DR, DxO, and the mysterious case of the binary camera

[32BT]

So, I have this great new piece of technology. It's an extremely high resolution camera. It can capture a gazillion giga pixels. Moreover, it can capture nano-second exposures. It does however only record 2 levels for each sensel. Either fully black, or fully white.

I point it at the sun and snap a picture. I get a perfectly white circle in a sea of noise-free black. Ooh, I just captured a dynamic range of maybe more than 1000k to 1.

DxO gets their hands on the same piece of equipment. Does anybody have a clue what then will be their final judgement on dynamic range?

[bjanes]

So, I have this great new piece of technology. It's an extremely high resolution camera. It can capture a gazillion giga pixels. Moreover, it can capture nano-second exposures. It does however only record 2 levels for each sensel. Either fully black, or fully white.

I point it at the sun and snap a picture. I get a perfectly white circle in a sea of noise-free black. Ooh, I just captured a dynamic range of maybe more than 1000k to 1.

DxO gets their hands on the same piece of equipment. Does anybody have a clue what then will be their final judgement on dynamic range?

The dynamic range would be high, but the tonal range and color sensitivity would be very low. The DXO score would be very low. So what is your point?

Bill

[Tony Jay]

He's just trying to get his head around how the scoring works.

Regards

Tony Jay

[Bart_van_der_Wolf]

Does anybody have a clue what then will be their final judgement on dynamic range?

Depends on the amount of binning (or a sampling over time), so it's variable. Also a solid white disk means overexposure, you should aim for some of the sensels staying black, and some white. Since that would have a Poisson Noise distribution, the optimal level of exposure would be based on an exposure level metering that avoids clipping the upper tail of that distribution. The new camera will thus also use a new exposure metering system, but you knew that . The read noise will then determine the Dynamic range for that particular amount of binning.

The visionary imaging chip was described by Dr. Eric Fossum as a Quanta Image Sensor (QIS).

Cheers,
Bart

[PierreVandevenne]

Ooh, I just captured a dynamic range of maybe more than 1000k to 1.

The dynamic range would be high

Hmmm, would it, really?

[EduPerez]

Consider that each photosite in a sensor is a mere photon counter. When you increase the resolution, each photosite receives a smaller portion of the light; when you lower the exposure time, each photosite receives less light. At the end, the number of photons that reach a photosite is so low you can count them on a one by one basis; then, lower bit-depth sensors start to make sense.

[32BT]

Depends on the amount of binning (or a sampling over time), so it's variable.

Yes, suppose we have a 1000 x 1000 binary sensels for every 8mpx sensel used in DxO normalized results.

And suppose they behave chaotically, but perfectly gaussian relative to 18% gray?
(Or even better, perfectly symmetrically gaussian relative to a perceptual 50%)?