This is why no RAW on the G7

[aaykay]

Could you please provide a link to that article?

Thanks,
Bernard
[{POST_SNAPBACK}][/a]

I could not find a direct link but here is something that will help, since these are comments by Thom in dpreview (on the "lossy" compression used by Nikon in their RAW format) :

[a href=\"http://forums.dpreview.com/forums/read.asp?forum=1038&message=20888361]http://forums.dpreview.com/forums/read.asp...essage=20888361[/url]

I stated earlier that the D80 uses a better RAW version (non-compressed NEF ?) than the D40/D50 but I was mistaken.  It uses the same compressed "lossy" RAW format, as mentioned below (actual data stored in the "lossy" Nikon RAW in the low-end cameras like the D80/D50/D40 etc., seem to be around 6-7 bits, which is essentially JPEG territory).......thus no surpises that high-quality JPEGs and "RAW" (quote/unquote!) could not be easily distinguished.  The "compressed RAW" (= lossy "RAW") in the D2X and the D200 can supposedly be turned off but in case of the D80/50/40, the user does not have such an option and is forced to accept 6-7 bit of stored data, as "RAW" :

http://forums.dpreview.com/forums/read.asp...essage=20888513

Note: "Lossy" compression essentially throws away data in the compression process, while "lossless compression" as used by Canon, Sony etc., compresses the RAW files without throwing anything away....decompress the file and you get all the information back.

Bottomline, when I shoot RAW, I expect to obtain what came off the sensor.  Not a processed "lossy" ***interpretation*** of what came off the sensor.  A processed "lossy" version is not RAW, in my Lexicon.

[jani]

You win.

Why, thank you.

What is the prize, anyway, for providing factual information instead of speculation and handwaving?

[bjanes]

It uses the same compressed "lossy" RAW format, as mentioned below (actual data stored in the "lossy" Nikon RAW in the low-end cameras like the D80/D50/D40 etc., seem to be around 6-7 bits, which is essentially JPEG territory).......thus no surpises that high-quality JPEGs and "RAW" (quote/unquote!) could not be easily distinguished.  The "compressed RAW" (= lossy "RAW") in the D2X and the D200 can supposedly be turned off but in case of the D80/50/40, the user does not have such an option and is forced to accept 6-7 bit of stored data, as "RAW" :

[{POST_SNAPBACK}][/a]

According to this analysis, the compressed Nikon NEF files actually have 9.4 bits of data information. They contain 653 levels, whereas 8 bit has 256 levels. In practice, the image differences between compressed and noncompressed NEFs are minimal, and most users of Nikon cameras that give the user an option use compressed NEFs. If you are manipulating the highlight data extensively, noncompressed NEFs may be better.

[a href=\"http://www.majid.info/mylos/weblog/2004/05/02-1.html]http://www.majid.info/mylos/weblog/2004/05/02-1.html[/url]

[aaykay]

IF that is the case, it seems to me that some folks might feel the G7 was lacking because they didn't get a latent image, but something more akin to a Polaroid print.  Some folks, on the other hand, are quite happy because they have save a trip to the drug store, and got a processed negative they can print or fiddle with if they want.  I have no data, but it seems to me many many more folks had their film processed by a third party (one dunk fits all) than got their hands wet.  Maybe another factor for Canon's decision.
[a href=\"index.php?act=findpost&pid=90356\"][{POST_SNAPBACK}][/a]

Howie, all good points.  One point you forgot is that you can obtain a JPEG from RAW (in **exactly** the same format as the camera's own JPEG), by opening the RAW file in the software that came with the camera, and "save as" JPEG.  Takes 2 seconds...that is assuming you don't want to do any post processing whatsoever and simply want to replicate **exactly** what the camera does within itself.

[aaykay]

.....and most users of Nikon cameras that give the user an option use compressed NEFs.

......except in case of the D80/D40/D50/D70, where the user does not have the **option** to obtain an uncompressed NEF, as per Thom above.

[jani]

I could not find a direct link but here is something that will help, since these are comments by Thom in dpreview (on the "lossy" compression used by Nikon in their RAW format) :

http://forums.dpreview.com/forums/read.asp...essage=20888361

Thanks, aaykay, that's really interesting reading.

To my tastes, using lossy raw compression makes about as much sense as taking JPEG images at less than full resolution.

I'm still sitting on the fence regarding my sentiments with the raw format availability in the PowerShot G series, since I still remember how dissatisfied I was with it in the S40.

I do agree with Jonathan's remark that Canon seems a bit, ehrm, divided in their own opinion of what their cameras are about.

But we (I hope I can say "we") already realized that when they put the print button on the EOS 5D; printing when a compatible printer is connected is just a matter of using the SET button anyway.

[bjanes]

......except in case of the D80/D40/D50/D70, where the user does not have the **option** to obtain an uncompressed NEF, as per Thom above.
[a href=\"index.php?act=findpost&pid=90376\"][{POST_SNAPBACK}][/a]

That is why I added the qualification to my statement. I do not understand your style of debating. After shown to be wrong in one area, you respond with a non sequitur argument. My assertion stands.

[howiesmith]

Why, thank you.

What is the prize, anyway, for providing factual information instead of speculation and handwaving?
[a href=\"index.php?act=findpost&pid=90370\"][{POST_SNAPBACK}][/a]

You are welcome.  For a prize, if you feel you need one, anything your very active imagination can come up with.  I suggest a nice knit cap to keep your head warmer.

[aaykay]

That is why I added the qualification to my statement. I do not understand your style of debating. After shown to be wrong in one area, you respond with a non sequitur argument. My assertion stands.
[a href=\"index.php?act=findpost&pid=90382\"][{POST_SNAPBACK}][/a]

Sorry, not to be argumentative but if the "shown to be wrong" portion was about the "9.4 bits" etc....well I thought I pointed out in the post prior to that from Thom Hogan, who disagreed, as below (quoting from him):

-------------------------------------
Not on the D50, D70, or D80. And that "9 bits" part is a little misleading. There are two levels of compression that occur. The net result is that 12 bits of original data becomes (usually) 6-7 bits of stored data and represents about 9-10 bits of "perceptual" data. (How's that for convoluted?)
---------------------------------------

6-7 bits is 6-7 bits.  9-10 bits of "perceptual data" is not 9-bits of real informational data.  Or did I miss something ?

[bjanes]

Sorry, not to be argumentative but if the "shown to be wrong" portion was about the "9.4 bits" etc....well I thought I pointed out in the post prior to that from Thom Hogan, who disagreed, as below (quoting from him):

-------------------------------------
Not on the D50, D70, or D80. And that "9 bits" part is a little misleading. There are two levels of compression that occur. The net result is that 12 bits of original data becomes (usually) 6-7 bits of stored data and represents about 9-10 bits of "perceptual" data. (How's that for convoluted?)
---------------------------------------

6-7 bits is 6-7 bits.  9-10 bits of "perceptual data" is not 9-bits of real informational data.  Or did I miss something ?
[a href=\"index.php?act=findpost&pid=90412\"][{POST_SNAPBACK}][/a]

I would suggest that you re-read Thom's explanation, which is quoted below for your convenience. I think you did miss something.

"Thus, Nikon's "Compressed NEF" format generally manages to take 12 bits of data and put it into 6 to 7 bits. Upon decompression, visually the results are equivalent to about 10 bits of real data. (Of course, you can turn Compressed NEF off on most non-consumer Nikon bodies [D100, D200, D1 series, D2 series])."

He states that 12 bits of data are compressed into 6-7 bits for storage. However, that second compression is lossless as he explains further in his e-book on the D200. The reduction of the 4096 levels in the 12 bit raw to 653 levels does throw away data, but the highlights are rich in levels and the result is not perceptible. When it is re-expanded it contains 9-10 bits of perceptual data, which is exactly what was quoted in my reference. A bit depth of 9 corresponds to 512 levels and a depth of 12 bits to 1024 levels; thus, the 653 levels of the compressed NEF is between 9 and 10 bits of real data and nothing is convoluted. Nikon refers to this as visually lossless and it is quite a bit better than 8 bit JPEG. Canon's raw files are also losslessly compressed and contain 12 bits of data in fewer than 12 bits of storage.

It is interesting that you quote Thom when he appears to support your thesis, but ignore his analysis when it does not coincide with your preconceptions about in camera JPEG as compared to raw.  

[aaykay]

It is interesting that you quote Thom when he appears to support your thesis, but ignore his analysis when it does not coincide with your preconceptions about in camera JPEG as compared to raw. 

I think we can keep on going back and forth on this topic but bottomline, the Nikon "RAW Format" is not what I typically associate with RAW, which is losslessly compressed information that came off the sensor, without any additional processing by the camera's processing engine.  

The Nikon RAW in the higher-end cameras are minimally processed with some loss in data but the lower end Nikon dSLRs (D80/D70/D50/D40) irretrievably throws away significant amounts of data and try to "perceptually" maintain the appearance of data that is truly not available and then calls it "RAW".  To put it differently, that approach fails the "smell test", however we wish to spin it.

[Ray]

Contrast is also lower; the camera was set to its default contrast setting of 0, and raising it would reduce JPEG highlight latitude even further. I left the EXIF data intact, so you can poke through it with Bridge or whatever to your heart's content. Enjoy!
[a href=\"index.php?act=findpost&pid=90238\"][{POST_SNAPBACK}][/a]

Jonathan,
According to Steve's Digicam review of the SP-350, the contrast settings range from plus 5 to minus 5. A default setting of zero would appear to be an average setting. The midpoint contrast setting on the 5D is also the default setting. As I mentioned before, the difference in retention of highlight detail in jpegs (on the 5D)varies by 1.66 stops from a minimum to maximum contrast setting. If the aim is to get the maximum DR from a jpeg shot, the contrast setting should be at a minimum, not the default.

If it's true the contrast adjustments on the SP-350 go from -5 to +5 and that a setting of 0 has resulted in a 2/3rds stop DR disadvantage for the jpeg, I'd be interested to know how the DR compares at -5 contrast and minimum saturation settings. My estimate of 1/10 of a stop might not be too far out   .

Interestingly, on the 5D, the underexposed jpeg shot, despite the minimum contrast and saturation settings, still produces better (more vibrant) color than the RAW image overexposed by one stop and pulled back in ACR. In other words, I still have to do some work, in addition to EC adjustment, just to get the RAW image looking as good as the unadjusted jpeg.

[NikosR]

I
The Nikon RAW in the higher-end cameras are minimally processed with some loss in data but the lower end Nikon dSLRs (D80/D70/D50/D40) irretrievably throws away significant amounts of data and try to "perceptually" maintain the appearance of data that is truly not available and then calls it "RAW".  To put it differently, that approach fails the "smell test", however we wish to spin it.
[a href=\"index.php?act=findpost&pid=90443\"][{POST_SNAPBACK}][/a]

Would you care to substantiate and support your statemens above? Specifically with regards to:

1. Uncompressed NEFs in D2 series and D200 'are minimally processed with SOME LOSS OF DATA'

2. Compressed NEF's 'throws away SIGNIFICANT amounts of data'.

I think its the first time I hear the first, and the second has been beaten to death for years now, with no-one having been able to demonstrate (on a picture basis ) any (not to mention substantial) loss of visible data. At least to my knowledge.

I am eager to hear your explanation. Myself, I think you have no clue what you're talking about. I'm convinced of this, since you mention in your previous posts something about compressed NEF providing 6-7 bits of data etc. It is a well known fact that NEF compression is not a linear one as you seem to imply, but one based on psycho-visual theory. People have analysed NEF's and have published info about how this is accomplished, Thom Hogan included. You obviously not only do not know what you're talking about, but you can't even read correctly Thom's analysis.

The only fact the anyone has been able to show (theoretically) is that the compression gives away some highlight bit accuracy that could be used for recovering lost highlights during conversion without incurring highlight posterisation. Even for this, I have not seen any visual evidence to support it, being a Nikon dSLR user for more time than I care to remember.

[Jonathan Wienke]

The article also likened RAW to the latent image on a piece of film.  Is converting RAW yourself like custom wet processing film yourself, and JPG more like sending your film to the corner drugstore to be processed?

IF that is the case, it seems to me that some folks might feel the G7 was lacking because they didn't get a latent image, but something more akin to a Polaroid print.

Yes, RAW is like having an undeveloped digital negative you can process over and over again however you like. And camera JPEGs are basically Polaroid prints. Omitting RAW in a camera that is being marketed to "serious photographers" is simply inexcusable.

[Jonathan Wienke]

If it's true the contrast adjustments on the SP-350 go from -5 to +5 and that a setting of 0 has resulted in a 2/3rds stop DR disadvantage for the jpeg, I'd be interested to know how the DR compares at -5 contrast and minimum saturation settings. My estimate of 1/10 of a stop might not be too far out   .

I doubt it; minimum-contrast JPEGs in my Canon DSLRs still have about a half-stop less highlight latitude than RAW. And even at contrast setting 0, the camera JPEGs are pretty flat and washed-out looking; -5 would be even more so. I'll do a test when I get around to it.

[bjanes]

Would you care to substantiate and support your statemens above? Specifically with regards to:

1. Uncompressed NEFs in D2 series and D200 'are minimally processed with SOME LOSS OF DATA'

2. Compressed NEF's 'throws away SIGNIFICANT amounts of data'.

I think its the first time I hear the first, and the second has been beaten to death for years now, with no-one having been able to demonstrate (on a picture basis ) any (not to mention substantial) loss of visible data. At least to my knowledge.

I am eager to hear your explanation. Myself, I think you have no clue what you're talking about. I'm convinced of this, since you mention in your previous posts something about compressed NEF providing 6-7 bits of data etc. It is a well known fact that NEF compression is not a linear one as you seem to imply, but one based on psycho-visual theory. People have analysed NEF's and have published info about how this is accomplished, Thom Hogan included. You obviously not only do not know what you're talking about, but you can't even read correctly Thom's analysis.

The only fact the anyone has been able to show (theoretically) is that the compression gives away some highlight bit accuracy that could be used for recovering lost highlights during conversion without incurring highlight posterisation. Even for this, I have not seen any visual evidence to support it, being a Nikon dSLR user for more time than I care to remember.
[{POST_SNAPBACK}][/a]

To the uninformed such as aaykay, all bits are the same. However, a scientific analysis shows that the bits in the shadows of a digital image are far more critical than in the highlights because of the nature of human vision. This principal is explained in detail by Norman Koren:

[a href=\"http://www.normankoren.com/digital_tonality.html]http://www.normankoren.com/digital_tonality.html[/url]

The brightest stop of a 12 bit digital capture contains 2048 tones, but the eye can distinguish only about 70 of these as indicated by the Weber-Fechner law. Therefore, it is possible to throw away many of these imperceptible tones in processing. However, the shadows contain far fewer levels and are liable to posterization.

In converting from a 12 bit linear image to an 8 bit gamma 2.2 image, one goes from  4096 to 249 levels as indicated by Bruce Lindbloom's levels calculator. Since very few devices are capable of more than 8 bit output, an 8 bit gamma 2.2 file can contain nearly all the levels in the original file that are capable of reproduction in the print. This explains why a high quality JPEG file is visually lossless. If you nail white balance, exposure, contrast, and color saturation in a JPEG capture, the results are similar to a rendering from a 12 bit raw file; however, the raw file can withstand quite a bit more editing than the JPEG.

Likewise, the compressed NEF is visually lossless. Even though highlight data are thrown away, highlight recovery with ACR is quite effective in my experience and I usually shoot compressed NEF with my D200.  While I think that raw should be available for more advanced users, the loss of image information with the G7's JPEG is most likely minimal if you expose the shot correctly in camera.

Bill

[Jonathan Wienke]

While I think that raw should be available for more advanced users, the loss of image information with the G7's JPEG is most likely minimal if you expose the shot correctly in camera.

Assuming that exposure, white balance, color response, contrast, and tonality are all nailed perfectly in-camera, and that the camera's noise processing algorithm achieves the best possible tradeoff between keeping detail and removing noise, yes, JPEG doesn't degrade the image much compared to RAW.  But all of these conditions are rarely met simultaneously while shooting in the real world.

[howiesmith]

Assuming that exposure, white balance, color response, contrast, and tonality are all nailed perfectly in-camera, and that the camera's noise processing algorithm achieves the best possible tradeoff between keeping detail and removing noise, yes, JPEG doesn't degrade the image much compared to RAW.  But all of these conditions are rarely met simultaneously while shooting in the real world.
[a href=\"index.php?act=findpost&pid=90525\"][{POST_SNAPBACK}][/a]

Why not also assume that the majority of photographers and nearly all picture takers do not care whether these things are "nailed perfectly in-camera."  Nor do they require the ability to correct these things to perfect in their computer.  

Apparently the G7 just isn't for you and I suspect Canon in their arroganance and stupidity didn't even try to meet your personal requirements.  Maybe the G8 will be better.  Then again, maybe not.

[Jonathan Wienke]

Is converting RAW yourself like custom wet processing film yourself, and JPG more like sending your film to the corner drugstore to be processed?

You have no programming experience, but you casually dismiss the opinions of experienced programmers. Given your evident naivete about what RAW's advantages to the digital photographer, I really don't think you're qualified to meaningfully participate in this debate. Is anything you've contributed to this thread based on actual personal experience; something more concrete than blind faith in Canon's marketing department? Or are you just being contrarian for contrarianness' sake?

[howiesmith]

You have no programming experience, but you casually dismiss the opinions of experienced programmers. Given your evident naivete about what RAW's advantages to the digital photographer, I really don't think you're qualified to meaningfully participate in this debate. Is anything you've contributed to this thread based on actual personal experience; something more concrete than blind faith in Canon's marketing department? Or are you just being contrarian for contrarianness' sake?
[a href=\"index.php?act=findpost&pid=90531\"][{POST_SNAPBACK}][/a]

And on what do you base your statement that I have no programming experience?   Must one need to be a programmer to participate in this thread?  Or be a photographer?  And I don't really care what you think of my qualifications.

I do not dismiss, casually or otherwise, the experience of progammers.  Nor do I simply accept anything and everything they say, even that they are experienced programmers.  I am trying to learn more about RAW.

I was asking a question, hoping to gather more information.  Hence, the ? after the question.

"Is anything you've contributed to this thread based on actual personal experience[?]"  Yes.  "... something more concrete than blind faith in Canon's marketing department?"  Yes.  Actually I do not have blind faith in Canon or you.  "Or are you just being contrarian for contrarianness' sake?"  No, I simply don't blindly buy into your ideas.

I was merely asking a question.  Is that too threatening for you?