SL3 , am I the only one ?

[BernardC]

Advertisement (gone after registration)

9 hours ago, Chaemono said:

Dynamic range is measured in stops, which is a logarithmic scale. Each stop represents a doubling or halving of light intensity.

Great answer.

Every commercial raw format has been encoded on a  log scale, beginning with EXR (motion picture film scanning). Film density (HD curves) and sound pressure (dB) are also encoded that way. The alternative would be linear encoding, where the least significant bit would take-up half of the possible values.

12 hours ago, ZHNL said:

Why is that? the methodology is the same, they plotted peak DR range one can achieve at given ISO.  At ISO6400, assume base ISO SL3 is 100, that means 6 stop gain applied. With same headroom of ADC, the peak DR has to be about 6 stop, (if we trust SO400 10 stop performance from there website, assume this is the dual gain setting point, from ISO400 to ISO6400, there should 4 stop gain applied, with the same headroom, DR will be 10-4= 6 stop) nothing surprising here.

EI 6400 is in the second gain range of the sensor, so base EI isn't 100 in this case, but closer to 800. The AD converter doesn't throw-out higher values to get to higher speeds. They are still there, you just get a different middle-grey value with more highlight range and less shadow range (because middle grey is deeper into the shadows).

[Advertisement]

The Leica Q3 is available for pre-order at B&H Photo Video and Adorama

[SrMi]

2 hours ago, BernardC said:

Great answer.

Every commercial raw format has been encoded on a  log scale, beginning with EXR (motion picture film scanning). Film density (HD curves) and sound pressure (dB) are also encoded that way. The alternative would be linear encoding, where the least significant bit would take-up half of the possible values.

EI 6400 is in the second gain range of the sensor, so base EI isn't 100 in this case, but closer to 800. The AD converter doesn't throw-out higher values to get to higher speeds. They are still there, you just get a different middle-grey value with more highlight range and less shadow range (because middle grey is deeper into the shadows).

That is wrong. Here is Leica's table, p337, from SL3's manual. While Leica claims ISO 50 to be the (first) native ISO, I would avoid it.

Welcome, dear visitor! As registered member you'd see an image here…

Simply register for free here – We are always happy to welcome new members!

[SrMi]

8 hours ago, FlashGordonPhotography said:

No it isn’t. It’s a daft system. Absolutely NO ONE buys a SL3 or A7R5 to shoot it at 8MP. P2P’s measurements have exactly zero real world use. Not only that they’re a raw score. Ignoring things like PDAF banding etc that does affect real world USABLE DR. You get better information from DP Reviews highly flawed test chart. Those using P2P’s charts to prove a point only prove they have no experience with the cameras they’re procrastinating about.

According to P2P an iPhone 14 has as much DR as an A7R5 at ISO 6400. If you believe that I’ve got a bridge to sell you.

Gordon

I still maintain that the main problem with P2P is not its measurements but how people misread it (the way a drunk person uses a lamp-post). The main issue with the PDR chart is that the ISO is not calibrated: when comparing cameras, one does not know whether the graph needs to be moved left or right.

However, the PDR chart is only one of many published charts, though it is the easiest one to misunderstand 😜. Other interesting charts are the read noise, PTC, sensor heatmaps, etc.

Sidenote: I sometimes consult DPR test charts, though I would caution anyone using them. The test images regularly use different exposures and have misfocus or camera vibrations. Also, the DPR score is useless to me.

Yes, the iPhone 14 has more PDR than the A7r5 at ISO 6400 (PDR is a special version of DR, like EDR is a special one). The reason is twofold, as depicted in the PDR graph: heavy NR and the iPhone stopping increasing ISO after ISO 400. At ISO 6400, we are comparing with the iPhone at ISO 400 and with lots of NR, while A7r5 is at ISO 6400.

In that context, it would be smarter if manufacturers stopped increasing ISO at a relevant high ISO number and instead used a metatag to adjust brightness. That would add a lot of flexibility in post-processing and capture.

 

[BernardC]

17 minutes ago, SrMi said:

That is wrong. Here is Leica's table, p337, from SL3's manual. While Leica claims ISO 50 to be the (first) native ISO, I would avoid it.

It depends on the camera and how it is setup (where middle-grey is placed on the scale). I said "approximately 800," but evidently Leica picked EI 640 instead, which is 1/3 of a stop less. Close enough?

Leica's video EIs are higher because video is a transmissive format (prints are reflective), so you need more highlight range. The "not-log" video EI is for SDR, which is for very old sets that have a severely limited highlight range. In other words: SDR can't display bright highlights, so you don't need to protect for highlight clipping in your source files (they would clip in your output file anyway).

[SrMi]

5 minutes ago, BernardC said:

It depends on the camera and how it is setup (where middle-grey is placed on the scale). I said "approximately 800," but evidently Leica picked EI 640 instead, which is 1/3 of a stop less. Close enough?

Leica's video EIs are higher because video is a transmissive format (prints are reflective), so you need more highlight range. The "not-log" video EI is for SDR, which is for very old sets that have a severely limited highlight range. In other words: SDR can't display bright highlights, so you don't need to protect for highlight clipping in your source files (they would clip in your output file anyway).

Do we agree that the two "native" ISOs for still photography with SL3 are 50/100 and 320?  You referenced P2P which is relevant only for still, not for video.

[ZHNL]

6 hours ago, FlashGordonPhotography said:

You don’t. How can you not take resolution into account when choosing a sensor? And nobody is deciding between an X2D and an iPhone. It’s a stupid metric.

I’d give it some consideration, despite its limitations as an accurate indicator, if they made the DR measurements at several output sizes. But they don’t 8MP is a useless measurement. People read this and make judgments on sensor IQ and those assumptions are often wrong becausee the data’s flawed for real world use.

And as an experiment I did compare my A7R5 to an iPhone at ISO 6400. Not even close. Even at 8MP. Not to mention the phone relied on computational factors to keep its DR reasonable. So not actually a real sensor measurement at all. But P2P doesn’t mention that.

Gordon

This discussion is only for one aspect of the sensor, DR. Period. 
Who claim you buy camera based on only DR? And who claimed you should be in the market of decision based on DR X2D iso6400 vs iPhone base? Please don’t change topic direction this way. 
Why you need make purchase decision based P2P? The stuff we discussed is fundamental and objective. 

[BernardC]

Advertisement (gone after registration)

18 minutes ago, SrMi said:

Do we agree that the two "native" ISOs for still photography with SL3 are 50/100 and 320? 

I explained why. Still photography is traditionally printed, so "paper white" is just 2.5 stops past middle grey. That means you don't need much dynamic range above middle grey (it can't be printed). The recommended video EI places middle grey deeper (darker) because projectors and HDR TVs are much brighter, and clipped highlights look terrible when projected. The camera is doing the exact same thing in either case, and total dynamic range is the same, the only difference is the number that we define as middle-grey.

Many people explain this using completely different terms: "I exposed at base ISO and I was able to lift the shadows by 6 stops, but the highlights clipped early!" "I exposed at a higher EI and then I couldn't lift the shadows as much, but the highlights didn't clip." It's a different way of describing the same elephant.

[ZHNL]

3 hours ago, BernardC said:

Great answer.

Every commercial raw format has been encoded on a  log scale, beginning with EXR (motion picture film scanning). Film density (HD curves) and sound pressure (dB) are also encoded that way. The alternative would be linear encoding, where the least significant bit would take-up half of the possible values.

EI 6400 is in the second gain range of the sensor, so base EI isn't 100 in this case, but closer to 800. The AD converter doesn't throw-out higher values to get to higher speeds. They are still there, you just get a different middle-grey value with more highlight range and less shadow range (because middle grey is deeper into the shadows).

Check P2P curve, 2nd gain point is at iso 340-360 region and at iso 400, you get 10 stop DR. Rest become a simple math first order calculation. 
the key point is not how many DR in number. As mentioned, if you don’t have clear definition of acceptable noise level, or target output size, you can’t compare DR. P2P provides both information and offer a platform so that you can compare DR cross platform. 
To Gordon’s point, you can certainly compare m11 vs X2D at output 64M resolution to make purchase decision but that info can be directly interpolate from P2P curve. 

Edited April 22, 2024 by ZHNL

[SrMi]

2 minutes ago, BernardC said:

I explained why. Still photography is traditionally printed, so "paper white" is just 2.5 stops past middle grey. That means you don't need much dynamic range above middle grey (it can't be printed). The recommended video EI places middle grey deeper (darker) because projectors and HDR TVs are much brighter, and clipped highlights look terrible when projected. The camera is doing the exact same thing in either case, and total dynamic range is the same, the only difference is the number that we define as middle-grey.

Many people explain this using completely different terms: "I exposed at base ISO and I was able to lift the shadows by 6 stops, but the highlights clipped early!" "I exposed at a higher EI and then I couldn't lift the shadows as much, but the highlights didn't clip." It's a different way of describing the same elephant.

Native ISO has nothing to do with the output medium but the raw data. If you want the most information in the SL3's DNG (least noise, most details), you shoot at ISO 100 with maximized exposure that does not clip highlights. Post-processing is responsible for matching that data into the output medium.

[BernardC]

3 minutes ago, SrMi said:

Native ISO has nothing to do with the output medium but the raw data.

Note that Leica calls it "base ISO" not "native ISO." They call it base because that's the lowest ISO you should use. I wish they used the term EI instead of ISO, but that's nit-picking. The main point to remember is that changing "ISO" on your camera just changes the value that your raw converter will use as middle grey (within each "gain" range).

[SrMi]

3 minutes ago, BernardC said:

Note that Leica calls it "base ISO" not "native ISO." They call it base because that's the lowest ISO you should use. I wish they used the term EI instead of ISO, but that's nit-picking. The main point to remember is that changing "ISO" on your camera just changes the value that your raw converter will use as middle grey (within each "gain" range).

For still photography:

Changing the ISO value does not only change what middle grey will be used unless ISO is fixed and metadata is used to adjust the brightness (very rare). With fixed exposure, increasing ISO reduces noise and reduces the maximum possible exposure because of clipping.

[BernardC]

1 hour ago, SrMi said:

For still photography:

Changing the ISO value does not only change what middle grey will be used unless ISO is fixed and metadata is used to adjust the brightness (very rare). With fixed exposure, increasing ISO reduces noise and reduces the maximum possible exposure because of clipping.

That's not how modern sensors work. It's really easy to test, so don't take my word for it. Take two images using the same manual exposure, one at EI 100 and one at EI 200. Open them both in your favourite raw converter and match the exposure (it's probably a slider). They will be absolutely the same. I've done this with recent Leica cameras, but the outcome should be the same with any camera made in the past 15 years or so. The only exception is for dual gain sensors, if you switch between gain settings. EI 100 and 200 should both use the same signal path, but 400 and 800 might not.

[SrMi]

31 minutes ago, BernardC said:

That's not how modern sensors work. It's really easy to test, so don't take my word for it. Take two images using the same manual exposure, one at EI 100 and one at EI 200. Open them both in your favourite raw converter and match the exposure (it's probably a slider). They will be absolutely the same.

Yes, but that is not the point. Why would I waste exposure at ISO 100? At ISO 100, I can use a twice-long exposure as with ISO 200 without clipping highlights. Therefore, I can have better DR at ISO 100. Of course, you can reduce exposure so that you do not have any clipping at ISO 200, but that is wasteful.

Remember that noise is determined by exposure.

Edited April 22, 2024 by SrMi

[BernardC]

13 minutes ago, SrMi said:

Remember that noise is determined by exposure.

Yes, and you are using exactly the same f-stop and shutter speed in the test I outlined, so the sensor sees exactly the same amount of light.

Anyway, it's up to you if you want to test this using your own camera and software. Either you get unrecoverable highlight clipping at EI 200, or you don't.

[Chaemono]

Reid Reviews did this test to determine the native ISO of the SL3.

[FlashGordonPhotography]

7 hours ago, SrMi said:

That is wrong. Here is Leica's table, p337, from SL3's manual. While Leica claims ISO 50 to be the (first) native ISO, I would avoid it.

Welcome, dear visitor! As registered member you'd see an image here…

Simply register for free here – We are always happy to welcome new members!

Wow. That is ambiguous.

Elsewhere Leica have said base ISO is 100 and 50 is a pull. You are absolutely correct to avoid ISO 50.

Gordon

[ZHNL]

11 hours ago, BernardC said:

That's not how modern sensors work. It's really easy to test, so don't take my word for it. Take two images using the same manual exposure, one at EI 100 and one at EI 200. Open them both in your favourite raw converter and match the exposure (it's probably a slider). They will be absolutely the same. I've done this with recent Leica cameras, but the outcome should be the same with any camera made in the past 15 years or so. The only exception is for dual gain sensors, if you switch between gain settings. EI 100 and 200 should both use the same signal path, but 400 and 800 might not.

Can you clarify what you mean the same manual exposure? did you take half the exposure time of ISO100 to match the exposure or you match the manual exposure time? 

There are 4 major noise before light become raw digital number: photon shot noise, sensor electric thermal noise, Read noise and quantization noise for ADC. in general, you will always want to using base ISO to minimize the noise. High ISO reduce the exposure time for the same exposure at cost of reduce light/signal goes into sensor.

Let's using one example to illustrate:

Base ISO100, correct exposure, DR = Signal/(shot+themral+read+Quantization) (equation is for illustration only, uncorrelated noise should be RSS)

ISO200 DR= 2X(1/2 Signal)/( more shot noise +2X(1/2 Thermal) +read + Quantization + gain noise (very low for modern sensor))=Signal/(more shot noise + thermal + read + Quantization+ gain noise (very low for modern sensor))

so DR degrade. 

For Morden sensor, what you could do to achieve same DR:

ISO 100 under exposure by 1 stop but do post process in raw converter to boost exposure by 1 stop. This will match ISO 200 correct exposed DR (which will be worse than ISO 100 correct exposed image as illustrate example above)

In theory, even this case, you will still prefer using ISO100 under exposure and do post process because even modern analog ISO/gain has low noise, it will still add additional noise to final image. Post Process is pure digital gain, no noise added. 

I don't want to further complicate this matter to talk about dual gain case but we should take advantage of dual gain whenever it is possible. to get best DR, you can in theory only use two ISO, base and dual gain ISO and do rest in post.  

Anyway, this seems complicated topic, implementation details may vary but first order underneath concepts are all fundamental EE and signal processing concepts. 

[BernardC]

5 hours ago, ZHNL said:

Can you clarify what you mean the same manual exposure? did you take half the exposure time of ISO100 to match the exposure or you match the manual exposure time? 

I mean the exact same. If the first shot is 1/125 at f:8.0, then the second shot is also 1/125 at f:8.0. The only difference is the ISO setting.

Some people claim that the second shot (where the camera is set to EI 200) will have one stop less dynamic range, compared to the first shot (where the camera was set to EI 100). This contradicts real-world tests done by myself and others.

If you try this test and find that you've lost dynamic range, can you tell me which raw converter you used, and where the DR was lost (highlights, shadows, split between both?).

[Advertisement]

Hallo,

Look here: Leica L-Mount

[SrMi]

2 hours ago, BernardC said:

Some people claim that the second shot (where the camera is set to EI 200) will have one stop less dynamic range, compared to the first shot (where the camera was set to EI 100). This contradicts real-world tests done by myself and others.

Nobody claims that ISO 100 shot will have more DR than ISO 200 if you keep the exposure constant.

[BernardC]

1 hour ago, SrMi said:

Nobody claims that ISO 100 shot will have more DR than ISO 200 if you keep the exposure constant.

That's what the PtP graphs show.

Exposure is a red herring in this case. Either you get 12 stops of DR or you don't. Exposure only determines which 12 stops you capture (for instance if you had a perfect scene that went from absolute blackness to supernova, with a constant gradient between the two).