Dynamic range and sensor size

[a_c_skinner]

Can someone give me a logically persausive explantion of the fact that smaller sensors have lower DR?

This is the comparison of D800 FX and DX . Same sensor, same pixels. I cannot make the logical leap that is needed to grasp why trimming off the outside of the frame reduces the DR of what is left. Surely cropping in processing won't do the same (rationally it cannot) but I'm at a loss to understand why.

 

[D Cox]

If you view two images on the same sized screen or print, the cropped one will be enlarged more.

Greater enlargement generally reduces signal amplitudes at higher spatial frequencies (line-pairs per image height). In other words, the enlarged image is less sharp. The amplitude of noise is less affected by enlargement. so the signal-to-noise ratio goes down.

It isn't clear from the graph you linked how DR was measured. If DXO were involved at all, I wouldn't trust it.

 

[thinkinginimages]

Can someone give me a logically persausive explantion of the fact that smaller sensors have lower DR?

This is the comparison of D800 FX and DX . Same sensor, same pixels. I cannot make the logical leap that is needed to grasp why trimming off the outside of the frame reduces the DR of what is left. Surely cropping in processing won't do the same (rationally it cannot) but I'm at a loss to understand why.

--
Andrew Skinner

Tough one because the technology changes fast and a blanket answer isn't really appropriate.

Technically a smaller sensor can have trouble gathering light because the cell sites are smaller. Beyond the threshold of being able to register that light accurately you get clipping and blocking. We ran into this with small sensors with high megapixel counts, because people relate that count with "better".

Of course that depends on the sensor, and a lot to the processing power and programming of the camera.

But, we're into something new now. Some cameras are powerful enough see the sensor more as a collection of sensors and can adapt a bit. They can add a bit of exposure to dark areas and subtract a bit from bright areas.

You may be assuming "constants" here. The sensor may not be the exact same sensor and the "computer", software, and firmware in the cameras may not be the same.

 

[fishy wishy]

Larger sensors can fit more magic elves in them, stupid.

 

[Rick Knepper]

Can someone give me a logically persausive explantion of the fact that smaller sensors have lower DR?

This is the comparison of D800 FX and DX . Same sensor, same pixels.

It's not the same pixels after cropping.

I cannot make the logical leap that is needed to grasp why trimming off the outside of the frame reduces the DR of what is left. Surely cropping in processing won't do the same (rationally it cannot) but I'm at a loss to understand why.

 

[D Cox]

Can someone give me a logically persausive explantion of the fact that smaller sensors have lower DR?

This is the comparison of D800 FX and DX . Same sensor, same pixels.

It's not the same pixels after cropping.

Really ? What happens to them ?

I can imagine the ones that are left out being disappointed or jealous.

I cannot make the logical leap that is needed to grasp why trimming off the outside of the frame reduces the DR of what is left. Surely cropping in processing won't do the same (rationally it cannot) but I'm at a loss to understand why.

 

[golfhov]

Can someone give me a logically persausive explantion of the fact that smaller sensors have lower DR?

because they are smaller. Imagine having an acoustic concert with half the players.

This is the comparison of D800 FX and DX . Same sensor, same pixels. I cannot make the logical leap that is needed to grasp why trimming off the outside of the frame reduces the DR of what is left. Surely cropping in processing won't do the same (rationally it cannot) but I'm at a loss to understand why.

DX Crop Mode Photographic Dynamic Range

 

[fishy wishy]

Writers used to say that medium format used to let them see more in the shadows. You are thinking of it clearly per pixel, but you certainly improve the lower end of the range by collecting more light with more pixels. Yes, you also get improved sensor measurements. Why wouldn't you?

 

[tbcass]

I suspect you own a camera that scores poorly in DR by DXO. Because DXO's sensor measurements are so rigidly controlled and sound engineering wise I trust them more than anybody else.

 

[tbcass]

Just like signal to noise ratio DR decreases when less light is captured. Among sensors using the same technology larger sensors have more DR because they capture more total light than smaller sensors.

 

[Gerry Winterbourne]

Can someone give me a logically persausive explantion of the fact that smaller sensors have lower DR?

This is the comparison of D800 FX and DX . Same sensor, same pixels. I cannot make the logical leap that is needed to grasp why trimming off the outside of the frame reduces the DR of what is left. Surely cropping in processing won't do the same (rationally it cannot) but I'm at a loss to understand why.

This is a much simplified explanation which I hope will help you understand the basic principles.

Dynamic range is the ratio between (a) the darkest point where details in a photo (the signal) can be made out without being obscured by noise and (b) the brightest point where details are lost because of clipping.

Although we think of light as being constant it is, in fact, a quantum phenomenon and is subject to random variation. It's that variation that causes (indeed, actually is) noise - with a constant signal things would always be the same brightness or darkness, but noise makes something that should be the same randomly brighter or darker.

The noise in light itself (nothing to do with the electronics in the camera) varies with the inverse square of the amount of light - which is measured by the number of photos. Four times as many photons suffer only twice the amount of noise.

At point (b) there are many, many photons so there is little variation and noise is effectively the same for both sizes of sensor. What concerns us is the dark end.

When measuring noise on the sensor (not individual pixels) the DX size is under half the area of the full frame so it collects under half the number of photons. This makes it noisier - by a factor of about 0.7 - so that would reduce the DX DR because it lifts the bottom end (point (a). It's too simplistic to say that it just knocks 0.7 stops off the sensor DR but the measured difference on the charts is 11.41 - 10.6 = 0.81 stops so it's not grossly untrue.

 

[tbcass]

I suspect your very detailed explanation may go right over the OP's head. ;-)

 

[a_c_skinner]

A rather rude reply but this is DPR ...

 

[Ilfotografo]

Can someone give me a logically persausive explantion of the fact that smaller sensors have lower DR?

This is the comparison of D800 FX and DX . Same sensor, same pixels. I cannot make the logical leap that is needed to grasp why trimming off the outside of the frame reduces the DR of what is left. Surely cropping in processing won't do the same (rationally it cannot) but I'm at a loss to understand why.

This is a much simplified explanation which I hope will help you understand the basic principles.

Dynamic range is the ratio between (a) the darkest point where details in a photo (the signal) can be made out without being obscured by noise and (b) the brightest point where details are lost because of clipping.

Although we think of light as being constant it is, in fact, a quantum phenomenon and is subject to random variation. It's that variation that causes (indeed, actually is) noise - with a constant signal things would always be the same brightness or darkness, but noise makes something that should be the same randomly brighter or darker.

The noise in light itself (nothing to do with the electronics in the camera) varies with the inverse square of the amount of light - which is measured by the number of photos. Four times as many photons suffer only twice the amount of noise.

At point (b) there are many, many photons so there is little variation and noise is effectively the same for both sizes of sensor. What concerns us is the dark end.

When measuring noise on the sensor (not individual pixels) the DX size is under half the area of the full frame so it collects under half the number of photons. This makes it noisier - by a factor of about 0.7 - so that would reduce the DX DR because it lifts the bottom end (point (a). It's too simplistic to say that it just knocks 0.7 stops off the sensor DR but the measured difference on the charts is 11.41 - 10.6 = 0.81 stops so it's not grossly untrue.

AH! Finally a serious attempt to explain a phenomenon that is not widely understood.

Excellent!

 

[a_c_skinner]

Thank you, that plus a private reply make sense. I'd not appreciated that noise limited DR in that way, though it makes sense.

 

[tbcass]

A rather rude reply but this is DPR ...

Don't be offended. It was tongue in cheek.

 

[a_c_skinner]

It was however close to the truth! I assumed it was TIC, unless of course you've read some of my attempts to explain things!

 

[Chaplain Mark]

Can someone give me a logically persausive explantion of the fact that smaller sensors have lower DR?

This is the comparison of D800 FX and DX . Same sensor, same pixels.

It's not the same pixels after cropping.

Really ? What happens to them ?

I can imagine the ones that are left out being disappointed or jealous.

I cannot make the logical leap that is needed to grasp why trimming off the outside of the frame reduces the DR of what is left. Surely cropping in processing won't do the same (rationally it cannot) but I'm at a loss to understand why.

It's the pixels that remain on the sensor who are disappointed or jealous that they remain captive behind the shutter, while their peers have been liberated to see the world outside the camera 24-7.

They're in a bad mood, thus the reduced DR.

 

[J A C S]

Can someone give me a logically persausive explantion of the fact that smaller sensors have lower DR?

This is the comparison of D800 FX and DX . Same sensor, same pixels. I cannot make the logical leap that is needed to grasp why trimming off the outside of the frame reduces the DR of what is left. Surely cropping in processing won't do the same (rationally it cannot) but I'm at a loss to understand why.

This is a much simplified explanation which I hope will help you understand the basic principles.

Dynamic range is the ratio between (a) the darkest point where details in a photo (the signal) can be made out without being obscured by noise

More precisely, the noise floor, where the signal is equal to the noise; the latter measured by its standard deviation. This is the "engineering DR". Bill's definition of his "photographic DR" is different: is a low endpoint with an SNR of 20 when adjusted for the appropriate Circle Of Confusion (COC) for the sensor. The CoC is proportional to the sensor size. As such, the PDR is affected both by read/sensor noise and photon noise.

and (b) the brightest point where details are lost because of clipping.

Although we think of light as being constant it is, in fact, a quantum phenomenon and is subject to random variation. It's that variation that causes (indeed, actually is) noise - with a constant signal things would always be the same brightness or darkness, but noise makes something that should be the same randomly brighter or darker.

The noise in light itself (nothing to do with the electronics in the camera) varies with the inverse square of the amount of light - which is measured by the number of photos.

Direct square root, not inverse. The NSR however is proportional to the inverse square root.

Four times as many photons suffer only twice the amount of noise.

At point (b) there are many, many photons so there is little variation and noise is effectively the same for both sizes of sensor.

The photon NSR is always a certain constant (in stops) above that of a large sensor, regardless of the luminosity level. It is well visible in the midtones at least.

What concerns us is the dark end.

When measuring noise on the sensor (not individual pixels) the DX size is under half the area of the full frame so it collects under half the number of photons. This makes it noisier - by a factor of about 0.7 - so that would reduce the DX DR because it lifts the bottom end (point (a). It's too simplistic to say that it just knocks 0.7 stops off the sensor DR but the measured difference on the charts is 11.41 - 10.6 = 0.81 stops so it's not grossly untrue.

You are ignoring completely the read noise, which is the whole point of the engineering DR and an important factor for the Bill's DR. Smaller sensors have a slightly higher NSR of that read noise, plus a higher photon NSR, and Bill's DR combines both. Those numbers are adjusted for sensor areas to reflect prints of the same size, which explains the differences.

 

[Chaplain Mark]

Thank you, that plus a private reply make sense. I'd not appreciated that noise limited DR in that way, though it makes sense.