Leica SL Sensor Measurements Published at PhotonsToPhotos

Started 4 weeks ago | Discussions
Shop cameras & lenses ▾
bclaff OP Veteran Member • Posts: 4,025
Re: Leica SL Sensor Measurements Published at PhotonsToPhotos

Jared_Willson wrote:

bclaff wrote:...

At any rate, here are the values I got assuming an SNR of 10 dB constituted the noise floor:

ISO 50: 11.3 EV 13.1 / 10.81 (EDR / PDR)
ISO 100: 10.9 EV 12.2 / 9.88
ISO 200: 11.1 EV 13.3 / 10.10
ISO 400: 10.7 EV 12.4 / 9.14
ISO 800: 9.6 EV 11.5 / 8.15
ISO 1600: 8.2 EV 10.5 / 7.17

I have added my values in BOLD next to yours.
The are from Read Noise in DNs versus ISO Setting

And I also added the PDR values from Photographic Dynamic Range versus ISO Setting

I doubt we actually have any disagreement here.

No disagreement; just discussion.

1) I'm sure you are right about the Leica changing things up during firmware revisions. No reason to doubt--you've looked and I haven't.

2) I agree that engineering dynamic range is fairly well defined--DR = 20 log (Full Well / Read Noise). I, like you, do not find those values particularly realistic for photographers. I was reporting out my version of Photographic Read Noise, and we don't quite have the same assumptions for that. For example, you use a SNR of 20 as the noise floor. I was using 10. That's why I reported out a DR that is slightly larger than you did in your original post. I wasn't trying to contradict your values--just show that the trends I found were in agreement with the trends you were finding. It's on the definition of a useful photographic dynamic range that I don't believe there is a strong consensus.

I didn't even realize your intent.
FWIW, PDR is SNR 20 normalized over the Circle Of Confusion (COC), not per pixel like EDR.

3) You're right that I was jumping the gun in assuming the lack of linearity near the clipping point is because Leica applies a curve. It could just be the anti-blooming gate kicking in.

4) Like you, I am confused why DXO isn't showing a blip in their data, even if Leica have changed the ISO for the crossover point. I definitely see it--just as you do. I've run the tests twice now on my camera, and it is definitely there. It's not huge, but it's definitely too big to just be a simple measurement error.

Regards,

-- hide signature --

Bill ( Your trusted source for independent sensor data at http://www.PhotonsToPhotos.net )

bclaff OP Veteran Member • Posts: 4,025
Re: Leica SL Sensor Measurements Published at PhotonsToPhotos

vbd70 wrote:

Jared_Willson wrote:

...

Jared, thank you for this. Unfortunately for me, I don't do night landscapes necessarily, in fact I mostly do very long exposures during the day; and as you said I do them for expressive reasons, so for me - taking your example as an example - if I need, say, 8 minutes at ISO 100 for the clouds to be blurred just so, going for 1 minute at ISO 800 is not an option.

So in short - you are saying that exposure time being equal I am better off using ISO 200 rather than ISO 50 or keeping noise down? Thanks!

Yes, if exposure time isn't an issue going to ISO 200 is a "no brainer".

Regards,

-- hide signature --

Bill ( Your trusted source for independent sensor data at http://www.PhotonsToPhotos.net )

Jared_Willson Contributing Member • Posts: 505
Re: Leica SL Sensor Measurements Published at PhotonsToPhotos

bclaff wrote:

Jared_Willson wrote:

bclaff wrote:...

At any rate, here are the values I got assuming an SNR of 10 dB constituted the noise floor:

ISO 50: 11.3 EV 13.1 / 10.81 (EDR / PDR)
ISO 100: 10.9 EV 12.2 / 9.88
ISO 200: 11.1 EV 13.3 / 10.10
ISO 400: 10.7 EV 12.4 / 9.14
ISO 800: 9.6 EV 11.5 / 8.15
ISO 1600: 8.2 EV 10.5 / 7.17

I have added my values in BOLD next to yours.
The are from Read Noise in DNs versus ISO Setting

And I also added the PDR values from Photographic Dynamic Range versus ISO Setting

I doubt we actually have any disagreement here.

No disagreement; just discussion.

1) I'm sure you are right about the Leica changing things up during firmware revisions. No reason to doubt--you've looked and I haven't.

2) I agree that engineering dynamic range is fairly well defined--DR = 20 log (Full Well / Read Noise). I, like you, do not find those values particularly realistic for photographers. I was reporting out my version of Photographic Read Noise, and we don't quite have the same assumptions for that. For example, you use a SNR of 20 as the noise floor. I was using 10. That's why I reported out a DR that is slightly larger than you did in your original post. I wasn't trying to contradict your values--just show that the trends I found were in agreement with the trends you were finding. It's on the definition of a useful photographic dynamic range that I don't believe there is a strong consensus.

I didn't even realize your intent.
FWIW, PDR is SNR 20 normalized over the Circle Of Confusion (COC), not per pixel like EDR.

Can you point me to a reference on this?  Like you, I agree that something like your Photographic DR is a more useful indicator of the useful dynamic range of a camera, and that it is significantly different from engineering DR  which simply assumes that the noise floor is where the read noise and the signal value are equal (SNR 1).  I also understand that the Circle of Confusion is the blur size for a given format that will still appear to be point-like on an 8"x10" print at a viewing distance of 10" (at least traditionally).  I'm not sure how/why the two concepts are related.  Just that you measure your SNR of 20 across a different number of pixels, depending on the pixel pitch and overall size of the sensor.  Again, I'd love to see a reference if you can point me to one.  I'm curious how you developed this process and particular set of standards.

3) You're right that I was jumping the gun in assuming the lack of linearity near the clipping point is because Leica applies a curve. It could just be the anti-blooming gate kicking in.

4) Like you, I am confused why DXO isn't showing a blip in their data, even if Leica have changed the ISO for the crossover point. I definitely see it--just as you do. I've run the tests twice now on my camera, and it is definitely there. It's not huge, but it's definitely too big to just be a simple measurement error.

Regards,

-- hide signature --

Bill ( Your trusted source for independent sensor data at http://www.PhotonsToPhotos.net )

-- hide signature --

- Jared Willson

bclaff OP Veteran Member • Posts: 4,025
Re: Leica SL Sensor Measurements Published at PhotonsToPhotos

Jared_Willson wrote:

bclaff wrote:

2) I agree that engineering dynamic range is fairly well defined--DR = 20 log (Full Well / Read Noise). I, like you, do not find those values particularly realistic for photographers. I was reporting out my version of Photographic Read Noise, and we don't quite have the same assumptions for that. For example, you use a SNR of 20 as the noise floor. I was using 10. That's why I reported out a DR that is slightly larger than you did in your original post. I wasn't trying to contradict your values--just show that the trends I found were in agreement with the trends you were finding. It's on the definition of a useful photographic dynamic range that I don't believe there is a strong consensus.

I didn't even realize your intent.
FWIW, PDR is SNR 20 normalized over the Circle Of Confusion (COC), not per pixel like EDR.

Can you point me to a reference on this? Like you, I agree that something like your Photographic DR is a more useful indicator of the useful dynamic range of a camera,

Try the links under "Further Reading" on the PDR Chart page.

Regards,

-- hide signature --

Bill ( Your trusted source for independent sensor data at http://www.PhotonsToPhotos.net )

vbd70
vbd70 Veteran Member • Posts: 5,456
Re: Leica SL Sensor Measurements Published at PhotonsToPhotos

Jared_Willson wrote:

vbd70 wrote:

Jared,

thank you for the comprehensive explanation - much appreciated indeed.

I see both white-ish speckles AND black, bigger ones; I thought either appeared randomly, but now I have a better understanding of the phenomenon.

Would making the back frame shorter than the actual exposure help in limiting these? Some manufacturer use to have a black frame shorter than the actual exposure, and I thought this might be the reason why.

Thanks again, best

Vieri

-- hide signature --

Vieri Bottazzini, Leica Ambassador, ABIPP EP
https://vieribottazzini.com
https://vieribottazziniworkshops.com

I haven't seen anyone scaling their dark frames, but it is certainly possible to do. It's a little more process intensive. Here is the standard process:

Take a light frame, for sake of argument 5 minutes duration

Take a dark frame, the same five minutes

Subtract the dark frame from the light frame. That's your new final exposure.

Here is what you would do for a scaled dark:

Take a light frame

Take a bias frame (shortest possible exposure with the shutter closed--just read noise)

Take a shortened dark frame, say half the length of your light

Subtract the bias from the dark frame

Multiply all values in the dark frame by two

Subtract the bias from your light frame

Subtract the new 2x dark from the light

The result is your new, final exposure

Scaled darks work fairly well as long as the exposures aren't long enough for hot pixels to be clipping.

I'm not certain how best to address them aside from what you are already doing. In astrophotography, most people choose to use master darks (which will create a better result than a single dark frame) and dither their images--each image is slightly mis-aligned from the last, and then they are averaged together using a statistical method such as "sigma clip" to completely remove hot and cold pixels. That is technically possible with a terrestrial image as well. You could take five separate two minute exposures or so with the tripod moved ever so slightly between frames, then use a sigma clip combine to average the frames. Of course, Photoshop and Lightroom don't have sophisticated stacking algorithms so you'd have to use some astronomy software for that, and most astronomy software doesn't do a great job on aligning terrestrial images since it expects point sources to be available, but it is definitely doable. I have done it occasionally. I wouldn't recommend it, though, unless you just can't accomplish what you want to with the techniques you are already trying.

Perhaps an incremental solution would be to cut your exposure duration significantly (to lower the thermal noise), boost your ISO to compensate, then take multiple exposures and average them together in Photoshop without any re-alignment. You wouldn't get the benefits of dithering, but you would start out with lower thermal noise. This comes at the cost of higher read noise in your image, but I bet you could find a reasonable balance. For example, instead of one single 8 minute exposure at ISO 100, try 8 separate one minute exposures at ISO 800 averaged in Photoshop as separate layers. You can do this with the "smart object" tool. That should give you an overall cleaner result without requiring any more time when shooting. I'm not quite sure what it will do to the clouds, but it might work quite well. Give it a shot.

-- hide signature --

Jared Willson

Thank you Jared, most informative as always. My Pentax 645z used to take a dark frame slightly shorter than my exposure.

About your last point, as I answered below to another of your previous posts I do long exposures for expressive reasons, so for me - taking your example as an example - if I need, say, 8 minutes at ISO 100 for the clouds to be blurred just so, going for 8 times 1 minute at ISO 800 is not an option, unfortunately.

Best,

Vieri

-- hide signature --

Vieri Bottazzini, Leica Ambassador, ABIPP EP
https://vieribottazzini.com
https://vieribottazziniworkshops.com

 vbd70's gear list:vbd70's gear list
Panasonic LX100 Leica SL (Typ 601) Leica SL 24-90mm F2.8-4 +5 more
vbd70
vbd70 Veteran Member • Posts: 5,456
Re: Leica SL Sensor Measurements Published at PhotonsToPhotos

bclaff wrote:

vbd70 wrote:

Jared_Willson wrote:

...

Jared, thank you for this. Unfortunately for me, I don't do night landscapes necessarily, in fact I mostly do very long exposures during the day; and as you said I do them for expressive reasons, so for me - taking your example as an example - if I need, say, 8 minutes at ISO 100 for the clouds to be blurred just so, going for 1 minute at ISO 800 is not an option.

So in short - you are saying that exposure time being equal I am better off using ISO 200 rather than ISO 50 or keeping noise down? Thanks!

Yes, if exposure time isn't an issue going to ISO 200 is a "no brainer".

Regards,

-- hide signature --

Bill ( Your trusted source for independent sensor data at http://www.PhotonsToPhotos.net )

Bill,

Thank you. Two more questions.

1. You definitely say that if I get, i.e., 8 minutes @ ISO 50 I'd be better off adding one more ND and go to 8 minutes @ ISO 200 for less noise, correct?

2. Help me understand this - from your graphs it shows that ISO 50 is better for photographic dynamic range, but that ISO 200 is better for read noise. I was under the impression that the two were linked: the better the read noise, the better the DR. Is it not so?

Thank you very much - between you and Jared I learned quite a lot so far with this thread, much appreciated! Best,

Vieri

-- hide signature --

Vieri Bottazzini, Leica Ambassador, ABIPP EP
https://vieribottazzini.com
https://vieribottazziniworkshops.com

 vbd70's gear list:vbd70's gear list
Panasonic LX100 Leica SL (Typ 601) Leica SL 24-90mm F2.8-4 +5 more
bclaff OP Veteran Member • Posts: 4,025
Re: Leica SL Sensor Measurements Published at PhotonsToPhotos

vbd70 wrote:

bclaff wrote:

vbd70 wrote:

Jared_Willson wrote:

...

Jared, thank you for this. Unfortunately for me, I don't do night landscapes necessarily, in fact I mostly do very long exposures during the day; and as you said I do them for expressive reasons, so for me - taking your example as an example - if I need, say, 8 minutes at ISO 100 for the clouds to be blurred just so, going for 1 minute at ISO 800 is not an option.

So in short - you are saying that exposure time being equal I am better off using ISO 200 rather than ISO 50 or keeping noise down? Thanks!

Yes, if exposure time isn't an issue going to ISO 200 is a "no brainer".

Bill,

Thank you. Two more questions.

1. You definitely say that if I get, i.e., 8 minutes @ ISO 50 I'd be better off adding one more ND and go to 8 minutes @ ISO 200 for less noise, correct?

Provided you have apparent noise then it will be lower at ISO 50 than ISO 200.
The chief advantage of any lower ISO is having the option to gather more light (your NDs affect this).
If you're unable to use a longer exposure or larger aperture then in your case you may see no difference.

2. Help me understand this - from your graphs it shows that ISO 50 is better for photographic dynamic range, but that ISO 200 is better for read noise. I was under the impression that the two were linked: the better the read noise, the better the DR. Is it not so?

Read noise and dynamic range are loosely correlated but lower read noise doesn't necessarily mean higher dynamic range.
As you raise the ISO setting you increase internal amplification (gain) and highlights get clipped from the top end.
So, although you get better shadows by raising ISO, you lose highlights; it's a trade-off.

Thank you very much - between you and Jared I learned quite a lot so far with this thread, much appreciated! Best,

Regards,

-- hide signature --

Bill ( Your trusted source for independent sensor data at http://www.PhotonsToPhotos.net )

Joe Pineapples II
Joe Pineapples II Senior Member • Posts: 1,228
Source of noise in long exposures
1

vbd70 wrote:

Jared, thank you for this. Unfortunately for me, I don't do night landscapes necessarily, in fact I mostly do very long exposures during the day; and as you said I do them for expressive reasons, so for me - taking your example as an example - if I need, say, 8 minutes at ISO 100 for the clouds to be blurred just so, going for 1 minute at ISO 800 is not an option.

So in short - you are saying that exposure time being equal I am better off using ISO 200 rather than ISO 50 or keeping noise down? Thanks!

The tricky thing is that Bill's measurements are highly relevant to people doing "normal" exposures who want to extract the best performance from their cameras, while your images are being affected by thermal noise issues due to the very extended exposure times that are more like those encountered in astrophotography. Because you are blurring movement in your scenes, then the technique of stacking multiple images would be a possible approach. However, the aesthetic effect of say an 8-minute exposure vs a stack of 8 1-minute exposures would be a matter of personal taste.

In general, techniques by themselves are neither good nor bad - only photographs are good or bad!

J.

Keyboard shortcuts:
FForum MMy threads