What do you think the base ISO speed of the X-Trans 3 Sensor is? Why do you think that?

[Astigmatism]

I have a new X-T30 II and am learning about it. One mystery is the base ISO of the X-Trans 3 Sensor. I haven't anyplace in a reference from Fujifilm, including my user manual, that says. I also haven't noticed anything at PhotonsToPhotos, or in my RawDigger software, either. But there are many online references that say, and they seem equally split between 160 and 200, with a few scattered oddballs like 640 or 800 or 1000.

 

FWIW, I think the working definition of "base ISO" would be "the ISO setting that puts the metering point at 18% (or 12%) of full scale in the green channel when metering off of an 18% gray card, while simultaneously causing a 100% white Lambertian reflectivity card next to the gray card to give raw pixel values of 2^14 in the green channel, and just saturating the sensor wells (i.e. onset of clipping by the wells)." What I have been able to find doesn't even make it clear that there is a literal definition of "base ISO" from the International Standards Organization themselves.

 

What do you think?

 

Thank you!!

[Greybeard]

 Always assumed 160 because it's the lowest setting without going to an artificial low.

[Astigmatism]

54 minutes ago, Greybeard said:

the lowest setting without going to an artificial low

As indicated how, the "L" that appears on the settings below that? Does Fuji say "artificial" somewhere?

[jerryy]

The 160 vs 200 you are seeing refers to the lowest ISO Fujifilm lists as the lowest number in the standard sensitivity range for the camera. Prior to the X-T3, this is ISO 200, with the improvements due to the new sensor in the X-T3 and later models, this became ISO 160.

For example:

https://fujifilm-dsc.com/en/manual/x-t2/technical_notes/spec/index.html  vs

https://fujifilm-x.com/global/products/cameras/x-t3/specifications/

The L and H refers to ISO settings outside of this range, which are “artificial”  because they are outside of the normal analog range for the sensor. Remember the sensors come with ADC (analog to digital convertor) chips.

https://photonstophotos.net/Charts/PDR.htm

Choose X-T3 or X-T4 or both for the chart. The results outside of the native range are software enhanced. They are very good results though.

Your X-T30 II is very close to the X-T4:

https://www.apotelyt.com/compare-camera/fujifilm-x-t30-ii-vs-fujifilm-x-t4

I do not think you will find an agreed to definition of base ISO similar to how ASA is defined for film. That ship snuck out of the harbor and sailed away a long time ago because the each of the early digital camera manufacturers essentially said ‘ this is what we want it to be and that is the way it is ‘ and there was no standards body with jurisdiction to say otherwise. At least today, the major manufacturers are sort of close to each other in how they use ISO.

[Astigmatism]

Great answer jerryy! Thank you!

I've wondered about several of the charts there on photonstophotos. Several of the charts show a sudden change between ISO 640 and 800. In particular the "Input-referred Read Noise versus ISO Setting" chart, in the first link below, has a step there that is overwhelming compared to the range of data nearly everywhere else. I don't know what to make of them. Maybe there's a shift between two different A/D amplifiers each of which has multiple gains, maybe there's also a digital change introduced there. I've tried to make sense of this by looking at RawDigger histograms (which are fascinating in several ways) -- do they start to have a pattern of empty bins? Or what they call "pits", nearly empty bins?

Any thoughts on this?

https://www.photonstophotos.net/Charts/RN_e.htm

https://www.photonstophotos.net/Charts/RN_ADU.htm

[jerryy]

There are several things going on. The recent Fujifilm cameras use two different approaches to ISO gain, one for the low ISO range and a different one for the higher ISO range. The changeover seems to occur at ISO 800, which is why those charts seem so odd as compared to the ones for other manufacturers — I do not completely know why ISO 800 is the changeover point. This can get complicated by the fact that Fujifilm uses an approach called ISO invariance (some other manufacturers also do this lately). This chart may help illustrate this:

https://www.photonstophotos.net/Charts/PDR_Shadow.htm#FujiFilm X-T3

At the low ISO range range, you will get an increase in shadow detail by bumping up the ISO a bit, but once you move into the higher ISO range, you do not get improvements in shadow detail by bumping up the ISO. What you can get by upping the ISO is an improvement by lowering read noise (random noise — some call it garbage — that is present in all electronic sensors from the small current needed to operate the sensor) but you trade that against lowering the available dynamic range. How much is lost is what those other charts are telling you.

This article may interest you:

https://www.photonstophotos.net/Aptina/DR-Pix_WhitePaper.pdf

 

Edited April 7, 2022 by jerryy

[Astigmatism]

I appreciate your post! Thank you! This is all fascinating, including the white paper. I also get the idea that on this sensor an ISO setting of 800 is a locally sweet spot -- that is, there seems little reason to pick 640, as it is a local "sour spot".

My impression is that "ISO invariance" amounts to a claim that the sensor has much more range than jpg files or monitors do, so that users can choose their ISO setting by whatever criteria they like without any image degradation. Maybe more than inkjet photo printing too, I mean to look that up. "ISO invariance" doesn't seem to be a specific technical feature, though. Is it?

[Astigmatism]

More on inkjet photo printing: lots of online references say they can deliver around 7 or 8 stops of dynamic range. Here are two discussions with lots of details:

https://forum.luminous-landscape.com/index.php?topic=44370.0

https://www.dpreview.com/forums/thread/2254402

For a long time I've daydreamed about a printing method to get very high dynamic range: print transparencies, one backwards, and laminate them together with careful alignment. Then illuminate them from behind with a bright light, maybe uncomfortably bright in the smallest highest highlights. I've also thought about making that bright light from three nearly monochromatic light sources chosen to increase the color gamut of the printing ink (you could certainly increase the gamut by filtering out the wavelengths that contribute least to color distinction).

[jerryy]

6 hours ago, Astigmatism said:

… "ISO invariance" doesn't seem to be a specific technical feature, though. Is it?

Yup, it is. The various search engines can direct you to lots of explanations and discussions, some very technically oriented, some less so. Take a look, it is out there.

[BobJ]

ISO invariance is a bit of a myth. All sensors are effectively ISO invariant. ISO is inherited from methods used to measure the sensitivity of film and is used in digital to indicate the amount of amplification applied to the charge accumulated on the capacitors in  the photodiodes. Think of it like turning up the volume on an am radio.  Each capacitor is charged during the exposure. When the capacitor is fully charged the brightest tone that can be shown is fixed. If the ISO is raised, the exposure is stopped before the capacitor is fully charged , hence the darkest tone, set by the noise, is pushed up the scale and dynamic range is reduced. It is possible to use more than one capacitor, each one having a value best suited to the ISO chosen. I think this is what is happening here. At ISO 800 the diode is switched over to a smaller value capacitor. I don't actually know that for sure though, but as a retired engineer it seems to make the most sense to me. By the way, circuit noise is almost non existent compared with the 'shot noise' from the random nature of the photons hitting the photodiodes. That is why bigger sensors with larger photodiodes exhibit less noise. The more photons captured the lower the noise. The base ISO is the sensitivity that the manufacturer has set to make best use of the maximum dynamic range possible with the capacitor and amplifier and is always the lowest value shown on the ISO control that is not an 'extended' one (up or down). Upping the ISO will not result in lower noise. You can easily prove that yourself with a bit of pixel peeping on the raw files (not jpeg as the noise reduction algorithm will confuse things).

[jerryy]

https://chrisbonney.uk/blog/iso-invariance-explained-better/

A bit more explanation.