R8R said:
Tif, psd, pdf, etc for printing. Jpg maybe for very small prints of snapshots.

A jpg file is a compressed version of the image. Depending on the level of compression, there are mild to severe compromises made in image quality.

PDF is just a container format, which will then contain i.e. a TIFF or JPEG image. That's similar to .mov for video, the container format doesn't specify what codec is used for the file itself.

As for the JPEGs: As I said, just check it out yourself, it's not expensive to do. Just be sure you select an image that has critical stuff in it, i.e. shadow detail, highlight detail, all that. For around 5 bucks, you can do a pretty nice series of one, two three images that give you a good comparison.

When in doubt, why worry about it anyway and just send TIFF. For larger-format "artsy" prints, I always send TIFF. I doubt it makes a difference, but if it did, then I'd be ok. For this kind of print work, I never send batches of photos anyway, but one or two, and who cares if it takes 20 seconds or ten minutes to upload. Just don't worry and send the large file, done.

sevencrossing said:
so if you are sending a file to be printed what is better TIFF or jpeg ?

Tif, psd, pdf, etc for printing. Jpg maybe for very small prints of snapshots.

A jpg file is a compressed version of the image. Depending on the level of compression, there are mild to severe compromises made in image quality.

More info (probably more than you wanted) here:
http://en.wikipedia.org/wiki/.jpg

chris_weinert said:

Anyway, my result was, that while there is a huge quality difference between the different labs, even with the best lab I could not see differences between an uncompressed (TIFF) and low-compression (JPEG highest) file. Then again, I tested this only at the highest resolution (300 dpi), might be different at lower resolutions, but who cares, actually.

I have noticed the same for the majority of my prints. The only time I have noticed a difference is when the image contains large swing in gradations of colors - Tiffs are better for that as they have a larger bit depth than jpegs. Printers anymore do a great job in "filling" in color and smoothing it out as well.

When I send my stuff out, I just go with whatever the printer prefers and send along or embed the color profiles.

TaoTeJared said:
Chris I'm sorry, but you are very mistaken in your understanding in the difference in Raw(NEF) files, what info they contain, and what you loose by converting to Tiffs. You are missing much information to build a better understanding. I could attempt to describe it, but there are loads of resources on the web that will do a much better job than I.

TTJ, thanks for clarifying – will go look for some "enlightenmen" then on the web. :-)

As for the printing:

I did a rather extensive test with different file formats, resolutions and different labs a year ago. This is something I'd advise everyone to test it themselves, since then you believe it best ;-) Or maybe you even see differences. It's not that much of work to put together a file with some "critical" content, and it won't cost much, either.

Anyway, my result was, that while there is a huge quality difference between the different labs, even with the best lab I could not see differences between an uncompressed (TIFF) and low-compression (JPEG highest) file. Then again, I tested this only at the highest resolution (300 dpi), might be different at lower resolutions, but who cares, actually.

I wrote a bunch of stuff and just deleted it since it can be summed up in an easy fact - Tiffs are in no way better or equal than Raw files.

Chris I'm sorry, but you are very mistaken in your understanding in the difference in Raw(NEF) files, what info they contain, and what you loose by converting to Tiffs. You are missing much information to build a better understanding. I could attempt to describe it, but there are loads of resources on the web that will do a much better job than I.

As I said above, as a final "print" they are excellent for that, but not for the beginning of the workflow. They are better than jpegs but not Raw files. I would put Raw files at the top, then PSD, then Tiff, and then everything else below that. There are other file formats that are arguably better than Tiffs and even PSD files, but they have either not caught on, or are in their infancy and not mainstream yet.

20 years ago, Tiffs where the default and were the best - technology has moved forward and that is not the case anymore.

sevencrossing said:
so if you are sending a file to be printed what is better TIFF or jpeg ?

Tiff or psd file for sure. There is a much wider color gamut.

One other thing to watch out for is color space manipulation during the NEF to TIFF conversion. Depending on which converter you use it may be converting to sRGB or aRGB space, both of which are significantly less than 12, 14 or 16 bits. (even though it will happily store the smaller value in the most significant of the 16 bits) Google "linear raw conversion" for lots of juicy reading. Also this looks interesting as well:

http://www.nx101.com/dynamicrange.html

That's nx101 dot com if the link gods punish me. One other thing to note is that RAW file conversion is as much an art as science and it has improved significantly over the last many years. If you keep the RAW file someday a better converter may be available to extract data from it. And finally, a 36megapixel image would yield a 200MB+ TIFF file!

Eric, TTJ and Ironheart: Thanks for your input again.

TaoTeJared said:
Tiff files do not retain the raw data. It is no different than making a print from a negative, and re-scanning it, then editing it.

Hm, I think this is not a good comparison, because a re-scan from a print loses a LOT of information, first of all simply from the print itself, that is a fraction of the dynamic range of the negative, and then from the scanner. So there's two steps of destructive processing in between, while with a file conversion and even an extension of the dynamic range to 16 bit, with no settings applied, there should, and that's the point: should be none. And of course the larger data space should contain all the RAW data from the original. If they didn't they would need to be eliminated, and why (and how) would they? It's like cutting off a 3-ft.-wide table so it can fit through a 5-ft.-wide door.

TaoTeJared said:
Case in point - when you convert a NEF file, the white balance (and other settings) are set and applied. Therefor it changes the original tone of the file which can limit your ability to process it. If you WB Bracket a set of sunsets, the colors will swing across the spectrum. RAW files, you can change those back to Neutral (remove the WB) - Tiff files you can not do this.

would disagree, in the sense that a RAW file implicitly already DOES contain a WB setting: the one of the sensor without any modification. RAW data are often treated like they were a mystical black box that contains all this information – which they are not. It's simply the unprocessed image data from the sensor in a 14-bit file, plus some metadata. There can't be more information in there than 14-bit per channel. And it IS an image file, so of course, there IS a color value per pixel, with a color depth of 42 bits, which implies that there IS a white balance, simply what the sensor measured. Even with a RAW image, the white balance doesn't do anything else than re-calculating, hence shifting this color information of every pixel, using an algorithm.

Now if I modify the settings and then convert, then the original data are lost and it's replaced with the new data. But if don't apply any changes (possible in Aperture and I guess in LR, too) and say "open with PS", you should, theoretically, just get a 1:1 copy of the original information, and hence be able to modify WB etc. the same way as with the original RAW file.

Now that I wrote this, I'm realizing that you wrote "can limit your ability to process it", and not that they do. I guess we're all on the same page here, just putting it differently, as Eric already said.

And yeah,

Ironheart said:
Try it out for yourself. Try noise reduction per and post conversion and be your own judge. Same for the WB. If all you are worried about is pulling data out of shadows, then the 16 bit TIFF conversion should preserve most of it. But try it and let us know.

I already did that meanwhile. Better than philosophizing here for hours, I do get that myself. ;-)

One advantage of the Aperture editing over Photoshop is that the algorithms for the shadow recovery are a lot better than those of the Adobe Camera Raw module, so it's not really comparable. The advantage with ACR is that you can have the original file as a smart object, meaning non-destructively processed RAW. I'm not sure if how this can be done with a file linked directly from Aperture yet.

I'll post here if I figure that out or if I have results that are worth showing.

Thanks,
Chris

TaoTeJared said:
...
Tiffs are best as a the "final" print/save in a workflow - not at the beginning of it.

so if you are sending a file to be printed what is better TIFF or jpeg ?

Last I checked a TIFF file is larger than the corresponding RAW so there is no advantage especially when you consider the non-destructive editing you get by default with RAW. For example, As TTJ pointed out WB adjustments are easily changed or backed out from RAW but after conversion would be difficult. Also noise reduction algorithms seem to work better on RAW files.

But don't take anyone's word for it. Try it out for yourself. Try noise reduction per and post conversion and be your own judge. Same for the WB. If all you are worried about is pulling data out of shadows, then the 16 bit TIFF conversion should preserve most of it. But try it and let us know.

Tiff files do not retain the raw data. It is no different than making a print from a negative, and re-scanning it, then editing it.

Case in point - when you convert a NEF file, the white balance (and other settings) are set and applied. Therefor it changes the original tone of the file which can limit your ability to process it. If you WB Bracket a set of sunsets, the colors will swing across the spectrum. RAW files, you can change those back to Neutral (remove the WB) - Tiff files you can not do this.

Tiffs are best as a the "final" print/save in a workflow - not at the beginning of it.

Chris, no mistake. My last post agrees with you. What I meant to convey is that if any processing is done prior to the conversion to TIFF, then as R8R indicated the conversion can irreversibly lock the associated changes into the image. If those processing steps resulted in any clipping, then data ends up being lost.

The RAW file always has the initial image data at maximum fidelity. A TIFF file might have some losses (depending on any processing done to the RAW before conversion to TIFF). Any losses are a consequence of processing done pre-conversion.

I agree that you should be able to take the camera RAW data into a TIFF without any losses. In fact I think you could take such TIFF data and extract the original RAW data.

As I see it, losses can only arise when the RAW file is processed and subsequently converted to a TIFF. In the most minimal case this can be integer round off and in more severe cases channel clipping (which ends up shifting colors). Since virtually any processing will likely result in integer rounding, then a subsequent conversion to TIFF will have some, perhaps imperceptible, loss.

Eric, thanks for the follow-up. The part I don't understand is the clipping. Clipping occurs when the data "space", in other words the bit-depth is overdriven, meaning the data is literally clipped off. If you have a array with data in it, but export that array of data into a different format with the same dynamic range (bit-depth) without changing anything, there should be no clipping. Clipping occurs only if you either push the exposure up (and hence, values beyond the 100% value) and then save the file destructively (i.e. export the file to a different format, applying the exposure shift), or if you downsample the data into a smaller space, like convert it to a 8-bit-per-channel format. Then of course, it has to be clipped off to fit into the smaller data space, or it could be compressed into the 8-bit space, but then you lose the tonal differentiation, because between 0 and 1 you now only have 256 steps instead of 16384 (14-bit RAW).

Clipping is the stuff that you can see in a histogram when you push the exposure and values get pushed beyond "full white" for example. But if you simply do nothing to the 14-bit depth data and just port it to another format, the dynamic range is still 14 bits, or well even 16. So no clipping, nothing cut off, nothing lost.

Or where is my mistake in the logic?

Chris. I can certainly see that if during processing you move the value any of the channels associated with a pixel up or down to clipping, then you can't recover. For example, I can easily see processing a RAW image to give me deeper blacks. If this change is saved as an instruction on top of a RAW image, then the data is still there. However if the image is flattened into a single TIFF, then the shadow data might be lost (clipped to 0). Same thing can happen when pushing saturation values for an effect. In other words, I can easily see how subsequent processing can induce clipping (high or low) into some pixels. When that happens data is lost when the image is flattened.

On the other hand, if you started with a unprocessed RAW and saved it immediately into a TIFF, then it seems to me that all that has been done is to convert the Bayer representation of the data into an RGB representation of the data. I don't think this transformation can induce clipping of the data, so it should be completely reversible and hence no data would be lost.

R8R, thanks for your post!

R8R said:
You might not be able to recover highlights and shadows in the same way. (if at all). Overall contrast and brightness? Yes. Recovery of an image using all that original dynamic range? Limited at best.

See, and this is what I don't understand. Why? What happens to "all that dynamic range", in other words: bits? We're talking about a 14bit/channel to 16bit/channel conversion here, with plenty of room for all that information.

Thanks,
Chris

So just to add some confusion here - Nikon's NEF raw format is based, at least partially, on the TIFF format.

:)

A raw file contains the image data, plus a file header, EXIF, sensor metadata, a small image thumbnail, a jpeg image preview, etc etc.

All of that extra data is "tagged" with any adjustments, and is applied to the raw image data and then your viewer (LR, etc) displays the image with the adjustments applied.

Once you export to a TIF, (or any other finished file format) you are essentially making a "print" from a "negative". You are fixing the white balance settings, sharpness, contrast, highlight adjustments, shadow recovery, etc. All those sliders in LR.

You can adjust these things again with a finished file in an editing app, but you will be working on a limited range of adjustment. You might not be able to recover highlights and shadows in the same way. (if at all). Overall contrast and brightness? Yes. Recovery of an image using all that original dynamic range? Limited at best.

Gareth said:
A RAW file is not an image. It contains more information than can be seen in one image.

Well yeah, it is an image. It's a file that contains image data just like any other image file. There can't be more information in an image than defined by its bit depth, so it's an image file with 42 bit color depth. And of course, the full 14bit per channel can't be seen on a screen or printed out. Neither can 8bit, I'd say. But that's a different topic.

Gareth said:
So if you delete those RAWs you will be throwing away information. But what's the chance that you would ever go back to an old image and rework it from scratch from the RAW?

The chances are high that I want or need to re-edit an image that I've already retouched in Photoshop. I'm not talking about pushing some sliders, I'm talking about heavy retouching as for fashion or beauty. And of course if you re-work something, it's the best idea to have the original data and not some already processed image.

But the original point of the question was that I was wondering whether a converted image would still contain all the information of the RAW file, not theoretically (which is obvious, since 16 bit is more than 14), but practically (just to make the point, a RAW file could be downsampled to a 4-bit-per-channel JPEG quality 0.3 and then handed over to Photoshop to be converted to a 16-bit PSD).

But from everyone's comments, it looks like we all believe that everything is preserved and can hence be processed like the original RAW.

Thanks for all your input,
Chris

chris_weinert said:Now, when you use Aperture or Lightroom or well PS to convert the RAW file into a 16-bit TIFF, is there any loss of this information of the image?

Yes.

A RAW file is not an image. It contains more information than can be seen in one image. When you look at a RAW in LR (for example) you are looking at what the RAW would look like when the settings selected are applied and it is converted to an image (tiff, psd, jpg, etc). That is why RAW editing is non detructive. You are not editing an image, you are choosing what settings to apply to the RAW then exporting to make an image.

HDR works the same way. There is too much information in the file to be shown at one time. When you apply settings you are choosing to show some of the information and hide other information. This is called tone mapping.

When you export to the final image the information that you have chosen not to show is discarded. Only the settings you chose to keep are retained.

So if you delete those RAWs you will be throwing away information. But what's the chance that you would ever go back to an old image and rework it from scratch from the RAW?

On a side note, I often use TIF. It has many advantages. One is that you can save as lossless compressed. Another is that you can open it without Photoshop, and another is you can retain layers. I save all of my final print ready images as TIF, either to print myself, or to send elsewhere.

chris_weinert said:
Eric, thanks very much for the follow-up! What I don't understand:

Ok, the 16-bit setting is per channel, resulting in 48 bits for the pixel, so the array is larger than with the original RAW. Are you suspecting that the "upscaling" or re-mapping produces artifacts, similar to... dang, what do you call that in English... you know, using levels adjustments to enhance contrast?

You shouldn't see a difference but if you upscale it adds nothing at all and could introduce artifacts but unlikely. The less you change formats, color spaces, bit depth, the better in your workflow for a good result.

The only time that it could be advantages is if you create artwork with photos (vector drawings, 3d art, etc.)

msmoto said:
My D4 does the same thing. And, the noise is what is the downside. But, the effective ISO in pulling the shadows up is near 800-1600 in your shot.

Noise is the downside? Excuse me, but this is an almost-completely black area of an exposure, it's not a "shadow" in the more classical sense. I think this is pretty much magic. I don't know what other sensors are doing, but I remember from my 5D Mark II that it was *way* different!

It actually does remind me of film: When I started developing film myself back in the analog days, I was always amazed of the incredible detail that you could get out of a seemingly completely black area that was underexposed. I can't tell whether digital sensors have gotten there yet, but I'd doubt it.

golf007sd said:
@chris_weinert: Off Topic: That is a lot of wires...possible fire hazard if you ask me.

Haha, thanks for the hint. The picture actually only shows 30% of the cable chaos, of which 80% of is audio as well as signal cables (USB, MIDI, FW etc.). What would you advise?

Best,
Chris

@chris_weinert: Off Topic: That is a lot of wires...possible fire hazard if you ask me.

Yup, we could never even attempt this magic tricks with film. My D4 does the same thing. And, the noise is what is the downside. But, the effective ISO in pulling the shadows up is near 800-1600 in your shot. Actually, this is how one gets the Hi-1, 2,etc. ISO from the bodies. The camera is actually exposing with a native sensitivity and then simply lightening up the exposure, as I understand it. So with the extended ISO levels, this shadow detail tends to disappear.

msmoto said:
I am wondering what the results of your examples are and if you could post a couple on this thread, with and without correction?

It's not much different from the luminous landscape stuff, but here's a shot from under my desk... Here's the original:

Here's the processed image:

And here's a crop from the marked area:

Very grainy, yes, but still impressive given the fact that it was almost completely black before adjustments.

Best,
Chris

@chris_weinert - I'm think I was mistaken. At 16 bits / channel I agree that you shouldn't lose anything. I afraid I got my image formats confused and I was thinking that it was 16 bits/pixel, perhaps an indexed color model with a look up table. So long as the RAW pixel data is combined into a similar number of RGB pixels, I don't think that the demosaicing should cause a problem. I agree with you, a 16 bit TIFF really shouldn't have any losses.

Eric, thanks very much for the follow-up! What I don't understand:

Eric said:
As a consequence a raw image has the tonal gradient of a 42 bit color pixel. This can be mapped into a 16 bit color space and you can't see any difference. However, some post processing might result in subtle banding... This would be similar to what happens when post processing jpegs, but not as severe.

Ok, the 16-bit setting is per channel, resulting in 48 bits for the pixel, so the array is larger than with the original RAW. Are you suspecting that the "upscaling" or re-mapping produces artifacts, similar to... dang, what do you call that in English... you know, using levels adjustments to enhance contrast?