before this. The pef is closer, the flower is white under incandescent light. that is all I can say. any other ideas or what to look for is welcome of course.

Its a shame that I can't get a higher usable ISO with pef. I was hoping. I think the slight difference I see is because the pef flower has less color. oh well.

That article probably meant pef/dng was used and not jpg.

And my last comment is that the pef and dng were loaded into pscs6 and topaz was applied. to see what differences would come up.

ACR requires updates to work with PEF files from a new camera model, whereas no updates are needed for DNGs. That probably explains the variations, unless there are other presets being applied.

[stevebrot]

Yes, slip of the fingers earlier. Here is what exiftool reports for a K-1 Mk1 file:

$ exiftool IMGP0158.PEF | grep "Image Size"
Preview Image Size : 720x480
Image Size : 7392x4950

$ exiftool K1A_0182.DNG | grep "Image Size"
Preview Image Size : 720x480
Image Size : 7392x4950

So they are the same according to exiftool.

But the astro program PixInsight which uses dcraw as its raw converter states that PEF is 7392x4950 while DNG is 7376x4932. I know this is a fact for at least this one program since one cannot use a DNG calibration frame with PEF light frames (or vice versa) in PixInsight due to an image size mismatch.

lt seems that different converters render different file sizes.

- Jack

Interesting. I used dcraw to extract image information from both a DNG and a PEF and got information similar to what you found:

DNG

Thumb size: 7360 x 4912
Full size: 7392 x 4950
Image size: 7376 x 4932
Output size: 7376 x 4932

PEF

Thumb size: 7360 x 4912
Full size: 7392 x 4950
Image size: 7392 x 4950
Output size: 7392 x 4950

I then used dcraw to process both images to 16-bit TIFF with the following results:

TIFF from DNG : 7376 x 4932
TIFF from PEF: 7392 x 4950

According to the documentation, dcraw breaks with the camera intent and processes the edge pixels from the sensor output to yield somewhat larger pixel dimensions than specified by Pentax. Why the PEF contains more edge (outside-nominal-frame) pixels is a tidy mystery and a contradiction to the conventional reasoning that PEF and DNG are data identical.



P.S. ExifTool results were the same as yours.


Steve

(...One learns something new everyday...)

[jbinpg]

The edge pixels have been used by Pentax for image info for a long time. I remember that K20 image files had a problem with magenta banding that was remedied by code that Gordon Brown over on DPReview cooked up using the edge pixel info of the images themselves.

[WorksAsIntended]

Can you compare the visible result of the outside pixels?
There is a whole bunch of possibilities which information could be contained in those pixels. If a standard is defined and its headers get to small to store all information a lot of developers just add header information to the data part in order to not change the formal standard, which of course is a very bad idea!

---------- Post added 04-14-18 at 12:13 AM ----------

Another idea: the sensor has to have more pixels than the outcomming picture for pixel shift to work.
Pentax decided to save the complete raw data but defined in the header which pixels to use. As dng is an open standard it is propably easier to implement it correctly to the interprating software, PEF is propriatary though. It would not suprise me if some software reading the data just doesnt use that header correctly and nobody ever noticed it or decided it is not an important feature. That would also explain the different behaviour in different software.

[maw]

Since the sensor size is the same for K1 and K1 MrkII at 24 x 35.9 mm, why is the resolution different, even if only for a very small amount?

K1 = 7360 x 4912 = 36,152,320 ~ 36.15 mpx

K1 MkII = 7392 x 4950 = 36,590,400 ~ 36.59 mpx

I would like to remind you that these are actual data published by DxO Mark.

So or they have got it wrong (unintentionally), or you prefer to believe your software (each has different algorithms for measurement).
What seems strange to me is that with the same sensor there is a slight difference in pixels (photo sensors) between the two Pentax. Why?
That's why I jokingly referred to the red berry, mentioned by photoptimist.