This content has been marked as final. Show 24 replies
I believe you have to tell ACR to open them as 16-bit in to Photoshop. All
raw images are 16-bit so the D300 isn't doing anything that any other raw
supporting camera would do. But, by default raw images open as 8-bit in to
Photoshop. At the bottom of the ACR window under the image there is a string
of text that includes the size of the image, it looks like a web hyperlink.
Click on this and set your option for 16-bit.
Robert, that was a very quick and helpful response. took care of my problem. thank you very much.
>I am downloading 16 bit nefs (D300) from card reader with cs4 bridge.
You were probably talking about the bit depth of the gamma corrected rendering space in ACR but just so you know, the D300 can only output 12 or 14 bit NEFs. No current Nikon can give you files with more than 14 bits. There is no advantage in using 14 bit NEFs on the D300.
Happy to help. Happy shooting (photos that is!)
I have to take issue with your statement that there is no advantage to shooting at 14-bits instead of twelve.
Your own test images show the issue only too clearly: NOISE!! And the subsequent lack of definition.
I frequently use very high ISO settings on a D3 and ALWAYS shoot RAW 14-bit .nefs.
With today's large capacity CF cards, why would anyone even consider shooting at other than the full bit-depth of which their camera is capable?
>I have to take issue with your statement that there is no advantage to shooting at 14-bits instead of twelve.
On the D300 there is no difference between 12 and 14 bit capture. My images show that very clearly and knowing the physics this makes perfect sense. The noise is exactly the same in both 14 bit and 12 bit capture and there is no difference in detail in the two images. On a D3 at low ISO, it might matter as it has a far lower noise floor than a D300 (more than two stops better according to dxomark.com ). I have never tested it on a D3, so I cannot be sure. The Dxomark results suggest it might matter for ISO < 200.
>I frequently use very high ISO settings on a D3 and ALWAYS shoot RAW 14-bit .nefs.
You're wasting your card space. 14 bit will only matter at low ISOs where the noise floor can get near the least significant bits. At higher ISOs, the noise is so large that the lowest bits, even in a 12 bit file, are swamped completely and there is no point to 14 bit capture. This is just basic statistics and physics.
Look at the difference between 12- and 14-bits In Noise, image detail and the rough gradients in your OWN test images!
They clearly illustrate exactly what I am saying.
At 6,400 ISO, I do not need to exacerbate Noise although there is very little on a D3
And with 32 GB of space available in loaded cards at any time, I am hardly going to be concerned with "wasting card-space"!
Try a D3 sometime if you can.
>Look at the difference between 12- and 14-bits In Noise, image detail and the rough gradients in your OWN test images!
? You must be looking at different images. The messy garage? There is absolutely no difference. The 14 bits image actually looks slightly worse from a detail perspective but noise-wise and gradient wise they are identical. I don't know what you are seeing.
>At 6,400 ISO, I do not need to exacerbate Noise although there is very little on a D3
A D3 at ISO 6400 has 8 stops dynamic range, which means that it is physically impossible that 14 bits would make a difference as the noise floor is 4 stops above where the 12 bit file would cut off. The dynamic range needs to get near 11 or 12 stops before 14 bits can make a difference. On the D3 that happens at ISO 800. The D300 doesn't get there but for below ISO 200. This means that above ISO 800, the D3 cannot capture with low enough noise for 14 bits to matter and the D300 probably has no range where 14 bit matters. I have yet to see an image (and I tried a lot of different situations, not just my messy garage) where 14 bits matters one bit (pun intended).
>And with 32 GB of space available in loaded cards at any time, I am hardly going to be concerned with "wasting card-space"!
Not just card space, but also disk space, backup space, etc. If you shoot 100's or thousands of images a day, you're going to notice this quite quickly. But it's up completely up to you. I guess it is my frugal Calvinist upbringing that makes it hard for me to waste space when there is no conceivable difference whatsoever.
>Try a D3 sometime if you can.
I have played with it in the shop. Amazing camera! The high ISO performance is absolutely astonishing. I just haven't looked at 14 bit capture on it. I'll try it when I get around to it. I have heard it matters on the D3, but I would like to see for myself.
>? You must be looking at different images.
It seems that I was!
I had clicked on the link on your page and didn't notice that i had been transported to a different Site:
It was the images (with roll-overs) on that site which illustrated the difference between 12- and 14-bit capture so clearly.
>It was the images (with roll-overs) on that site which illustrated the difference between 12- and 14-bit capture so clearly.
Yeah, they show some difference but apart from those (slightly contrived) examples I have not seen any real world differences in natural scenes and I have not been able to reproduce his results even with test charts and using camera RAW and Lightroom. At ISO 100 and 200, there is potentially some difference on the D300, so it could be real. I just don't see it in my own testing.
>On the D300 there is no difference between 12 and 14 bit capture.
This statement is incorrect. The D300 has lower read noise at 14 bits, not because of the increased bit depth but due to the fact the frame rate is limited with 14 bits. Read noise is inversely correlated with read out rate
Show me the data! In actual tests it is just not there.
>Show me the data! In actual tests it is just not there.
I don't have data, but was relying on this analysis by Emil Martinec, who is quite knowledgeable in these matters. Look at the next to the last paragraph in the linked section of the paper. You may be correct
Are you setting up your workflow in Camera Raw 5.2. It is in the middle of the page at the very bottom (usually in blue) I have mine set up to: (my down loads are from a D200)
ProPhoto, 16 bit, 3872 x 2529, 300 dpi
I see you have said that they are nef files, but have you got some preference to change them to jpeg files as these type of files can only be 8 bit.
I just tried ACR 5.2 and opened in CS4 and they are all 16 bit files as ProPhoto exactly as to what I have set in my work flow of Camera Raw.
>Look at the next to the last paragraph in the linked section of the paper. You may be correct
very interesting paper! thanks. I looked at 12 vs 14 bit at ISO 6400 too yesterday evening, and I canot see any difference as you would expect. Any difference you se at 100% is the same as between two shots at identical settings and I wouldn't be able to pick out the 14-bit image in a double-blind comparison.
Remember that Emil's notes are regarding the read noise differences between 12-bit and 14-bit, which only shows up in the very deepest shadows. You would not expect to see any differences higher than that (e.g., quarter tones or midtones), unless the amplifier's frequency characteristics are fundamentally different (unlikely).
> Remember that Emil's notes are regarding the read noise differences between 12-bit and 14-bit, which only shows up in the very deepest shadows. You would not expect to see any differences higher than that (e.g., quarter tones or midtones),
Eric and Jao,
When I made my post above, I was relying on memory and I quoted Emil's post and provided the link without carefully re-reading the notes. My quote was somewhat out of context, since Emil's thesis was that there is little if any advantage of 14 bits over 12 with the D300 (and likely with the D3) because of noise as Jao has noted.
Emil postulated that with the D300 any improvement at 14 bits would be mainly from the lower read rate at that bit depth rather than the increased number of bits. As Eric notes, lower read noise would show up only in the deepest shadows.
I've been following this with interest, as I use Nikon 14-bit raw on a D300 - mainly because it seems to be the right thing to do, and not because I can see a significant difference.
I'm having trouble understanding the jargon though. Are you saying that the sensor data is read more slowly in 14-bit, which reduces the noise floor? and that is is only of any benefit when used at the lowest noise ISO (ie 200)?
Is there no other benefit to post-processing? I'm not clear where the extra 2 bits go; is it increased resolution over the same dynamic range, or is the dynamic range extended? Is this not also a factor in choosing 14 over 12?
Sorry if this is a bit off-topic, but you have to grab experts when you find some ;-)
>I'm having trouble understanding the jargon though. Are you saying that the sensor data is read more slowly in 14-bit, which reduces the noise floor? and that is is only of any benefit when used at the lowest noise ISO (ie 200)?
Yeah that is basically the idea. A different read-out path or set of amplifiers is used or it is simply "read slower" which can, in principle, reduce noise. Not something that is inherent to 14 vs 12 bits but more to the way the data is read. If the noise stayed the same and it is in the lower 2 bits of the 12 bits converter, there is no point as the only thing you're doing is more precisely imaging the noise. If however, the noise goes down a little going to the 14-bit readout method, you might get better signal to noise ratios and hence more shadow detail. This is all theory, and if you actually do the test with your camera, you'll see there is no difference in real world photos. The shadows would need to be really deep to see any of this. Basically you need to be boosting exposure by 3 or 4 stops AND applying shadow fill near 100. Not a realistic situation. To see it, you indeed need to be at a situation where the noise floor is significantly low, which means low ISOs on most cameras. At high ISO, there could be a difference because of differences in amplifiers, etc, but in practice that is not the case either on this camera. Other sources of noise drown it out completely.
>Is there no other benefit to post-processing? I'm not clear where the extra 2 bits go; is it increased resolution over the same dynamic range, or is the dynamic range extended? Is this not also a factor in choosing 14 over 12?
There could be some benefit if the noise characteristics were significantly different. Lower noise means higher dynamic range and therefore better capability to extract detail from shadow areas. However, from my testing, in practice this does not happen on the D300. Other cameras might well show higher dynamic range and better detail in shadows at 14 bits. You can see some of this by looking at the dynamic Range charts on DxOmark.com where sensors of different cameras are tested. You need to be significantly above 11 to 12 EV for 14 bit capture to make a difference.
>Sorry if this is a bit off-topic, but you have to grab experts when you find some
The easiest thing is to test it. In realistic scenarios, I just cannot see any difference. If you contrive a completely weird case, such as underexposing an entire image by 8 stops, or pushing up shadows by as many stops, you can get into the range where it might make a difference. Those are just not real situations though.
The most likely real-world situation where you would see a difference, if it exists, is for an accidentally severely underexposed image which you are attempting to salvage by digging up image detail out of the deepest shadows.
It is highly unlikely that you would see a difference in the midtones (e.g., near L* = 50), even if the difference existed.
> The D300 has lower read noise at 14 bits, not because of the increased bit depth but due to the fact the frame rate is limited with 14 bits
The read noise with 14bit is exactly four times greater than with 12bit. I am in favour of 14bit recording, but for the higher number of levels, not for the lower noise.
I've just done a bit of surfing on the subject, and I'm still convinced that 14 bit has got to be preferred over 12 bit - not necessarily for any reasons of noise, but purely for the same reasons why 16 bit PSDs are better than 8 bit - that is, the scope for extensive processing with minimum quantisation error provided by increased digital resolution.
If I understand correctly, the lower half of the dynamic range sees the best improvement in resolution (as there's relatively a lot more in the top half) - which has got to be a bonus when hitting a raw image with a series of mind-bending curves and adjustments in Camera Raw.
I wouldn't have expected noise to be a big consideration in choice of bit depth, unless it's the finer quantisation of noise.
> A different read-out path or set of amplifiers is used or it is simply "read slower" which can, in principle, reduce noise
I doubt that. Looking at the 12bit raw data very closely reveals, that it is numerically derived from the 14bit data. Note, that for the red and green channels this is not simply a division by four (a bit shift). This means, that the readout is always in 14bit, and the 12bit version is simply "dumbed down".
>I doubt that. Looking at the 12bit raw data very closely reveals, that it is numerically derived from the 14bit data.
Yeah I agree, that's why I posted it as a hypothetical. I see no difference on my D300, just slower operation and a slight difference in luminance in the very lowest tiers which might be coming from ACR instead of the actual file.
>I'm still convinced that 14 bit has got to be preferred over 12 bit - not necessarily for any reasons of noise, but purely for the same reasons why 16 bit PSDs are better than 8 bit - that is, the scope for extensive processing with minimum quantisation error provided by increased digital resolution
That is not a correct analogy. Raw data is generally linear, where the number you see is linearly related to the light intensity hitting the photosite. In a typical 12-bit RAW file from a typical camera, at typical ISOs the two (or more) least significant bits are always just pure noise, no matter where on the luminance curve you are. At higher luminances, your absolute magnitude of noise is higher because of photon-counting statistics. So if your noise floor is say two bits high in 12 bit precision, there is no benefit whatsoever to go higher bit depth, since you will simply image the noise at a higher precision whatever luminance you are looking at. In a tiff file, gamma compression is applied. This means that at high luminances, there is less precision than at low luminances, even to the point that the bit levels surpass the noise, and you can get posterization in midtones and highlights when using low bit depth because of this. RAW does not have that same issue as usually (say you're not using a D3), the bit steps are always smaller than the noise.
But really, this is something that is very easy to test. Just take a superhigh contrast scene (say shooting an object in profile against a sunset, or a bright summer day with very harsh shadows) and shoot the same scene in 12 and 14 bits with exactly the same settings exposing for highlights. Then compare how much you can restore in the shadows from the two files. I did this test with my D300 and there is absolutely no difference, whatever ISO you try. It just doesn't have the dynamic range. In real photography, there really is not much point to it. Now if you underexpose 6 stops or so on some really low-noise cameras such as a D3 and want to pull out the shadows in that underexposed file there is a potential benefit but not on most high megapixel crop-sensor DSLRs. You don't have to believe me, just test it yourself.