[willeisner]

Hello Pentaxian

Reading about future Pentax full frame and comments on MP camera may have, I wonder another question, the maximum bit
Pentax cameras are 14-bits processing both APS-C and medium format.
There are some medium format cameras that are in 16-bits processing.
Pentax could be the first 35mm camera with 16-bits?

[MadMathMind]

The sensor has to be 16 bit. Pentax doesn't make their own sensors. They buy them from Sony (and sometimes Toshiba; in the past, it was Samsung). If Sony doesn't make it, Pentax won't have it.

[willeisner]

I think the image processing 16-bit or 14 is connected to "capacity" of the company and not with the chip maker, we have the K5 series that uses 14-bit and k30 the same sensor with 12-bit.
We have cameras hasselblad with the same sensor that 645z features 16 bitz
against 14 Z

[disconnekt]

I read an article about this (can't remember what magazine though) not that long ago. From what I remember, the author said something to the affect that our eyes really won't notice the difference between the 2 bit jump (from 14 to 16), compared from a 12 to 14 jump. And another factor may be price of adding to or changing their hardware.

[Adam]

Hello Pentaxian

Reading about future Pentax full frame and comments on MP camera may have, I wonder another question, the maximum bit
Pentax cameras are 14-bits processing both APS-C and medium format.
There are some medium format cameras that are in 16-bits processing.
Pentax could be the first 35mm camera with 16-bits?

14 bit is more or less an industry standard for RAW files, so I'd expect the Pentax FF to follow it.

[willeisner]

Probably, under normal conditions it is not possible to differentiate 14 or 16 bits
but in extreme conditions, over and and underexposure considering files lossless and lossy have significant difference, when you retrieve the photo in any image editor.

[RuiC]

A jump from 14 to 16 bits has its benefits on the end quality of the image, I believe it. But it has its costs mainly on processing power which leads to more heat produced, higher power batteries, higher costs and space.

[biz-engineer]

16 bits makes sense for large pixels (rem. CMOS cell johnson noise = k.T/C [ http://en.wikipedia.org/wiki/Johnson%E2%80%93Nyquist_noise ] ). For exemple , K-3 has a 14 conversion, if we compare the SNR at 800 iso / -2ev and 3200 iso / 0ev, we gain almost nothing at 3200 iso, which means that the quantization noise of the 14 bits conversion is already smaller than the noise of the CMOS cell at 800 ISO. Except being a marketing feature, having 16 bits on a 36Mpixels FF sensor instead of 14 bits would be no benefit. In addition, increasing resolution from 14bits to 16bits is achieved by noise shaping (integration+digital decimation), which everything else remaining the same would mean a slower data acquisition, = slower frame rate. So, on medium format OR full frame 12 Mpixels, it makes sense to have a 16 ADC, but it does not bring any benefit for other cases.

[normhead]

People can't tell the difference between 12 bit and 14 bit.... so it doesn't make any difference because (wait for it).. it doesn't make any difference discernible .) There is simply no empirical evidence that it does. And without empirical evidence backing it up, it's a notion, not a theory.

But it is amazing how many notions get called theories on this site.

The important question, is not "is their a theoretical difference". The question is "Given the limits of human perception, does it make any difference to people." Most of us are still using 8 bit printers and monitors so what possible difference could going 16 bit make to us?

[wombat2go]

As I read, some monitors are only 6 bit with hardware (in the monitor) dithering to 8 bit.

Better monitors are 8 bit with dithering to 10 bit by a LUT in the monitor.
I have one of those, and it does provide a better image when viewing a 16 bit tif compared to an 8 bit jpeg. But I am not sure how it does that over a dvi link.

The monitors with Trucolor work over the DisplayPort link and can run at least 30 bits ( 3 * 10 bits).
I have just set up a new Eizo with Nvidia Quadro for DisplayPort and driver capable of 48 bits and I have set the linux o/s Xorg to render 48 bits.
I am still learning about it . Yesterday I saw that film scanner banding artifacts appear a a lower level on 10 bit monitor compared to 8 bit.

Also I think there is an interaction between bit depth and spatial pixel resolution.
Most of the time, the video sent to a monitor is either an interpolation, or a downsampling of the image file.

So our dslrs are at least 12 bits, which puts them at 64 times the resolution of a TrueColor monitor and gives a bit to play with with backgrong processing in rendering to a monitor and also in pp if we are so inclined.

The tif file can go up to 32 bit floating point. I suppose this is needed for applications in science and astronomy. It is also useful for pp like fft transformation.

[DeadJohn]

...The question is "Given the limits of human perception, does it make any difference to people." Most of us are still using 8 bit printers and monitors so what possible difference could going 16 bit make to us?

Astrophotographers who need to do more extreme processing might benefit from 16 bits. The final output will be displayed in less than 16 bits but the extra bits of data can help uncover faint details during processing. The types of astrophotos that most benefit from 16 bits are probably being done with dedicated, cooled CCD cameras anyway rather than with DSLRs.

[RGlasel]

Sensors don't directly read light intensity with anything close to 16, 14, or even 12 bits of resolution. The camera's processor calculates 12 or 14 bits of information based on the light meter output as well as the signal from the sensor. The camera designers have to balance EV range with noise within the physical capabilities of the sensor. The cost of obtaining 16 bit RAW files from the camera in terms of reduced EV range and increased noise will probably outweigh the imperceptible value of increased bit depth.

[wombat2go]

Sensors don't directly read light intensity with anything close to 16, 14, or even 12 bits of resolution

The ADC converts the Bayer sensors' mos well voltages (integrated charge).
A 12 bit ADC has a quantization dynamic range of 83dB.
The usable DR of the well depends on its noise floor and saturation voltage,
and also on the various tricks the manufacturers are developing like partial reset wide dynamic range sensors etc.

I expect the DR will keep on increasing.

[biz-engineer]

There's a line amplifier in front of the ADC, that's why if you use ISO6400 you get let noise than if you use ISO100 underexposed by 7 stops, and that's where it's better to have a 14 bits ADC rather than a 12 bits ADC. From 12 bits to 14 bits won"t help to get a better DR, but it will allow a cleaner higher ISO settings.

People can't tell the difference between 12 bit and 14 bit.... so it doesn't make any difference because (wait for it).. it doesn't make any difference discernible .) There is simply no empirical evidence that it does. And without empirical evidence backing it up, it's a notion, not a theory. But it is amazing how many notions get called theories on this site. The important question, is not "is their a theoretical difference". The question is "Given the limits of human perception, does it make any difference to people." Most of us are still using 8 bit printers and monitors so what possible difference could going 16 bit make to us?

normhead, I see that you have a K5 and K3, and you are not interested in your extra bits, either when you upgrade a camera, you also have to upgrade your glasses, or you could do a good action by giving the bits you don't use to people who need them more :-))))

[jonby]

There's a line amplifier in front of the ADC, that's why if you use ISO6400 you get let noise than if you use ISO100 underexposed by 7 stops, and that's where it's better to have a 14 bits ADC rather than a 12 bits ADC. From 12 bits to 14 bits won"t help to get a better DR, but it will allow a cleaner higher ISO settings.

I'm sure that DPR did a test where they came to the conclusion that with a particular camera (think it was a Nikon D810), more or less the same shadow information was captured whether shooting at high ISO, or underexposing at low ISO and lifting everything in post. From what I remember, apart from minor differences in colour balance and tonality, largely down to how they had processed it, there was very little difference in the output from an ISO100 shot, underexposed by say 6 stops, and boosted in post, and an ISO6400 shot as is. It may not have been a D810, but it was definitely a Nikon with the Sony sensor. They compared it to a Canon model - think it was the 5DIII - which was a completely different story and there was huge benefit to using the high ISO.

This seems at odds to what you are saying, at least for some cameras/sensors.

I would provide a link but I can't find it now. Anyone remember this? It wasn't that long ago.