[Spodeworld]

I know that a 14-bit raw has an advantage in shadow area detail, but does it have an advantage in noise, all else being equal?

[Adam]

Theoretically speaking, yes, there should be a difference, since better noise reduction can be achieved if very subtle variations from pixel to pixel are recorded. However, in practice, I think that this difference would be so small that 12bit vs 14bit will yield the same results.

[Lowell Goudge]

I know that a 14-bit raw has an advantage in shadow area detail, but does it have an advantage in noise, all else being equal?

this all depends.

generally with any A-D conversion, there is some noise in the order of perhaps +/-1 least significant bit. In this respect, clearly if the entire issue was to be caused by the A-D then yes 14 bit would have less because the least significant bit represents a smaller change, BUT in many cases the noise we see is the sensor itself, and the impact of things like dark current and the randomness of photons hitting one specific "well" of the sensor. I am not sure where that accuracy and noise level lies, and then of course, there are the actual amplifiers themselves.

[wombat2go]

It looks like typical cameras post 2011 or so have a dynamic range of 8 or 10 F stops based on tolerable sensor noise, at say iso 400
The way I look at it (in Hex ,for a 16 bit ADC which cameras don't have yet)
Full bright = FFFF
Down 10 stops = 00FF
So for a camera with 12 bit adcs, the resolution at sensor noise is still 1 part in 16
Or, put the other way, the noise level is still 16 times the resolution - so lsb jitter should not be an an issue ?

[vodanh1982]

It looks like typical cameras post 2011 or so have a dynamic range of 8 or 10 F stops based on tolerable sensor noise, at say iso 400
The way I look at it (in Hex ,for a 16 bit ADC which cameras don't have yet)
Full bright = FFFF
Down 10 stops = 00FF
So for a camera with 12 bit adcs, the resolution at sensor noise is still 1 part in 16
Or, put the other way, the noise level is still 16 times the resolution - so lsb jitter should not be an an issue ?

Unfortunately, they are not linear.
Are 14-bit Raw Images Really Any Better Than 12-bit Raw? - Photo Tips @ Earthbound Light

[Anvh]

Unfortunately, they are not linear.
Are 14-bit Raw Images Really Any Better Than 12-bit Raw? - Photo Tips @ Earthbound Light

That's already some time ago.
There is a big differnce between the K5 and the D300, i believe K5 was one of the first where there was 14bit was beneficial.
You can compare the K5 with the K30, same sensor, but different process.... but also 14bit vs 12bit file.

[Na Horuk]

Um.. it should, especially if you are increasing brightness during post processing. But it won't be super dramatic. The other thing is that K-5 allows minimal ISO to be 80, while K-30 has minimal ISO 100. So 14 bit at ISO 80 might be less noisy than 12 bit at ISO 100, if you brighten the shadows in post. Though, at those ISO, noise will be practically nonexistent to begin with.

[Spodeworld]

I guess I was asking, all other settings being equal, whether a 14-bit image will yield less noise than a 12-bit one.

Um.. it should, especially if you are increasing brightness during post processing. But it won't be super dramatic. The other thing is that K-5 allows minimal ISO to be 80, while K-30 has minimal ISO 100. So 14 bit at ISO 80 might be less noisy than 12 bit at ISO 100, if you brighten the shadows in post. Though, at those ISO, noise will be practically nonexistent to begin with.

[Spodeworld]

I'm not even sure how to interpret what that means

So for a camera with 12 bit adcs, the resolution at sensor noise is still 1 part in 16
Or, put the other way, the noise level is still 16 times the resolution - so lsb jitter should not be an an issue ?

[top-quark]

14-bit raw files of the Pentax K-5 are definitely noticeably more malleable than the 12-bit raws of the K-r. However, this isn't a comparison of like-with-like. I have no idea whether you can recover equal amounts of information with a K-30 / K-01. 14-bit raw should improve noise performance simply because there's more signal for the same amount of noise. You should be able to see differences if you use techniques like ETTR, since most of the tone information is in the higher bits (so 12-bit should give 4096 tone levels per channel while 14 gives you a theoretical 16384, 12288 of those being contained in those two extra bits). That said, I would be surprised if the sensor were capable of showing that many discrete levels. If anyone has a K-5 and a K-30 and can photograph the same scene using ETTR, we should be able to see whether there are any real-world differences.

[wombat2go]

Spode
What my post meant was, for a reasonably well exposed photo, the black level/noise level is higher than the resolution of a 12 bit sensor.
So the extra resolution of the 14 bit adcs is not yet available with present sensor noise levels.
I note that Pentax uses 12 bit in the k-01 for example.

Vodanj put a useful link about reclaiming dark images below noise levels. In that link I think the camera had a user selectable adc resolution ( 12 ~14 bit) and don't know if the noise reduction algorithms stayed the same in the selection. I think probably the noise filter would be tailored to the resolution selection contributing to the better , but still unusably noisy dark example shown, at 14 bits.

Anyway I just try to expose properly in the initial shot or discard them, and I never tried to reclaim such an underexposed shot.

[wombat2go]

f anyone has a K-5 and a K-30 and can photograph the same scene using ETTR, we should be able to see whether there are any real-world differences.

That is a good idea, one test is worth 1000 words.

At present we have the monitors and printers with standard 8 bit sRGB so that is a further degradation. But the focus on the forums seems to be on theoretical camera performance rather than all the steps in the chain to a viewable image.

[Class A]

There aren't many ISO levels at which 14bit are an actual advantage, but certainly at ISO 80 they are useful.

Have a look at this comparison by Guillermo Luijk.

Also bear in mind that Pentax had to choose between adding better video and focus peaking to the K-5 II(s) by using the K-30 image processor, or staying with the 14-bit K-5 image processor. I personally am glad that they chose to keep the K-5 processor.

[Heie]

I have both a 12-bit (K-30) and a 14-bit (K-5 IIs), so I thought I would offer some help. The immediate thing that came to mind was the sharpness difference of the two could have been an issue with the perceived noise. So then I came up with the following idea - take a shot of the same scene with the same lens on both cameras, and give 100% crops of the out of focus areas. This is my dead K-5 (waiting for Jenny's 5 year old brother to have his birthday so I can give it to him - he loves cameras and loves playing with them. With it comes a $10 manual lens that lets him look through the viewfinder and twist the focus ring ) on my bed and some sunlight coming through the window. The K-5 IIs is a hair out of focus because I "focused and recomposed" on both shots, but apparently a bit too much on the K-5 IIs. It's my fault, not that of the camera. Anyway, here they are. Top row is resized to 795x1200

FA 77 Limited, 1/2000s, f/1.8, ISO 8000


The crop shows highlights, midtones, and severe shadows/darks.


I don't see any difference.

-Heie

[Class A]

FA 77 Limited, 1/2000s, f/1.8, ISO 8000

At ISO 8000 you surely won't see a difference.

GordonBGood (posting at DPRreview) once wrote that he expects ISO 640 to be the threshold.