[Thomasbrowphoto]

I would love to see Pentax update the 645Z firmware to support 16bit colour like the Phase and Hasselblad have done with the same sensor. I'm unsure if it can be done by just firmware or if it requires heavier hardware. 16bit colour is nice. Im sure I'm not the only one that would want this?

[Adam]

I think it's a hardware thing. We'll see how the medium format line evolves- I'd bet that the next model will be a big improvement, just like the 645Z was compared to the 645D.

[interested_observer]

You have to look at the entire image processing chain. Here is a block diagram of the overall process.

The light hits the sensor, which when the shutter is tripped the sensor is read (pixel by pixel), the ISO as an amplification is applied (as an analog signal), which then goes through an Analog to Digital Converter (ADC). This is a piece of hardware, that quantizes the analog signal into a set of digital bits. The ADC has a specific output (8, 10 12, 14, 16, ... etc.) bits wide. Then the now digital word that represents that pixel in terms of how much light it collected is passed to the digital signal processor and the general processor for further processing and saving to the SD card.

The ADC is done in hardware due to the speed required. To do it in software (which would still use hardware for a large part of the operation) would be cumbersome, slow, and just consume a lot of power. So, no it would not be a change that could be done in firmware.

Here is a simplified high level block diagram for a digital camera.

When you think about it, you are taking a 16MP sensor - 16 million pixels (or what ever size - 50MP is even worse), taking each one, processing it into a image, and then storing it. Doing this 7 or so times a second. That's a lot of bits flying around. Then there is the battery that powers the whole shebang for about 600 images at a wack. It really is a wondrous Ruth Goldberg contraption - that you can hold in your hand....

[joergens.mi]

That can only be done in hardware. And it is not trivial. The resolution is increased by a factor of 4 and all components up to the ADC including the pcb must be in a quality to cope with that. If the full range of the ADC is 0 to 1 Volt the whole System must be able to sense Signals below 0.01526 mV with low noise.

[biz-engineer]

That can only be done in hardware. And it is not trivial. The resolution is increased by a factor of 4 and all components up to the ADC including the pcb must be in a quality to cope with that. If the full range of the ADC is 0 to 1 Volt the whole System must be able to sense Signals below 0.01526 mV with low noise.

Already 14bits effective is not trivial to achieve, but at low frequencies, increased resolution is achieved by oversampling and noise shaping (all 20bits/24bits audio conversion is done this way via sigma delta conversion). Increasing the ADC resolution takes a toll on frame per seconds because the of amount noise that has to be pushed out of band by oversampling, and the bottleneck of clocking speed limited by propagation over the sensor, already clocked at 400MHz which itself generate noise. If the light sensitive cell itself can't achieve 14bits of noise level, 16bits is a waste. If you go to 16bits depth, you have to use larger pixels and for a given clock speed, you lower the number of frames per second. So, you can be happy that you already own the best trade-off, already decided by engineers for your benefit.

[CDW]

Already 14bits effective is not trivial to achieve, but at low frequencies, increased resolution is achieved by oversampling and noise shaping (all 20bits/24bits audio conversion is done this way via sigma delta conversion). Increasing the ADC resolution takes a toll on frame per seconds because the of amount noise that has to be pushed out of band by oversampling, and the bottleneck of clocking speed limited by propagation over the sensor, already clocked at 400MHz which itself generate noise. If the light sensitive cell itself can't achieve 14bits of noise level, 16bits is a waste. If you go to 16bits depth, you have to use larger pixels and for a given clock speed, you lower the number of frames per second. So, you can be happy that you already own the best trade-off, already decided by engineers for your benefit.

Agreed. The 14 vs 16 bit issue has been discussed to death on photography forums. The extra two bits buy you nothing that translates to visibly improved print quality. Phase uses the same sensor in their IQ250 and it is also a 14 bit system.

Phase One announces IQ250 50MP CMOS medium format back: Digital Photography Review

[interested_observer]

A recent thread on 12 and 14 bit...

• https://www.pentaxforums.com/forums/173-general-photography/318180-14-bit-vs-12-bit-raw.html

I have to agree with Biz. Going to 16 bit does nothing unless the entire system was designed with that in mind - starting with the sensor is able to utilize the additional resolution. If the sensor is unable to resolve to 16 bit values, having a 16 bit A/D converter does not buy you a lot - other than bragging and marketing rights.

[Kolor-Pikker]

A recent thread on 12 and 14 bit...

• https://www.pentaxforums.com/forums/173-general-photography/318180-14-bit-vs-12-bit-raw.html

I have to agree with Biz. Going to 16 bit does nothing unless the entire system was designed with that in mind - starting with the sensor is able to utilize the additional resolution. If the sensor is unable to resolve to 16 bit values, having a 16 bit A/D converter does not buy you a lot - other than bragging and marketing rights.

The sensor itself does not have bits, as it's a fully analog component, but it does posses a S/N ratio that can be quantized into digital information via the ADC. If the sensor is sufficiently noisy, then even if you have a 14-bit+ ADC, you'll simply be digitizing more of the noise floor rather than obtaining more meaningful information. Pure white is the saturation point of a sensor, so all additional dynamic range extends into the shadows, therefore the noise floor dictates dynamic range.

But you are right, the system as a whole needs sufficiently low noise, a high-depth ADC, and encoder in order for 16-bit to make sense. Few sensors can actually break past 14 stops worth of signal, and there are no 16-bit ADCs that are usable for high-end imaging (to my knowledge), so any camera that claims 16-bits simply wraps the data in a 16-bit container, which does potentially offer benefits for hardware-level processing, but not for actual image quality.

It is actually possible for a sensor to posses more SNR than its hardware is capable of quantizing, it simply means that the noise will be below the visible threshold. In the case of the 645Z you have to really crank the shadows just to grab a look at the noise floor, so I'm guessing that the SNR is actually greater than what can be described in 14-bit, so this sensor could actually benefit from 16 bits, should the technology to reach that deep exist.