X2D, GFX 100x long exposure noise at base ISO

[JimKasson]

In previous posts, I looked at 15 minute dark field exposures made with the Hasselblad X2D and the GFX 100S at ISO 800 and 1000 respectively.

There appear to be many people whose main use for long-exposure images is for subject blurring. Those people typically use neutral density filters and base ISO. I received requests to conduct the same sort of 15-minute exposure tests at the base ISO for both cameras.

I made two exposures.

• GFX 100S, LENR off, ISO 100, 14-bit precision, 15 minutes, 45mm f/2.8 GF lens and lens cap, f/32.
• X2D 100C, ISO 64, 14-bit precision, 17 minutes, 38mm f/2.5 XCD lens and lens cap f/32.

I reported on the results here:

https://blog.kasson.com/gfx-100s/long-exposure-noise-in-x2d-100c-gfx-100x-part-4/

Takeaway:

The X2D has more noise when measured in the raw file, but when viewed visually in a Lightroom developed image, the GFX 100S appears to have more noise. This is likely because the GFX spectra are biased towards lower frequencies, while the X2D ones are nearly flat.

Jim

--
https://blog.kasson.com

 

[yogi4fitness]

In previous posts, I looked at 15 minute dark field exposures made with the Hasselblad X2D and the GFX 100S at ISO 800 and 1000 respectively.

There appear to be many people whose main use for long-exposure images is for subject blurring. Those people typically use neutral density filters and base ISO. I received requests to conduct the same sort of 15-minute exposure tests at the base ISO for both cameras.

I made two exposures.

• GFX 100S, LENR off, ISO 100, 14-bit precision, 15 minutes, 45mm f/2.8 GF lens and lens cap, f/32.
• X2D 100C, ISO 64, 14-bit precision, 17 minutes, 38mm f/2.5 XCD lens and lens cap f/32.

I reported on the results here:

https://blog.kasson.com/gfx-100s/long-exposure-noise-in-x2d-100c-gfx-100x-part-4/

Takeaway:

The X2D has more noise when measured in the raw file, but when viewed visually in a Lightroom developed image, the GFX 100S appears to have more noise. This is likely because the GFX spectra are biased towards lower frequencies, while the X2D ones are nearly flat.

Jim

Thanks for the comparison Jim! Do the darkfield images of the two cameras explain why the GFX100S images have a magenta shift and the X2D images have a green shift in AWB images.

 

[JimKasson]

In previous posts, I looked at 15 minute dark field exposures made with the Hasselblad X2D and the GFX 100S at ISO 800 and 1000 respectively.

There appear to be many people whose main use for long-exposure images is for subject blurring. Those people typically use neutral density filters and base ISO. I received requests to conduct the same sort of 15-minute exposure tests at the base ISO for both cameras.

I made two exposures.

• GFX 100S, LENR off, ISO 100, 14-bit precision, 15 minutes, 45mm f/2.8 GF lens and lens cap, f/32.
• X2D 100C, ISO 64, 14-bit precision, 17 minutes, 38mm f/2.5 XCD lens and lens cap f/32.

I reported on the results here:

https://blog.kasson.com/gfx-100s/long-exposure-noise-in-x2d-100c-gfx-100x-part-4/

Takeaway:

The X2D has more noise when measured in the raw file, but when viewed visually in a Lightroom developed image, the GFX 100S appears to have more noise. This is likely because the GFX spectra are biased towards lower frequencies, while the X2D ones are nearly flat.

Jim

Thanks for the comparison Jim! Do the darkfield images of the two cameras explain why the GFX100S images have a magenta shift and the X2D images have a green shift in AWB images.

No. The average of all four raw channels is close — exceptionally so in the X2D case — so I am tentatively putting that down to black point errors in the raw development.

 

[FDecker]

Thank you Jim, interesting as always.

I have a hard time to understand the results, though. Looking at the spectral analysis, the X2D is clearly superior to the GFX. In any aspect. Still, the EDR you calculated is lower. The full scale should be roughly the same, right? So, how can this be?
As you write, also to the eye, the image X2D is more pleasing.

 

[JimKasson]

Thank you Jim, interesting as always.

And somewhat surprising, to me at least.

I have a hard time to understand the results, though. Looking at the spectral analysis, the X2D is clearly superior to the GFX. In any aspect. Still, the EDR you calculated is lower. The full scale should be roughly the same, right? So, how can this be?
As you write, also to the eye, the image X2D is more pleasing.

The spectral results are normalized, and say nothing about the total amount of the noise, only about its frequency distribution.

 

[FDecker]

Thank you Jim, interesting as always.

And somewhat surprising, to me at least.

I have a hard time to understand the results, though. Looking at the spectral analysis, the X2D is clearly superior to the GFX. In any aspect. Still, the EDR you calculated is lower. The full scale should be roughly the same, right? So, how can this be?
As you write, also to the eye, the image X2D is more pleasing.

The spectral results are normalized, and say nothing about the total amount of the noise, only about its frequency distribution.

Normalized in what way? Aren‘t you calculating the FFT over the spatial noise distribution, taking the same area in both cases?

 

[JimKasson]

Thank you Jim, interesting as always.

And somewhat surprising, to me at least.

I have a hard time to understand the results, though. Looking at the spectral analysis, the X2D is clearly superior to the GFX. In any aspect. Still, the EDR you calculated is lower. The full scale should be roughly the same, right? So, how can this be?
As you write, also to the eye, the image X2D is more pleasing.

The spectral results are normalized, and say nothing about the total amount of the noise, only about its frequency distribution.

Normalized in what way? Aren‘t you calculating the FFT over the spatial noise distribution, taking the same area in both cases?

Normalized for amplitude. A flat spectrum will give a noisy flat line at zero, no matter the amplitude of the signal.

 

[FDecker]

Thank you Jim, interesting as always.

And somewhat surprising, to me at least.

I have a hard time to understand the results, though. Looking at the spectral analysis, the X2D is clearly superior to the GFX. In any aspect. Still, the EDR you calculated is lower. The full scale should be roughly the same, right? So, how can this be?
As you write, also to the eye, the image X2D is more pleasing.

The spectral results are normalized, and say nothing about the total amount of the noise, only about its frequency distribution.

Normalized in what way? Aren‘t you calculating the FFT over the spatial noise distribution, taking the same area in both cases?

Normalized for amplitude. A flat spectrum will give a noisy flat line at zero, no matter the amplitude of the signal.

Still confused:

This is my understanding of your data: You take the amplitude data from the sensor area, i.e. the sensor noise, and run a 2D FFT, finally showing a vertical and horizontal line from the 2D FFT spectrum to us. Or maybe you take a horizontal and vertical line from the sensor data as a representative of the 2D sensor data and run a 1D FFT on these single lines. In both cases, the noise amplitudes in the spectra refer to the noise amplitudes in the sensor data. As you have no data but noise, the spectrum should be centered around zero. Even if you have to suppress a DC offset in the spectrum (which shouldn't be there unless you have a delta spike somewhere in the sensor data), the noise amplitudes must be on the same scale for the X2D and the GFX data. But looking at the scale and amplitudes of the plots, it seems that the noise amplitudes for the GFX are considerably higher.

What do I miss?

 

[JimKasson]

GFX 100S


X2D in Lr


X2D in Phocus

EV +9 stop Exposure boost. Cropped full-res images. I tried to post the uncropped ones, an DPR upload hung. Exposure times in DPR's interpretation of the EXIF metadata are wrong.

--
https://blog.kasson.com

 

[yogi4fitness]

GFX 100S


X2D in Lr


X2D in Phocus

EV +9 stop Exposure boost. Cropped full-res images. I tried to post the uncropped ones, an DPR upload hung. Exposure times in DPR's interpretation of the EXIF metadata are wrong.

--
https://blog.kasson.com

wow thats remarkable how clean the phocus processed x2D crop looks. Would it be worth comparing phocus processed x2D vs Lr processed GFX100s noise?

 

[JimKasson]

GFX 100S


X2D in Lr


X2D in Phocus

EV +9 stop Exposure boost. Cropped full-res images. I tried to post the uncropped ones, an DPR upload hung. Exposure times in DPR's interpretation of the EXIF metadata are wrong.

wow thats remarkable how clean the phocus processed x2D crop looks. Would it be worth comparing phocus processed x2D vs Lr processed GFX100s noise?

Compare the first image above to the third to do that.

--

the last word the last word - Photography meets digital computer technology. Photography wins -- most of the time.

Photography meets digital computer technology. Photography wins -- most of the time.

blog.kasson.com

 

[yogi4fitness]

GFX 100S


X2D in Lr


X2D in Phocus

EV +9 stop Exposure boost. Cropped full-res images. I tried to post the uncropped ones, an DPR upload hung. Exposure times in DPR's interpretation of the EXIF metadata are wrong.

wow thats remarkable how clean the phocus processed x2D crop looks. Would it be worth comparing phocus processed x2D vs Lr processed GFX100s noise?

Compare the first image above to the third to do that.

I meant amplitude of the noise quantitative comparison.

On a separate note is there any other way of processing the GFX raw files (in C1 or some other processing SW) that will give as clean a result as image 3?

 

[JimKasson]

GFX 100S


X2D in Lr


X2D in Phocus

EV +9 stop Exposure boost. Cropped full-res images. I tried to post the uncropped ones, an DPR upload hung. Exposure times in DPR's interpretation of the EXIF metadata are wrong.

wow thats remarkable how clean the phocus processed x2D crop looks. Would it be worth comparing phocus processed x2D vs Lr processed GFX100s noise?

Compare the first image above to the third to do that.

I meant amplitude of the noise quantitative comparison.

Oh. I only do that comparison on raw files.

On a separate note is there any other way of processing the GFX raw files (in C1 or some other processing SW) that will give as clean a result as image 3?

I don't know of one, but I turned raw developer noise reduction off for these shots.

--

the last word the last word - Photography meets digital computer technology. Photography wins -- most of the time.

Photography meets digital computer technology. Photography wins -- most of the time.

blog.kasson.com