Sensor sensitivity

[VinceGalloro]

Hello. Reading here about the Pentax 645D, I was struck by the minimum ISO of the camera being 200. As the lengthy, informative interview that was published on Luminous Landscape made clear, Pentax aimed this camera squarely at landscape shooters. Why isn't there a lower native ISO option than 200? It can get down to ISO 100 equivalent, but still, why not ISO 50? The D3X has native ISO 100, and ISO 50 equivalent, but again, I wonder why it doesn't go even lower, to a native ISO 50 and then extended ISOs below that.

I can guess why a camera that aims to give much higher usable ISOs (and not designed with landscape shooters in mind, at least not primarily), like the D3/D3s, would bottom out at ISO 200 -- I can imagine that a sensor that is usable at those sky-high ISOs just can't be made any less sensitive than ISO 200. But can sensors for cameras that are not seeking to push the limits on the high end instead push the limits on the low end?

I'm wondering if there is a technical reason that sensors can't be made to native ISO 50, or maybe even ISO 25. Or maybe that, while they can be made to that spec, doing so makes too much of a tradeoff w/r/t other features that photographers want in a sensor.

[tbosley]

I'm wondering if there is a technical reason that sensors can't be made to native ISO 50, or maybe even ISO 25. Or maybe that, while they can be made to that spec, doing so makes too much of a tradeoff w/r/t other features that photographers want in a sensor.

I've owned several older Leaf backs that start off at 25 and 50...  

[ErikKaffehr]

Hi,

It's called quantum efficiency. A sensor captures something like a 80% of the photons falling on it. There are two ways of reducing ISO:

Leaving out the micro, lenses, gives about one stop.
Putting an ND filter in the front of the sensor

Essentially there is zero benefit of keeping sensor sensivity down. You can actually achieve the same thing by adding an ND filter in front of the lens. In some light conditions you may be able to overexpose, something Leica calls "pulling" on the S2.

Most MFDBs don't have microlenses.

This is really a question of technology moving forward. Digital sensor technology is quite sensitive to light, so base ISO tends to be pretty high. Why would a camera maker "cripple" the high ISO performance of a camera when you can achieve the exactly same effect with a cheap peace of ND glass in front of the lens?

Best regards
Erik

Hello. Reading here about the Pentax 645D, I was struck by the minimum ISO of the camera being 200. As the lengthy, informative interview that was published on Luminous Landscape made clear, Pentax aimed this camera squarely at landscape shooters. Why isn't there a lower native ISO option than 200? It can get down to ISO 100 equivalent, but still, why not ISO 50? The D3X has native ISO 100, and ISO 50 equivalent, but again, I wonder why it doesn't go even lower, to a native ISO 50 and then extended ISOs below that.

I can guess why a camera that aims to give much higher usable ISOs (and not designed with landscape shooters in mind, at least not primarily), like the D3/D3s, would bottom out at ISO 200 -- I can imagine that a sensor that is usable at those sky-high ISOs just can't be made any less sensitive than ISO 200. But can sensors for cameras that are not seeking to push the limits on the high end instead push the limits on the low end?

I'm wondering if there is a technical reason that sensors can't be made to native ISO 50, or maybe even ISO 25. Or maybe that, while they can be made to that spec, doing so makes too much of a tradeoff w/r/t other features that photographers want in a sensor.

[bjanes]

Hi,

It's called quantum efficiency. A sensor captures something like a 80% of the photons falling on it. There are two ways of reducing ISO:

Leaving out the micro, lenses, gives about one stop.
Putting an ND filter in the front of the sensor

Essentially there is zero benefit of keeping sensor sensivity down. You can actually achieve the same thing by adding an ND filter in front of the lens. In some light conditions you may be able to overexpose, something Leica calls "pulling" on the S2.

I agree with Erik that quantum efficiency is a desirable attribute and the resulting increase in ISO is an asset and not a liability in most cases. Another way to lower the effective ISO of a sensor would be to increase the full well capacity so that more electrons could be collected. As Roger Clark points out, "As a pixel holds more electrons, the charge density increases. There are finite upper limits to the charge density of electron storage". Full well capacity is basically related to pixel area, and for a given quantum efficiency, the number of electrons collected is also related to pixel area. For example, the Nikon D3 sensor is generally well regarded, but as Roger's Figure 1 shows, its full well is no greater than the smaller pixel Canon 5D Mark 2, and there would be some room for improvement with the D3 full well.

[ErikKaffehr]

I absolutely agree with Bill, good point.

What I intended to say that relatively high sensitivity is an inherent property of the process and there is little technical merit to decrease ISO by artificial means so I don't really see it happening. Doubling the full well capacity would essentially decrease ISO, at least in terms of exposing to the right. On the other hand pixels tend to shrink and having more pixels has some merits so I don't think we are going to see a big raise in well capacity.

in MFDBs the premium sensors are essentially nearly full frame 645 and going to having like 6 my (micron) pitch, so I don't see them increase well capacity either. One of the reasons MFDBs have low ISOs is that they don't have microlenses, mostly. The Leica S2 has microlenses and a base sensivity of 160 ISO, but has an option to "pull" ISO to 80. Pulling ISO could be achieved by "exposing to the right" under certain conditions.

The simplest way to add low ISO capability would probably to have an optional replacement for the front assembly of the sensor having an ND-filter like it was done on the Mamiya ZD, which had an optional OLP (Optical Low Pass) filter. This would be very much doable, at least if enough people would be willing to pay for it.

Best regards
Erik

I agree with Erik that quantum efficiency is a desirable attribute and the resulting increase in ISO is an asset and not a liability in most cases. Another way to lower the effective ISO of a sensor would be to increase the full well capacity so that more electrons could be collected. As Roger Clark points out, "As a pixel holds more electrons, the charge density increases. There are finite upper limits to the charge density of electron storage". Full well capacity is basically related to pixel area, and for a given quantum efficiency, the number of electrons collected is also related to pixel area. For example, the Nikon D3 sensor is generally well regarded, but as Roger's Figure 1 shows, its full well is no greater than the smaller pixel Canon 5D Mark 2, and there would be some room for improvement with the D3 full well.

[Anders_HK]

Hello. Reading here about the Pentax 645D, I was struck by the minimum ISO of the camera being 200. As the lengthy, informative interview that was published on Luminous Landscape made clear, Pentax aimed this camera squarely at landscape shooters. Why isn't there a lower native ISO option than 200? It can get down to ISO 100 equivalent, but still, why not ISO 50? The D3X has native ISO 100, and ISO 50 equivalent, but again, I wonder why it doesn't go even lower, to a native ISO 50 and then extended ISOs below that.

I can guess why a camera that aims to give much higher usable ISOs (and not designed with landscape shooters in mind, at least not primarily), like the D3/D3s, would bottom out at ISO 200 -- I can imagine that a sensor that is usable at those sky-high ISOs just can't be made any less sensitive than ISO 200. But can sensors for cameras that are not seeking to push the limits on the high end instead push the limits on the low end?

I'm wondering if there is a technical reason that sensors can't be made to native ISO 50, or maybe even ISO 25. Or maybe that, while they can be made to that spec, doing so makes too much of a tradeoff w/r/t other features that photographers want in a sensor.

Vince,

I shoot landscapes and portraits as an advanced amateur and use nearly exclusively ISO 50 & 100. On rare occasions do I use ISO 200 or 400. That applies to film and digital alike. Simply, ISO 200 as base ISO would not be acceptable to me.

For digital I use Leaf Aptus 65 with base ISO 50. Simply it gives the optimum image quality. Higher ISO also with digital result in more contrast, and by pushing away from base ISO it results in noise. Now, for main steam amateur shooters it may perhaps not matter, but for someone that is very serious in photography it does.

All Leaf and Phase One digital backs have base ISO at 50 or less (25), only the new Aptus-II 8 & 10 models have higher, at ISO 80.

Perhaps Pentax tried to make the sensor cheaper in order to compete with Nikon & Canon? I do not know. However, there are quality refurbished Leaf and Phase-One digital backs available for good prices from Leaf and Phase One agents and under warranty. Those are also very tried and reliable products that are capable of exceptional image quality.

As some replies already indicate, in this forum it is popular to argue technology as in the lines of Nikon, Canon and the like. For optimum highest image quality it takes low ISO with a medium format digital back and this is one important reason to why people use them. Perhaps some will argue my post. I will ignore that. Simply it is tiring and waste of my time argue with folks who simply will not see. If you have any more questions, please feel free to drop me a private message instead.

Regards
Anders

[Bart_van_der_Wolf]

I'm wondering if there is a technical reason that sensors can't be made to native ISO 50, or maybe even ISO 25. Or maybe that, while they can be made to that spec, doing so makes too much of a tradeoff w/r/t other features that photographers want in a sensor.

Hi Vince,

IMHO it's important to understand that silicon based sensors have an inherent sensitivity to various wavelengths of light, and that sensitivity is not that much different between sources. There can however be several design compromises that will reduce that inherent sensitivity. Examples of such sensitivity reducing compromises are the signal connection architecture to off-load the signal to other places which reduce the effective light sensitive surface area, and/or color filters with certain transparency characteristics. Some designs are more efficient in avoiding the loss of quantum efficiency, e.g. back illumination to avoid obstruction from connection gates.

In addition, there is the matter of definition of ISO. If one adopts a Saturation (clipping) based definition, like in 'expose-to-the-right' (ETTR), then the capacity to collect the energy of a large numbers of photons will influence the amount of exposure needed to reach that saturation level. In other words, sensels with large capacity wells or capacitance, will have a lower equivalent ISO. That has nothing to do with sensitivity of silicon.

So ultimately, it's the combination of design characteristics and per sensel surface area that will determine the ISO. Those characteristics also define things like color accuracy, dynamic range, color uniformity, etc., i.e. other aspects of image quality (which can be one reason to want low ISO capability).

The question thus becomes, why does one want low ISO? Is it for quality reasons, or does one need a low efficiency sensor design for the purpose of allowing low ISO capability (e.g. in bright lighting conditions)? One may hope it's the quality aspect that prevails, since Neutral Density filters can come a long way in reducing the scene brightness (although unfortunately also the viewfinder).

One trend to keep a critical eye out for is increasing the transparency of the CFA filter structure, since that may negatively impact color accuracy, i.e. the capability to successfully recombine RGGB filtered sensels into RGB pixels. The quest for high ISO performance may lead to poorer color quality.

Cheers,
Bart

[bjanes]

I shoot landscapes and portraits as an advanced amateur and use nearly exclusively ISO 50 & 100. On rare occasions do I use ISO 200 or 400. That applies to film and digital alike. Simply, ISO 200 as base ISO would not be acceptable to me.

For digital I use Leaf Aptus 65 with base ISO 50. Simply it gives the optimum image quality. Higher ISO also with digital result in more contrast, and by pushing away from base ISO it results in noise. Now, for main steam amateur shooters it may perhaps not matter, but for someone that is very serious in photography it does.

There is no doubt that shooting at base ISO will give the best quality, but your implication that a higher base ISO per se degrades image quality is unfounded. Could you please elaborate and post references?

[ErikKaffehr]

Hi,

Just a few points. It seems that the base ISO for the Kodak CCD used in the Leica S2 and the Pentxa is about 160. At least the Leica uses micro lenses. It's reasonable that a microlens free design, like most MFDBs would have something like a sto less speed. There may also be differences in the CGA.

Another issue, it seems, with older MFDBs that the higher ISO settings were not achieved using variable pre amplifiers but fixed preamps and tagging the file with higher ISO also called "fake ISOs", these factors could at least in part explain why MFDBs work best at low ISOs. The few MFDBs that sport microlenses also tend to have about twice the ISO compared with the microlens free designs.

Best regards
Erik

There is no doubt that shooting at base ISO will give the best quality, but your implication that a higher base ISO per se degrades image quality is unfounded. Could you please elaborate and post references?

[JonathanBenoit]

There is no doubt that shooting at base ISO will give the best quality, but your implication that a higher base ISO per se degrades image quality is unfounded. Could you please elaborate and post references?

Obviously a comparison test will be done, but by definition 200 iso will have more noise than 100 iso or lower. Unless somehow they are using a sensor that has the sensitivity of 200 iso with the same amount or lack of noise as 100 iso.
I think it makes more sense that they are using cheaper sensors, but I have no idea for sure.

[bjanes]

IMHO it's important to understand that silicon based sensors have an inherent sensitivity to various wavelengths of light, and that sensitivity is not that much different between sources. There can however be several design compromises that will reduce that inherent sensitivity. Examples of such sensitivity reducing compromises are the signal connection architecture to off-load the signal to other places which reduce the effective light sensitive surface area, and/or color filters with certain transparency characteristics. Some designs are more efficient in avoiding the loss of quantum efficiency, e.g. back illumination to avoid obstruction from connection gates.

One trend to keep a critical eye out for is increasing the transparency of the CFA filter structure, since that may negatively impact color accuracy, i.e. the capability to successfully recombine RGGB filtered sensels into RGB pixels. The quest for high ISO performance may lead to poorer color quality.

Bart,

Thanks for adding your insights to the discussion. One example of compromises in CFA filter characteristics is discussed on the DXO comparison of the Canon EOS 500D and the Nikon D5000. The former camera tries to push up the global sensitivity of the sensor by applying a less selective filter, requiring higher color matrix coefficients which push up color noise. The Canon has higher resolution with smaller pixels, likely leading to this compromise in an effort to maintain sensitivity.

Bill

[Anders_HK]

Hello. Reading here about the Pentax 645D, I was struck by the minimum ISO of the camera being 200. As the lengthy, informative interview that was published on Luminous Landscape made clear, Pentax aimed this camera squarely at landscape shooters. Why isn't there a lower native ISO option than 200? It can get down to ISO 100 equivalent, but still, why not ISO 50? The D3X has native ISO 100, and ISO 50 equivalent, but again, I wonder why it doesn't go even lower, to a native ISO 50 and then extended ISOs below that.

I can guess why a camera that aims to give much higher usable ISOs (and not designed with landscape shooters in mind, at least not primarily), like the D3/D3s, would bottom out at ISO 200 -- I can imagine that a sensor that is usable at those sky-high ISOs just can't be made any less sensitive than ISO 200. But can sensors for cameras that are not seeking to push the limits on the high end instead push the limits on the low end?

I'm wondering if there is a technical reason that sensors can't be made to native ISO 50, or maybe even ISO 25. Or maybe that, while they can be made to that spec, doing so makes too much of a tradeoff w/r/t other features that photographers want in a sensor.

http://www.leaf-photography.com/news_0410_js.asp

[bjanes]

Obviously a comparison test will be done, but by definition 200 iso will have more noise than 100 iso or lower. Unless somehow they are using a sensor that has the sensitivity of 200 iso with the same amount or lack of noise as 100 iso.
I think it makes more sense that they are using cheaper sensors, but I have no idea for sure.

For most normal photography, noise is determined by shot noise and read noise added in quadrature. Shot noise is determined by the number of photons that are collected and has little to do with the ISO. Except for the deep shadows, shot noise predominates with all decent digital sensors. Read noise at low ISO for many DSLRs is actually larger than at higher ISOs due to limitations in the analog to digital converter at the high readout rates required for high frame rates. This limits the dynamic range at low ISO. For an example, look at the DR for the Nikon D3 on the DXO site. The DR vs ISO curve flattens below ISO 800 due to ADC and electronics limitations. I wouldn't say that Nikon is using a cheap ADC, but rather they accept higher read noise for a fast frame rate.

Comparison tests have been done by DXO. For example, the SNR (screen) of the D3x is 39 dB at an effective base ISO of 78 and that of the Phase One P65+ is 39.2 dB at an effective base ISO of 44. If one compares the print SNRs, the P65+ has an advantage of 48 dB as compared to the D3x SNR of 43.9. This is due to the fact that the P65+ has a larger total sensor area and collects more photons and not due to the difference in ISOs.

[BJL]

by definition 200 iso will have more noise than 100 iso or lower. Unless somehow they are using a sensor that has the sensitivity of 200 iso with the same amount or lack of noise as 100 iso.

Not "by definition" at all when comparing different sensors or films. The latter is exactly the point: the base ISO of sensors increases in newer models mostly because of improved sensitivity, due to improvements in microlenses (gapless is the latest trend) and lower dark noise (improved amplifier designs), so that S/N ratio at base ISO is the same or better.

I would be wiling to bet that the IQ at base ISO is consistently improving in rational IQ comparisons (like viewing images at equal size) along with the increases in base ISO. In particular when the base ISO increase comes from adding microlenses in sensors like that of the Leica S2 and Pentax 645D.

[JonathanBenoit]

Not "by definition" at all when comparing different sensors or films. The latter is exactly the point: the base ISO of sensors increases in newer models mostly because of improved sensitivity, due to improvements in microlenses (gapless is the latest trend) and lower dark noise (improved amplifier designs), so that S/N ratio at base ISO is the same or better.

I would be wiling to bet that the IQ at base ISO is consistently improving in rational IQ comparisons (like viewing images at equal size) along with the increases in base ISO. In particular when the base ISO increase comes from adding microlenses in sensors like that of the Leica S2 and Pentax 645D.

Adding microlenses or having more noise than with base iso 100. either way, the IQ suffers

[Bart_van_der_Wolf]

Adding microlenses or having more noise than with base iso 100. either way, the IQ suffers

Hi Jonathan,

IQ doesn't necessarily suffer from the use of micro-lenses, in fact it can improve the quantum efficiency and reduce light fall-off, which in turn boosts base ISO requiring lower amplification and thus noise. Perhaps you are referring to something else?

Cheers,
Bart

[ErikKaffehr]

Hi,

It is not my opinion that micro lenses affect IQ negatively. The effect they have is mainly that photons that would normally be lost are detected. On the other hand, micro lenses will increase the fill factor and thus reduce aliasing artifacts. The very same effects may reduce perceived resolution.

My understanding is that micro lenses play well with telecentric lens designs like DSLR/MF wide angles, but less so with more symmetric wide angle designs or shifted lenses.

Best regards
Erik

Adding microlenses or having more noise than with base iso 100. either way, the IQ suffers

[BJL]

Adding microlenses or having more noise than with base iso 100. either way, the IQ suffers

Why, apart from special cases like tilt-shift lenses? Adding micro-lenses mainly means that the sensor gathers light faster from the same scene, so each photo-site gets about as much light at the new base ISO of 200 with micro-lenses as it would at the old base ISO of 100 without micro-lenses. Same signal, same S/N ratio, but with options like less motion blur and/or more DOF depending on whether one increases shutter speed or f-stop.

And most recent DSLR base ISO increases are in sensor product lines that already had micro-lenses, through improving the micro-lenses to gap-less designs and such. Where is the worse noise and IQ there?

[JonathanBenoit]

Why, apart from special cases like tilt-shift lenses? Adding micro-lenses mainly means that the sensor gathers light faster from the same scene, so each photo-site gets about as much light at the new base ISO of 200 with micro-lenses as it would at the old base ISO of 100 without micro-lenses. Same signal, same S/N ratio, but with options like less motion blur and/or more DOF depending on whether one increases shutter speed or f-stop.

And most recent DSLR base ISO increases are in sensor product lines that already had micro-lenses, through improving the micro-lenses to gap-less designs and such. Where is the worse noise and IQ there?

Microlenses soften the image. I'm more interested in tack sharp images than better noise reduction at iso 200

[Bart_van_der_Wolf]

Microlenses soften the image. I'm more interested in tack sharp images than better noise reduction at iso 200

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