D7000: an ISO less camera

[Pierre Sottas]

The sensor of the D7000 has very low read noise at base ISO. Given the DXOMark results, I expect that the read noise remains fairly constant at all ISO. In other words, this means that the D7000 is an ISO less camera.

I did a simple test to confirm this hypothesis. Here is an image taken at ISO6400 (with the 50 1.8 at f/8, 1/500):



and here is an image taken at ISO100 (with the 50 1.8 at f/8, 1/500) with exposure pushed by 6.2 stops in RPP:



The differences that we see between these two images must be weighted with the cons to shoot at ISO6400:

• 6 stops lost in sensor DR;
• a file about 50% larger (here: 25.5M for ISO6400 vs 16.6M for ISO100).

With today constant improvements in read noise, I expect that ISO will become a metadata in a near future.

Pierre

 

[bansal98]

I just responded to a post by bobn2 where he made the same assertion.

http://forums.dpreview.com/forums/readflat.asp?forum=1021&message=36874258

 

[Graystar]

But...so what? Who wants to fix every low light image in post just to see if it's a keeper...or just to see it at all??

It's the same as changing ISO. Okay. So just change the ISO because...you know...it's the same thing.

.

 

[Pierre Sottas]

But...so what? Who wants to fix every low light image in post just to see if it's a keeper...or just to see it at all??

It's the same as changing ISO. Okay. So just change the ISO because...you know...it's the same thing.

6 stops less DR at ISO6400, not really the same thing.

Pierre

 

[BoBCita]

So let me see if I understand what the both of you are saying in layman terms: shooting at base ISO and raising brightness in PP gives the same result as shooting at higher ISO, with perhaps a moderate (

Possibly useful in well controlled shooting situations, but not sure about most real-life cases.

What do you think?

--
BoB
-----------------------------------------------------
http://www.flickr.com/photos/bobcita/

 

[BoBCita]

6 stops less DR at ISO6400, not really the same thing.

OK, but when you brighten up the image in PP, you actually throw away (most of) those 6 stops (which just reproduce the almost blacks... don't you?).

--
BoB
-----------------------------------------------------
http://www.flickr.com/photos/bobcita/

 

[Graystar]

But...so what? Who wants to fix every low light image in post just to see if it's a keeper...or just to see it at all??

It's the same as changing ISO. Okay. So just change the ISO because...you know...it's the same thing.

6 stops less DR at ISO6400, not really the same thing.

Yes...it's exactly the same thing. This is already been discussed ad nauseam in the forum over hundreds of posts for years.

Because you’re underexposed, the sensor is never saturated...both exposures collected the same amount of light. In one, the small signal was multiplied in the camera. In the other, the same small signal was multiplied in the computer. It’s the same exact process (other than mostly negligible quantization errors on the computer.)

.

 

[Pierre Sottas]

But...so what? Who wants to fix every low light image in post just to see if it's a keeper...or just to see it at all??

It's the same as changing ISO. Okay. So just change the ISO because...you know...it's the same thing.

6 stops less DR at ISO6400, not really the same thing.

Yes...it's exactly the same thing. This is already been discussed ad nauseam in the forum over hundreds of posts for years.

Because you’re underexposed, the sensor is never saturated...both exposures collected the same amount of light. In one, the small signal was multiplied in the camera. In the other, the same small signal was multiplied in the computer. It’s the same exact process (other than mostly negligible quantization errors on the computer.)

No. Please read this post from Bobn2, very informative:
http://forums.dpreview.com/forums/read.asp?forum=1021&message=36874258

Pierre

 

[Pierre Sottas]

6 stops less DR at ISO6400, not really the same thing.

OK, but when you brighten up the image in PP, you actually throw away (most of) those 6 stops (which just reproduce the almost blacks... don't you?).

You can use Nikon D-Lighting to push the dark tones only. I personaly like RPP but this requires more work. In short, shoot at base ISO, expose to the right as much as possible without blowing any highlights (ideally using UniWB) and finally push the dark areas with the curve that you like.

Pierre

 

[Graystar]

But...so what? Who wants to fix every low light image in post just to see if it's a keeper...or just to see it at all??

It's the same as changing ISO. Okay. So just change the ISO because...you know...it's the same thing.

6 stops less DR at ISO6400, not really the same thing.

Yes...it's exactly the same thing. This is already been discussed ad nauseam in the forum over hundreds of posts for years.

Because you’re underexposed, the sensor is never saturated...both exposures collected the same amount of light. In one, the small signal was multiplied in the camera. In the other, the same small signal was multiplied in the computer. It’s the same exact process (other than mostly negligible quantization errors on the computer.)

No. Please read this post from Bobn2, very informative:
http://forums.dpreview.com/forums/read.asp?forum=1021&message=36874258

You are obviously misunderstanding what that data is indicating. You do not get 14 stops of DR by using ISO 100. You get 14 stops of DR when the light you're recording saturates (or comes very close to) saturating the pixels. When you shoot at high ISO you are collecting less light. That is why DR starts to drop. For any given sensor, the DR of any particular image is based not on the saturation DR, but upon the amount of light that was collected for that image. The less light you collect, the less your DR will be.

.

 

[rhlpetrus]

http://forums.dpreview.com/forums/read.asp?forum=1034&message=36881855

--
Renato.
http://www.flickr.com/photos/rhlpedrosa/
OnExposure member
http://www.onexposure.net/

Good shooting and good luck
(after Ed Murrow)

 

[BoBCita]

You are obviously misunderstanding what that data is indicating. You do not get 14 stops of DR by using ISO 100. You get 14 stops of DR when the light you're recording saturates (or comes very close to) saturating the pixels. When you shoot at high ISO you are collecting less light. That is why DR starts to drop. For any given sensor, the DR of any particular image is based not on the saturation DR, but upon the amount of light that was collected for that image. The less light you collect, the less your DR will be.

This is what I meant (in layman terms) when I said that if you shoot at iso 100 a scene which would otherwise require 6400 and brighten up in PP you start from 14 stops of DR but 6 of those are just black...

But I am not a sensor expert... just a humble physicist...

--
BoB
-----------------------------------------------------
http://www.flickr.com/photos/bobcita/

 

[Pierre Sottas]

You are obviously misunderstanding what that data is indicating. You do not get 14 stops of DR by using ISO 100. You get 14 stops of DR when the light you're recording saturates (or comes very close to) saturating the pixels.

Did I say that I "get" 14 stops of DR by using ISO100? No. Only that the recording media can "record" either 14 or 8 stops depending on the choice of the user. Would you choose ISO6400 for a scene luminance of 14 stops?

Before the D7000, there were some advantages (noise wise) to boost the ISO in-camera. With the D7000 there is no advantage anymore, there only remains the risk to blow highlights in function of the scene DR.

When you shoot at high ISO you are collecting less light.

It is not when you change the ISO that you collect less light, it is when you change the aperture and/or the shutter speed.

Pierre

 

[spbStan]

But...so what? Who wants to fix every low light image in post just to see if it's a keeper...or just to see it at all??

It's the same as changing ISO. Okay. So just change the ISO because...you know...it's the same thing.

6 stops less DR at ISO6400, not really the same thing.

Yes...it's exactly the same thing. This is already been discussed ad nauseam in the forum over hundreds of posts for years.

Because you’re underexposed, the sensor is never saturated...both exposures collected the same amount of light. In one, the small signal was multiplied in the camera. In the other, the same small signal was multiplied in the computer. It’s the same exact process (other than mostly negligible quantization errors on the computer.)

No. Please read this post from Bobn2, very informative:
http://forums.dpreview.com/forums/read.asp?forum=1021&message=36874258

You are obviously misunderstanding what that data is indicating. You do not get 14 stops of DR by using ISO 100. You get 14 stops of DR when the light you're recording saturates (or comes very close to) saturating the pixels. When you shoot at high ISO you are collecting less light. That is why DR starts to drop. For any given sensor, the DR of any particular image is based not on the saturation DR, but upon the amount of light that was collected for that image. The less light you collect, the less your DR will be.

.

Precisely, the camera has 14 stops DR from noise floor to saturation. A 6 stop under exposed light collects has a DR of from the noise floor to the light level 6 stops below saturation.

It can certainly reduce high clipping, but despite the low noise floor, there is little data to work with at the low end. So this is useful in critical situations where clipping highs is not tolerable and where a little DR can be sacrificed.

A very low noise floor is a nice thing and will add to the tool set photographers have.

The more we find out about this camera the more options comes to mind. In some regards this is the most significant DX camera to be released....ever.
--
Stan
St Petersburg Russia

 

[Pierre Sottas]

yep, thank you, my test only confirms what bobn2 wrote initially.

Time now for the manufacturers to put ISO in the RAW metadata and for Nikon to improve their D-Lightning application.

Pierre

http://forums.dpreview.com/forums/read.asp?forum=1034&message=36881855

--
Renato.
http://www.flickr.com/photos/rhlpedrosa/
OnExposure member
http://www.onexposure.net/

Good shooting and good luck
(after Ed Murrow)

 

[GrahamD]

With low light, you start with a noisy image. If readout noise is low, then the noise is primarily down to photon statistics. If you raise the ISO, you increase the analogue gain before digitisation. If the gain is too high, you clip the highlights, not because you have saturated the detector, but because you exceed the range of the ADC. The argument is that with 14 bits DR and 14 bit ADC you don't need to raise the analogue gain, you just keep to base ISO and amplifiy the signal digitally (which is presumably what the D7000 does with ISO 12800 and 25600i.e., digitally amplified ISO 6400).

Personally, I don't think I would take it to that extreme. But if you are shooting raw and you need high ISO for one reason or another, there may be some merit in dialing in a couple of stops under exposure and setting the ISO a couple of stops lower than you would otherwise use. Of course, the preview on your LCD screen will be dark and pretty unuseable.

 

[Graystar]

Did I say that I "get" 14 stops of DR by using ISO100? No.

Yes, you implied it.

Only that the recording media can "record" either 14 or 8 stops depending on the choice of the user.

The DR is based on scene illumination. If your scene only has 6 stops of DR, then that's all you're going to record...even at ISO 100.

When you shoot at high ISO you are collecting less light.

It is not when you change the ISO that you collect less light, it is when you change the aperture and/or the shutter speed.

Duh. Would you choose ISO6400 for a scene luminance of 14 stops? No. Obviously, having less available light has something to do with using high ISO.

.

 

[panos_m]

Thanks for the test.

Pierre Sottas wrote:

With today constant improvements in read noise, I expect that ISO will become a metadata in a near future.

I hope they (camera manufacturers) will do that.

--
Panagiotis

 

[binary_eye]

In short, shoot at base ISO, expose to the right as much as possible without blowing any highlights (ideally using UniWB) and finally push the dark areas with the curve that you like.

If the shot, and therefore the JPEG preview, is underexposed, UniWB will be useless.

 

[Stujomo]

Kind of goes against the whole idea of exposing to the right.

I just wonder how lifting really dark images will affect the tonality of the image.

The sensor of the D7000 has very low read noise at base ISO. Given the DXOMark results, I expect that the read noise remains fairly constant at all ISO. In other words, this means that the D7000 is an ISO less camera.

I did a simple test to confirm this hypothesis. Here is an image taken at ISO6400 (with the 50 1.8 at f/8, 1/500):



and here is an image taken at ISO100 (with the 50 1.8 at f/8, 1/500) with exposure pushed by 6.2 stops in RPP:



The differences that we see between these two images must be weighted with the cons to shoot at ISO6400:

• 6 stops lost in sensor DR;
• a file about 50% larger (here: 25.5M for ISO6400 vs 16.6M for ISO100).

With today constant improvements in read noise, I expect that ISO will become a metadata in a near future.

Pierre

--
Stuart...

http://www.sjmphotography.co.nr/

http://sjmphotography.wordpress.com/