Image quality of 12-bit versus 14-bit RAW cameras

I really would like to hear people's thoughts/experiences on image quality in a 12-bit full frame camera sensor versus the 14-bit'ers.

In theory:
- More bits should mean higher tonal richness (more levels) on your images, which should make gradations softer and the image more robust against posterization.
- At the same time, a N-bit linear encoding can hold a maximum dynamic range of 2^N f-stops so more bits could make us expect to be able to capture more DR.

In practice:
- 12-bit RAW files provide nearly 4000 tonal levels for captured information, and moreover interpolated levels are already in the 16-bit range (60000 levels) so a 14 bit RAW doesn't produce a visible improvement in tonal richness.
- Regarding DR, it's limited by saturation in the highlights, and by noise in the shadows. Even if newer 14-bit cameras (40D, D3, A700,...) do have less noise in the shadows than the former ones, the improvement in noise is not enough to really need a 14-bit encoding not to loose DR. I have tested these cameras and they have a practical DR useful to the photographer of 9 f-stops, which can also be encoded in a 12-bit RAW file.

So IMO the conclusion is that today's 14-bit cameras don't offer any practical advantage over the previous 12-bit, just for the 2 extra bits. The improvement comes from other features.

BR

I have to agree here. I don't see any practical advantage on 14bit depth images. There is a theoretical advantage, but in practice it's negligable. Beside that, as _GUI_ said, there's no advantage in DR.

To put it in an analogy made by Michael Reichmann: the dynamic range is the distance between two floors in a building. Let's assume we put a stair between them. If the stair has 2 o 30 steps does not change the distante between floors, but adds for more intermediate steps. This is the bit depth. A 12 bit depth image or a 14 bit depth image have indeed the same DR. One has a theoretical improvement on intermediate steps, but between 12 and 14 bits, the difference -for practical purposes- is not of decisive importance. There are other factors which are more important.

Best regards,
Sebas.

By this logic, there's no advantage to scanning/one shot backs capturing in 16 bit either. No? It's been my understanding that the benefit of higher bit depth captures WAS the gathering of a wider range of exposure values, a longer staircase, not more steps on the staircase, equating with a possibility to record a greater range of highlights or shadows relative to ones choice of exposure. Similarly, I thought that the limitation of the current crop of sensors themselves (glutted photosites) prevented the real extraction of an extra stop or two. Is there no real dynamic range advantage to noninterpolated 16 bit captures?

First, I will clarify that when we say 12-bit or 14-bit, we are talking about the precision of the analog to digital (A/D) converter AND the bit depth of the resulting RAW file.

Using the dynamic range analogy, more bits do not necessarily correspond to a taller staircase-- only the number of steps. The tonal mapping/curve is what determines the height of the staircase.

On the other hand, if the A/D converter is linear (meaning that numerical values directly correspond to actual brightness), then a greater bit depth WILL correspond to a greater A/D dynamic range. Using the above staircase analogy, this is like saying that the steps are equally spaced. However, it is usually image noise -- not the 12/14-bit A/D dynamic range -- that ultimately limits the photo's visible dynamic range. A 14-bit linear A/D converter has a dynamic range of 14 f-stops, for example, but this is far beyond what is visible because of image noise (5-10 f-stops at the very most).

Further, A/D converters and CCD/CMOS sensors are not always linear; in which case the image bit precision/depth does not necessarily correlate with A/D dynamic range. Take a look at the last two sections on bit depth and tonal curve in the dynamic range tutorial. There's a lot more on this there, although be warned that this concept can be quite confusing.

At present the only tangible use I can think of for the extended 14-bit precision is more shadow tonality in a RAW photo which has undergone extreme shadow recovery. If this is true, then 14-bit is worth it in exchange for the marginally larger file sizes. However, I have no evidence of this in practice.

So is the practical conclusion from this that camera manufacturers are mainly using 14- bit as a marketing differentiator?

As a newcomer to DSLR (from film) I am reading technical articles on things such as this, but find myself wondering if the technical improvements are sometimes aimed at people who prefer to look at specifications rather than photographs. I am sure that this is just my cynicism about marketing - is it similar to the marketing battle to try to achieve the megapixel high ground?

Well, it's always good to have some skepticism when reading information. There is a lot of hype put there by the marketing people -not that I have something against them, they're doing their job-. For example, the Canon G9 or Sony H7/H9 cameras are usually found as 'prosumer' cameras, when the reality is that as they can be top examples of P&S digicams, they are not prosumer cameras (ok, we can start discussing what makes for a good photograph and from there question wether this cameras can or not take a good picture, so I'm restricting my argument to the empirical data, meaning sensor size and its relation with perfomance, mainly).

It is good to look at the photographs, but I think there's a middle step included, because the judgement of a picture is often biased by taste and sometimes it can become difficult to judge a camera performance by photographs only, more if we consider that in the digital scenario, cameras are very different than what they were from film (for example, in film if you were to shoot at night, you loaded the appropiate film, and if it didn't perform as expected you simply loaded another roll next time. The camera and lense were not in question and did not need to be changed if they were of good quality, say, a Pentax LX and a 50/1.4 Pentax-A lens. Now, if the camera doesn't perform as expected... well, I think it's clear enough the sadness when you see that in low light at ISO 1600 there's more noise than anything).

So, we should consider the technical information, the marketing hype at moderated (after all, they not always lie, they usually exaggerate), and the photographs in adverse conditions if possible (low light, high ISO, always RAW).

I hold the same position as McQ since I don't have evidence of a performance improvement in 14bit depth images and 14bit A/D conversors. Beyond a theoretical advantage, I don't think that there's a huge improvement unless there are others factors changed as well (which I haven't seen reported). Of course, consider that as I don't have any evidence, I could be incurring in a mistake. My common sense tells me that even if the theoretical improvement would be real, these advantages would not be decisive... perhaps with a couple of exceptions under certain conditions.

Best regards,
Sebas.

An interesting post Sebas, and well reasoned (despite the reference to a castrated sheep :cool: )

Essentially your argument seems to be that the photograph needs to be judged in the context of the technology that facilitated its capture.

I completely agree that judgements of a photograph in the range from awful to sublime, are very much in the eye of the beholder.

Nevertheless I remain if the personal view that the reason we take photographs is to deliver images that we and perhaps others, can admire, aspire to, talk about, and be challenged by either emotionally or intellectually. To me it is all about the end - the technical means for me is merely a tool that needs to be understood so that I can achieve that end.

And in saying that last part fo that last sentence, I see that probably we are of like mind even if we do approach the answer from different angles.