Sensor sensitivity

[JonathanBenoit]

Microlenses do not soften the image, other than by potentially reducing aliasing, and thus modifying the MTF.
Perhaps you are confusing them with an anti-aliasing filter, AKA optical low-pass filter (OLPF)?

Cheers,
Bart

You are correct. I mostly use a view camera for my work, so I did confuse the two. My brain has been trained to hate microlenses. I also have no use for ISO over 100, unless I was to capture aerial landscapes.

[AJSJones]

You are correct. I mostly use a view camera for my work, so I did confuse the two. My brain has been trained to hate microlenses. I also have no use for ISO over 100, unless I was to capture aerial landscapes.

Is there something about ISO100 that your brain has been trained to love?  Or is this based on experience with film and earlier generations of digital sensors?  Perhaps you need long exposure times for some aspect of your work?

[JonathanBenoit]

Is there something about ISO100 that your brain has been trained to love?  Or is this based on experience with film and earlier generations of digital sensors?  Perhaps you need long exposure times for some aspect of your work?

I'm always working with a tripod, mirror up, cable release, longer exposures than average. I actually shoot at ISO 50 with my H DB. I don't need to use anything faster. I have it easier than most. I'm sure there are a good deal of photographers like me. To have a camera that's base ISO is 200 is just silly. Is Pentax noise reduction at ISO 200 as good as Hasselblad or Phase ISO 100? I have no idea. It really doesn't make much sense that Pentax doesn't have base ISO 100. From the Landscape or Architectural photographer's point of view, it's a turn off. I can understand the appeal of high ISO for handheld photography. I'm sure the Pentax is geared toward this type of photographer specifically.

[ErikKaffehr]

Hi,

Low ISO gives normally the best quality, simply by collecting more photons. The best Signal Noise Ration (SNR) is always achieved exposing so the the pixels are near saturation (full well capacity) for the highlights. DR generally increases with exposure.

The DxO-measurements illustrate this quite handily:

[attachment=21514:Screen_s...37.09_AM.png]

Note the "hump" at 800 and 1600 ISO on the Phase One P65+, that is the effect of Sensor+ technology.

All three cameras achieve maximum SNR at minimum ISO and the SNR diminishes with similar slope for increasing ISO. The Nikon D3X has about one stop advantage which is consistent with it having micro lenses. This comparison is per pixel. The MFDBs have about twice the number of pixels and this would give an advantage in similarly sized prints:

[attachment=21515:Screen_s...54.54_AM.png]

Note also that in the normalized print size view the Sensor+ hump disappears.

The DxO measurements only measure sensor data and the SNR is just one of their measurements. Their measurements only relate to sensor data. Vendor software may have access to more data like individual camera calibration data so it may achieve better results.

Best regards
Erik

Is there something about ISO100 that your brain has been trained to love?  Or is this based on experience with film and earlier generations of digital sensors?  Perhaps you need long exposure times for some aspect of your work?

[AJSJones]

I'm always working with a tripod, mirror up, cable release, longer exposures than average. I actually shoot at ISO 50 with my H DB. I don't need to use anything faster. I have it easier than most. I'm sure there are a good deal of photographers like me. To have a camera that's base ISO is 200 is just silly. Is Pentax noise reduction at ISO 200 as good as Hasselblad or Phase ISO 100? I have no idea. It really doesn't make much sense that Pentax doesn't have base ISO 100. From the Landscape or Architectural photographer's point of view, it's a turn off. I can understand the appeal of high ISO for handheld photography. I'm sure the Pentax is geared toward this type of photographer specifically.

Your goal, and that of many others as you say, is obviously the absolute lowest possible noise (highest SNR). You can achieve this if you can shoot 50, based on Erik's examples. My question went towards why you picked 100 as acceptable and 200 not.

[ErikKaffehr]

Hi,

My view is simply that you want to absorb as many photons in the sensor as possible. That is always achieved by maximum exposure. You want to do that without clipping non-specular highlights, that's called expose to the right.

One interesting observation may be that on both the Nikon D3X and the Phase One P65+ the two lowest ISO are in reality identical.

[attachment=21516:Screen_s...34.54_AM.png]

The noise plots on DxO are based on real ISO.
[attachment=21517:Screen_s...54.54_AM.png]

The two lowest ISOs are plotted as a single point in the figure above. What this says is that ISO 50 and ISO 100 on the P65+ give the exactly same results if you expose to the right! The software may do different tricks based on ISO, however.

It strikes me how similar the sensors are. Other parameters tested by DxO show larger differences.

BR
Erik

Your goal, and that of many others as you say, is obviously the absolute lowest possible noise (highest SNR). You can achieve this if you can shoot 50, based on Erik's examples. My question went towards why you picked 100 as acceptable and 200 not.

[BJL]

Low ISO gives normally the best quality, simply by collecting more photons.

True when comparing different ISO settings on the same sensor; not true when comparing different sensors of different QE. For example, adding or improving the efficiency of microlenses over the same basic photosite design with the same full well capacity will increase the base ISO but, by the definition of base ISO, exposing each of these different sensor versions at base ISO and at shutter speeds adjusted in proportion to the ISO speed difference will collect equal numbers of photons: roughly, photosites at metered mid-tones will get about 1/10 of full well capacity.

[AJSJones]

So if one makes the "perfect" exposure based on the desired image, and this does not take advantage of the whole range of the sensor (by not exposing to the right) has one lost a significant part of the advantage of the low ISO at 50 or 100?  A shot exposed to the right at 200 ISO and darkened in PP may well be "cleaner" than one exposed correctly at ISO 100, might it not?  I would be more inclined to focus on exposing to the right all the time if I wanted maximum DR and "clean" images, no matter what ISO.  Is this what is commonly done by those who prefer using the lowest ISO possible?

[bjanes]

Hi,

One interesting observation may be that on both the Nikon D3X and the Phase One P65+ the two lowest ISO are in reality identical.
The noise plots on DxO are based on real ISO.

The two lowest ISOs are plotted as a single point in the figure above. What this says is that ISO 50 and ISO 100 on the P65+ give the exactly same results if you expose to the right! The software may do different tricks based on ISO, however.

It strikes me how similar the sensors are. Other parameters tested by DxO show larger differences.

BR
Erik

Erik,
As you note, the lowest ISO with the D3x is merely overexposure. If one keeps the f/stop and shutter speed the same, the raw files are identical. This shows that the electronic amplification has not changed and one is merely overexposing and losing "highlight protection". Nikon allows 0.5 EV headroom for "highlight protection", so many times one can get away with this overexposure.

Bill

[JonathanBenoit]

Your goal, and that of many others as you say, is obviously the absolute lowest possible noise (highest SNR). You can achieve this if you can shoot 50, based on Erik's examples. My question went towards why you picked 100 as acceptable and 200 not.

The noise level is visible at 200 on the H3DII. I dont have any reason to shoot that fast, at least yet. Phocus does a great job removing noise, but since I dont need to shoot at 200, why even bother if there are no benefits to my work?

[Jack Flesher]

As for your comment that ISO 50 or 100 being "best" for a landscape shooter, can you please explain why you feel it would be any better than 200 or 400 if all these ISO settings generated identical noise and identical DR; IOW if all generated identical final images from a purely technical standpoint?  

Any given camera manufacturer's 'base' ISO is usually the 'optimal' ISO as well, meaning best signal to noise ratio.  So it may be the case for Pentax that ISO 200 generates the cleanest, highest DR file and therefore why that is it's base ISO. (We won't know until we actually get our hands on a raw file.)  Some manufacturers expand ISO to generate lower ISO's than the base setting, but this usually trims DR to some measurable degree so is not advised for critical exposures, but it usually does not add to the base noise so may be desirable in some situations -- at least it's there as an option if wanted.

[JonathanBenoit]

As for your comment that ISO 50 or 100 being "best" for a landscape shooter, can you please explain why you feel it would be any better than 200 or 400 if all these ISO settings generated identical noise and identical DR; IOW if all generated identical final images from a purely technical standpoint?  

Any given camera manufacturer's 'base' ISO is usually the 'optimal' ISO as well, meaning best signal to noise ratio.  So it may be the case for Pentax that ISO 200 generates the cleanest, highest DR file and therefore why that is it's base ISO. (We won't know until we actually get our hands on a raw file.)  Some manufacturers expand ISO to generate lower ISO's than the base setting, but this usually trims DR to some measurable degree so is not advised for critical exposures, but it usually does not add to the base noise so may be desirable in some situations -- at least it's there as an option if wanted.

If they generated the same image technically, it would be great to use ISO 400. ISO 100 wouldn't be better. Having the same image technically would just give more flexibility.
I'd like to know what the factors are behind a base ISO of 200. I don't know the technical details of it. Micro lenses cant be the only factor. If their ISO 200 is better than 100, would this be the first time this has happened?

[Jack Flesher]

I'd like to know what the factors are behind a base ISO of 200. I don't know the technical details of it. Micro lenses cant be the only factor. If their ISO 200 is better than 100, would this be the first time this has happened?

Well first off, I think the root of the issue under discussion it this -- you have a pre-conception that ISO 200 on the Pentax is going to be inherently worse than ISO 50 or 100 on some other manufacturers digital back as respects noise and image quality.  It may well be the case, but we simply won't know for certain until the camera is out and we can look at some raw files.  

To answer your question of is this the first time.  For ISO 200, possibly yes, but the Leica S2's native or base ISO is already at 160.  As respects micro-lenses, they do increase the quantity of photons entering the site well, so it is likely to be a significant part of the higher base ISO. (Usually accounts for about a full stop.)  As for where the rest is coming from, it could be a couple of things, but the most likely is probably newer sensor efficiency. Or possibly one of the newer style Bayer matrices with interspersed panchromatic or 'clear' cells --- basically because there is no color filter, there is no filter factor for some percentage of the sensor sites and hence the same base sensor ends up with a higher net sensitivity.  So if those two technologies are combined on one sensor, it is not much of a stretch to envision netting a base ISO of 200 or even higher...

[BJL]

I'd like to know what the factors are behind a base ISO of 200. I don't know the technical details of it. Micro lenses cant be the only factor.

It seems that with Kodak's medium format sensors, microlenses probably are the only factor, or close enough.

Compare the specs for the two new Kodak sensors, the 50MP KAF-50100 (without microlenses) and the 40 MP KAF-40000 (with microlenses), both using the same basic TrueSense 6.0 micron pixel design.
The former has quantum efficiencies in the three primary colors of 22%, 22%, 16%
The latter has quantum efficiencies in the three primary colors of 42%, 44%, 38%
All other per pixel specs are very similar.

That almost exact doubling of QE fits perfectly with a doubling of base ISO, and with the KAF-40000 offering almost exactly the same S/N performance at ISO 200 as the KAF-50000 does at ISO 100.

[VinceGalloro]

I wanted to thank everyone who has posted in this discussion -- some of which, quite frankly, is over my head. Reading the discussion has helped me understand the choices made with products like the Pentax 645D, even if there is still some disagreement among those who posted here.