ISO performance on SL2-s vs SL2

[hansvons]

Advertisement (gone after registration)

10 hours ago, SrMi said:

Blown highlights in still photography are caused by metering.

Same in digital cinematography. Metering to protect the highlights by two stops is the same as "regular" metering with a higher ISO, e.g. ISO 400. 

 

10 hours ago, SrMi said:

With L-log, does ISO 800 correspond to ISO 100 in still mode? I notice that the lowest ISO for L-Log is ISO 400 (the lowest ISO in Photo mode is ISO 50).

Without more profound insights into maths and technology, I'd think that ISO 50 is a marketing-ISO, sacrificing DR. ISO 100 seems to be the minimal ISO without losing DR. However, at ISO 100, chances are high that highlights will be blown out as one operates the sensor at its upper limit. With ISO 200, you gain 1 "safety" stop, and at ISO 400, 2. Why would one not do that? To battle noise which isn't there at ISO 200/400? I chose ISO 800 as I like texture and prefer some leeway. The SL2-S provides astonishing juice in the shadows even at higher ISO, which is, IMHO, its most significant advantage over the SL2. 

Edited April 11, 2023 by hansvons

[Advertisement]

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

[drjonb]

3 hours ago, hansvons said:

Same in digital cinematography. Metering to protect the highlights by two stops is the same as "regular" metering with a higher ISO, e.g. ISO 400. 

 

Without more profound insights into maths and technology, I'd think that ISO 50 is a marketing-ISO, sacrificing DR. ISO 100 seems to be the minimal ISO without losing DR. However, at ISO 100, chances are high that highlights will be blown out as one operates the sensor at its upper limit. With ISO 200, you gain 1 "safety" stop, and at ISO 400, 2. Why would one not do that? To battle noise which isn't there at ISO 200/400? I chose ISO 800 as I like texture and prefer some leeway. The SL2-S provides astonishing juice in the shadows even at higher ISO, which is, IMHO, its most significant advantage over the SL2. 

I agree completely with your comments here.  SL2-S has great range and you can really bring out the dark areas with shadow slider in Capture One.  I use 400 and 800 ISO and I know the texture you are talking about, very satisfying.  This is the result of halving the number of pixels per 35mm sensor but doubling the pixel size.  Horses for courses…if you need to print large and need lots of detail/resolution, more and smaller pixels per given sensor diagonal of course helps - but expensive to manufacture.  But for me, the SL2-S sensor is more than adequate.  

[BernardC]

17 hours ago, SrMi said:

M11, SL2, Sony, Nikon, etc, use dual conversion gain sensors. Arris and other video cameras use dual output gain sensors. See this article:

What is dual gain and how does it work?

I'm not sure why you keep referring to this article. It plainly states that only older sensors applied increasing gain at each ISO, and that the Panasonic S1H, and Leica, and Nikon, and Sony, and Fuji, and others, use dual conversion gain similar to what is described in Panasonic's web page. It mentions the GH6 as an oddball because it combines the output from both read modes into the same file, and mentions that the GH5 doesn't do this (nor do the other MFT cameras mentioned).

The end result is very similar anyway: you get a "DR budget" which you can spend on highlights or shadows, but not on both.

[SrMi]

2 hours ago, BernardC said:

I'm not sure why you keep referring to this article. It plainly states that only older sensors applied increasing gain at each ISO, and that the Panasonic S1H, and Leica, and Nikon, and Sony, and Fuji, and others, use dual conversion gain similar to what is described in Panasonic's web page. It mentions the GH6 as an oddball because it combines the output from both read modes into the same file, and mentions that the GH5 doesn't do this (nor do the other MFT cameras mentioned).

The end result is very similar anyway: you get a "DR budget" which you can spend on highlights or shadows, but not on both.

I am reading different facts from that article:

a) Dual output gain (Panasonic) is not the same as dual conversion gain (Aptina).

b) Arri and GH6 use dual output gain (Panasonic), while Nikon, Sony, etc use dual conversion gain (Aptina).

[SrMi]

8 hours ago, hansvons said:

Without more profound insights into maths and technology, I'd think that ISO 50 is a marketing-ISO, sacrificing DR. ISO 100 seems to be the minimal ISO without losing DR

Jim Kasson has analyzed ISO 100 on Q2M (same as ISO 50 on SL2) and concluded:

a) you can get up to a third of a stop more engineering dynamic range

b) plenty of minuses (unreliable histogram, etc.).

Should you use ISO 100 on the Leica Q2 Monochrom?

8 hours ago, hansvons said:

However, at ISO 100, chances are high that highlights will be blown out as one operates the sensor at its upper limit. With ISO 200, you gain 1 "safety" stop, and at ISO 400, 2.

I do not understand that paragraph. If I clip highlights at ISO 100, f/8, 1/100, I will also clip highlights at ISO 200, f/8, 1/200. Highlight clipping is caused not only by saturated sensels but also by the ISO "amplification." What do you mean by a safety stop?

Ideally, in ISO invariant ranges, ISO change would only change a tag, not raw data. That way, highlight clipping would be under the control of the user, and we would have the "highlight safety stops" that you talked about. The only way to get highlight safety stops in SL2 is to expose for the given high ISO but apply negative EC to ISO.

8 hours ago, hansvons said:

The SL2-S provides astonishing juice in the shadows even at higher ISO, which is, IMHO, its most significant advantage over the SL2. 

Yes!

Some SL2 disadvantages may be mitigated by AI NR but not all. The more extreme the shadow lifting, the more SL2-S shines.

Edited April 11, 2023 by SrMi

[jaapv]

9 hours ago, hansvons said:

Same in digital cinematography. Metering to protect the highlights by two stops is the same as "regular" metering with a higher ISO, e.g. ISO 400. 

 

Without more profound insights into maths and technology, I'd think that ISO 50 is a marketing-ISO, sacrificing DR. ISO 100 seems to be the minimal ISO without losing DR. However, at ISO 100, chances are high that highlights will be blown out as one operates the sensor at its upper limit. With ISO 200, you gain 1 "safety" stop, and at ISO 400, 2. Why would one not do that? To battle noise which isn't there at ISO 200/400? I chose ISO 800 as I like texture and prefer some leeway. The SL2-S provides astonishing juice in the shadows even at higher ISO, which is, IMHO, its most significant advantage over the SL2. 

Actually, the changing of the ISO value does not occur on a sensor level but in the firmware processing chain after ADC conversion. The sensor only records the number of photons that it captures. 

[BernardC]

Advertisement (gone after registration)

2 hours ago, SrMi said:

Arri and GH6 use dual output gain (Panasonic), while Nikon, Sony, etc use dual conversion gain (Aptina).

The article specifically mentions the S1H and "Leica" in that category. One assumes that this includes Leica cameras that use similar sensors to the S1H (Q2, SL2).

[SrMi]

12 minutes ago, BernardC said:

The article specifically mentions the S1H and "Leica" in that category. One assumes that this includes Leica cameras that use similar sensors to the S1H (Q2, SL2).

S1H has a different sensor than Q2, SL2, or S1R.

[SrMi]

1 hour ago, jaapv said:

Actually, the changing of the ISO value does not occur on a sensor level but in the firmware processing chain after ADC conversion. The sensor only records the number of photons that it captures. 

In many cases, ISO is applied before ADC conversions:

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!

[jaapv]

 But after the sensor dump. Exactly like a volume control on your audio amplifier. 

[SrMi]

5 minutes ago, jaapv said:

 But after the sensor dump. Exactly like a volume control on your audio amplifier. 

Yes, after the sensor dump, which was the main point of your post.

[Ken Abrahams]

On 4/4/2023 at 2:33 AM, Sohail said:

Hi all, 

I'm seriously thinking of buying an SL2-s as a second body to my SL2.

A question: is the ISO performance on SL2-s markedly better than SL2. For my tastes, anything above 3200 on SL2 is not great. 6400 is so-so.

Thanks in advance.

Sohail

I purchased the SL2s (Reporter) today - it was what I wanted from the start, so thanks to the discussion here I will alternate with the two cameras to suit the scenes and desired outcomes.  

Ken 

[jaapv]

15 hours ago, SrMi said:

Yes, after the sensor dump, which was the main point of your post.

So no change to the sensor or data from it, but in the subsequent process. 

[BernardC]

17 hours ago, SrMi said:

S1H has a different sensor than Q2, SL2, or S1R.

True, but it uses a similar sensor to the S5, SL2-S, and fp. The sensor's origin is widely believed to be the same as the S1R, Q2, SL2.

That's mostly nit-picking. The main point is that the article explains two different ways of using "dual gain:" one way uses one gain setting for lower ISO, and another for higher ISOs. The other combines the two outputs at all ISOs.

According to the article, the second method is hardly ever used. Panasonic uses it in one (very specialized) camera.

In the end, they boil-down to the same thing: DR does not vary with small changes in ISO (it does when you switch signal paths in the first method, usually somewhere past ISO 1600). Changing ISO moves your middle grey mapping, so you get more shadow range at low ISOs, and more highlight range at higher ISOs.

[SrMi]

 

3 hours ago, BernardC said:

In the end, they boil-down to the same thing: DR does not vary with small changes in ISO (it does when you switch signal paths in the first method, usually somewhere past ISO 1600). 

With SL2 and most still cameras, the DR changes significantly at the dual conversion gain point, typically between ISO 200 (M11) and ISO800 (SL2-S). Since ISO change modifies the raw data, the maximum possible exposure is decreased as ISO increases. Therefore, since exposure determines the DR (visible noise), the top possible DR will be reduced with the increasing ISO.

3 hours ago, BernardC said:

Changing ISO moves your middle grey mapping, so you get more shadow range at low ISOs, and more highlight range at higher ISOs.

With Leica cameras, changing ISO modifies the middle grey mapping (brightness) and the raw data stored. So if you have highlight clipping three stops above the middle grey, you will have it at any ISO. Did I misunderstand what you mean by highlight range?

P.S.: You can look at raw data using a tool like RawDigger.

[BernardC]

11 minutes ago, SrMi said:

With Leica cameras, changing ISO modifies the middle grey mapping (brightness) and the raw data stored. So if you have highlight clipping three stops above the middle grey, you will have it at any ISO. Did I misunderstand what you mean by highlight range?

That isn't my experience. I've compared files with the same exposure (aperture, shutter speed) and different ISO and compensation values, and they are the same. That's not to say that either of us is wrong, we probably don't use the same camera model, firmware version, raw converter, etc.

[hansvons]

5 hours ago, BernardC said:

That isn't my experience. I've compared files with the same exposure (aperture, shutter speed) and different ISO and compensation values, and they are the same. That's not to say that either of us is wrong, we probably don't use the same camera model, firmware version, raw converter, etc.

That mirrors my experience. To me, with C1, the SL2S sensor is invariant until ISO 1600 or 3200 (I can't remember the exact ISO, tested the sensor at the beginning of 2020). YMMV.

[SrMi]

11 minutes ago, hansvons said:

That mirrors my experience. To me, with C1, the SL2S sensor is invariant until ISO 1600 or 3200 (I can't remember the exact ISO, tested the sensor at the beginning of 2020). YMMV.

I have shown that SL2-S is not ISO invariant from ISO 100 to ISO 3200. The P2P measurements confirm that. 
Can you demonstrate the ISO invariance from ISO 100 to ISO 1600?

P.S.: we are talking about still photography, correct?

Edited April 12, 2023 by SrMi

[Advertisement]

Hallo,

Look here: Leica L-Mount

[BernardC]

13 hours ago, SrMi said:

I have shown that SL2-S is not ISO invariant from ISO 100 to ISO 3200. The P2P measurements confirm that. 

Did I miss the part where you showed that? Do you mean that 100 and 3200 use different gain circuits, or that 100 is different from 200?

Have you tried the test that I suggested: pick a manual exposure, take a series of images at that exposure, switching the ISO between shots.

PtP isn't measuring what they think they are. Their method isn't well suited to measuring dynamic range. I also find it strange that their DR numbers go down by almost a full stop for every extra stop of ISO. Don't you think that manufacturers simply wouldn't bother with variable gain if it made images worse? Think about it, if boosting gain by one stop has the effect of losing one stop of DR, why would you do it? The DPR article that you linked mentions that digital cameras have not worked that way for a quite a while.

 

[BernardC]

On 4/11/2023 at 12:30 PM, SrMi said:

Jim Kasson has analyzed ISO 100 on Q2M (same as ISO 50 on SL2) and concluded:

a) you can get up to a third of a stop more engineering dynamic range

b) plenty of minuses (unreliable histogram, etc.).

Should you use ISO 100 on the Leica Q2 Monochrom?

That page says the same thing that others have been arguing: "(ISO 100) does not offer a material advantage over ISO 200". They are the same, so you may as well use 200. I'm not sure what he means by "engineering dynamic range." Is it a real thing, or his own name for something else? The fact that he found only a third of a stop of difference points toward experimental error. As we know, 1/3 stop is the minimum noticeable difference.