Hi All,

this should be pretty basic stuff, but my understanding is that most settings in the camera have no effect on the Raw file.

Does the Sharpness setting have any effect on the Raw file?

Cheers

Murray

The embedded jpeg will sport the results of the selected sharpening. The RAW file will also be tagged. There is, however, no impact on the actual RAW image data.

Is that right? I thought none of the settings would "tag" anything. What do you mean by "tag" and how does that rel;ate to settings in the RAW converter?

It tags it the same way it tags it with white balance, focal length and shutter speed. Most OEM converters will let you use camera settings as a starting point.

TIF is an acronym for Tagged Image File. Information about the image data (this is called "metadata", data about the data) is stored in so-called "tags", which are grouped in "directories". "Tagged" means a "tag" is inserted for example with the sharpening specification.



Some of the tags are standardized. Examples: image width and length, data organization. The so-called Exif data is a group of standardized tags in a directory.

Other tags are arbitrarily defined by the camera manufacturers (often violating the rules of TIFF). They are not and can not be standardized, for example because they are describing features unique to the camera and model.

Now to the raw conversion:

1. some in-camera selections are defined in Exif, but practically useless due to the small-mindedness of the designers. For example contrast and sharpness are defined as Normal, Soft or Hard - this is worthless.

2. other options are camera model specific, like HTP; of course they are not standardized.

Such selections are recorded in a non-standard directory, called Maker Note. The identification and format of such metadata is not published by the manufacturers; reverse engineering is required to find out these. For example when I was looking for the HTP record, I compared the Maker Notes of two otherwise identical shots.

Contrary to the wide-spread nonsense, that these records (tags) are "secret", raw processors could and some do use these; however, it is often a problem with the interpretation of these. For example Canon's record five levels of sharpness - but the meaning of "high" and "highest" etc. is open to interpretation, apart from the fact, that such primitive sharpening (i.e. based on a single parameter) would be a shame of any converter.

Thus raw processors could but mostly do not follow those instructions.

In general when you press the shutter button on the 5D2 (or similar camera), nearly your entire menu and button configuration at that moment get dumped into the raw file metadata, e.g., what exposure mode were you in? which AF point was selected? was Highlight Tone Priority on or off? what was the Picture Style? etc. It's as if you had an assistant who, every time you took a picture, scribbled down on a notecard your entire camera configuration at that very moment of exposure.

Chuck Westfal said several years ago, that the low-end cams WERE doing RAW sharpening. I'd not be surprised that this is or will start happening in all their cameras at some point in time, just as the heavy-handed chroma NR is becoming as standard 'feature' of most cams...

Can you provide a reference for that? I'm curious how one would even do sharpening on data that hasn't been demosaiced.

edit: removed an 'e'.

While I'm in no position to speculate what happens in the 'black box' on this or any other camera, I can say that 5D2 files can take a lot of sharpening in my experience. I do think that the AA filter is quite aggressive, in part because of the amount of sharpening I can throw at the files.

On the flip-side, I have never seen any unwanted sharpening taking place.

- N.

I am finding my 5D2 files a little soft and when I sharpen I bring out a pattern noise in blank areas very quickly. (Assume well exposed files at 100iso)

I am wondering if I a have a 'bad egg' that needs to go back to Canon or if this is a feature of the sensor. I am shooting it along side an Alpa/Sinar setup so it's bound to look a little soft...but the pattern noise is a worry...

Murray

Do you mind uploading the raw file of a sample exhibiting the pattern noise in a well exposed file at ISO 100?

Sure Gabor,

I always value your opinion!

Here are 2 crops from the CR2 file and a jpeg of the finished image which has had a few other exposures painted in for the window detail but has required loads of painting out of the pattern noise in the blank areas.

Raw File:
https://rcpt.yousendit.com/670596769/9ba9bf...2ad6b27af9a1587

The pattern noise gets much worse once 'normal' sharpening is applied. That's what surprised me about Nick's comment that the files could take a lot of sharpening.

Murray

Murray, I would consider this unit to be defective. I have seen enough 5D Mark II images and this behavior is not normal in production cameras. (The phenomenon is more commonly seen in preproduction hardware, however.)

Thanks,

I have contacted Canon and uploaded them a file to start the process...

Cheers

Murray

Murray,

The shot is well-exposed and there is no "customary" pattern noise in the raw channels. The blue channel in the black parts of the carpet is noisy, but not pattern noise and who would notice that anyway on that location. Those parts are in the 11th stop of the DR, naturally noisy, but the carpet's texture "destroys the bad noise".

All I see is vertical stripes on the wall, see the attached capture (hugely amplified!). However, this is not the typical pattern noise of the Canon sensors. Can't it be caused by the texture of the painting of the wall?

Have another example, from a different project, with dark, smooth areas for comparison?

I converted the raw with ACR using the default parameters, except I set the noise reduction and Blacks to zero in order to bring out any noise present there. Then sharpened in PS, but I don't see any noise.

Anyway, where are the two crops? I see only the full image attached.

I'd like to replicate the conversion. Did you convert it with ACR? What parameters? Which sharpening are you referring to? Scene sharpening in ACR, or output sharpening? Parameters?
I have a problem with WB: "as shot" yields softer colors (the tables and the carpet's white are more beige than in the picture you attached); on what should I pick WB, or what is the temp/tint in ACR?

Sidenote: have you noticed the blob on the narrow black wall at the far end, between the two windows, diameter about 25 pixels? I guess it is dirt on the sensor.

Hi Murray,
looks dodgy to me too!
Also serious CA with the 14mm lens!
However I get CA as well with my Nikkor 14-24 @ 14,but not as bad.
Best,

Willem.

Thanks Gabor,

I will clean the sensor!

I am converting using Lightroom 2.3 with a bit of sharpening eg .6 70% detail 25. Then when I apply any further sharpening eg a smart sharpen in PS say .8 80% then the pattern jumps out.

I have attached a crop with exaggerated sharpening to emphasize the problem.

It also becomes very bad when I use an HDR program such as Photomatix.

Sorry about the crops before - LR would not export cropped raw files and I forgot the change the text.

Murray

Now I see what you mean; my default sharpening in ACR is much-much lower.

This is not noise; such maze-like pattern is caused sometimes by the demosaicing algorythm, but here it is a faulty sensor. I have no doubt that you get a new one.

Although this certainly does not contribute to the solution, Murrey's camera needs a new sensor, but I thought a demonstration of the source of the maze-like pattern might be interesting.

First the pattern, in 300% view, then the raw channels separated. One green channel and the red show a vertical pattern, which is quite unique (perhaps this sensor is valuable as a collector's item?). Obviously this green channel is the one, which is in the same columns as the red.

The other green and the blue do not exhibit the pattern. The maze-like pattern is the result of demosaicing of this mix.

Murray,

I sharpen to similar levels on most files, but also add a lot of masking, so that there is virtually no sharpening at all in detail-free areas. I find this makes a huge difference (working in LR).

- N.

Gabor,

were these tests from the Raw file I sent or the cropped jpeg?

Cheers

Murray


PS Nick, masking the sharpening was exactly how I got around the problem but it's a lot of extra work!

Murray,

1. the crop with the maze is from the raw file converted by ACR,

2. the others are from the unadultered raw data, as the title says, Rawnalyze; those are the non-demosaiced raw channels (that's the reason there are two green channels).

One reference to this is Chuck Westfall's post that can be found at this link: Sharpening in 10D files also see his confirmation in post number 13 of the same thread.

In a much more recent post he states that no sharpening is used in .CR2 files as opposed to sharpening in some .CRW files: No sharpening in .CR2 files

Regards
Nigel

Something is wrong with that page. Anyway, I have already seen total nonsense written by Chuck Westfall; I put this one in the same basket.