Noct 58/0.95 - technical

[Roland Vink]

Interesting question. Film has reciprocity failure in low light with long shutter speeds - the sensitivity declines so even longer shutter speed is required to obtain a proper exposures, and color shifts can also occur.

I haven't heard of this with digital sensors. Maybe in theory in very bright light there is a limit to how fast the sensor can count photons? If so, it could occur even with slower lenses (or smaller apertures) if light was very bright (strong flash, direct sunlight) and in such cases there is more chance the sensor would be overloaded.

Could the microlenses limit the amount of light reaching the sensor when a very fast lens is used? Sensors work best when light is more or less collimated and strikes perpendicular. For a pixel near the centre of a sensor behind an f/1.0 lens, peripheral light will strike the sensor at about 63° so the efficiency of the sensor could be less. This effect could become even more pronounced towards the edges of the sensor as the light bundles become less perpendicular. If so, I would also expect to see color shifts towards the sides at very wide apertures, such as occurs when very wide non-retrofocus lenses are used. Has anyone observed this? Note that the last element of the Z Noct is concave (negative) which tends to collimate light, so it appears the designers at least attempted to minimise this effect.

[Jack Dahlgren]

My question was directed to anyone who has the hands-on experience with the Neo-Noct, but I would appreciate if you would perform a test to make sure of this aspect of the optical wonder.

Theoretically and practically, the relationship between the aperture setting and the shutter speed for the same exposure value is constant.  For example, 1/2000sec.@f1.4=1/1000@2.0=1/500@2.8, etc.  That is obvious.  But, because the physical opening (the speed) of the microlenses image sensor is said to be f1.2, the same relationship won't apply when the attached lens is faster than f1.2.

You may want to refer to this review of a budget 50mm/f1.1 lens by Christopher Frost:

https://www.youtube.com/watch?v=TRt2mNFxdDI

That makes no sense.
The sensor only knows how many photons fall on it.
The claim that it is saturated at f/1.2 and can not accept any more light is crazy.
It may be true with a certain light level, but the image would be completely overexposed.
The same could be true with any aperture given bright enough light.

[Eddie Draaisma]

Interesting question. Film has reciprocity failure in low light with long shutter speeds - the sensitivity declines so even longer shutter speed is required to obtain a proper exposures, and color shifts can also occur.

I haven't heard of this with digital sensors. Maybe in theory in very bright light there is a limit to how fast the sensor can count photons? If so, it could occur even with slower lenses (or smaller apertures) if light was very bright (strong flash, direct sunlight) and in such cases there is more chance the sensor would be overloaded.

Could the microlenses limit the amount of light reaching the sensor when a very fast lens is used? Sensors work best when light is more or less collimated and strikes perpendicular. For a pixel near the centre of a sensor behind an f/1.0 lens, peripheral light will strike the sensor at about 63° so the efficiency of the sensor could be less. This effect could become even more pronounced towards the edges of the sensor as the light bundles become less perpendicular. If so, I would also expect to see color shifts towards the sides at very wide apertures, such as occurs when very wide non-retrofocus lenses are used. Has anyone observed this? Note that the last element of the Z Noct is concave (negative) which tends to collimate light, so it appears the designers at least attempted to minimise this effect.

Maybe "Leica M" rings a bell...

[JohnMM]

Interesting discussion here ...

https://www.dpreview.com/forums/post/61063848

and in the references cited therein.

[Roland Vink]

The question at dpreview is stated as: "I understand that the sensors micro lenses have an effective working aperture of about F2 or smaller.  Does this imply that the sensor will not fully see/record all of the light from super speed lenses such as the Leica F1 noctilux ..."

My sense is that the aperture of the microlenses is not greatly relevant. Their job is simply to collect the light focused on them by the lens, and pass it down to the sensor. If the lens in front has a very fast aperture, the intensity of light focused on the microlenses may be greater, but they should still do their job the same as with a slower lens. Put it this way: imagine two sensors where the physical depth of the pixel is the same, the microlenses cover the same percentage of the surface but one has twice as many pixels as the other. The microlens on the high-res sensor is one stop slower than the microlens on the low-res sensor, yet both sensors have the same overall light-gathering ability regardless of the lens put in front.

As discussed earlier, the angle of incidence may be greater with very fast lenses which may reduce the efficiency of the microlenses, but that is not closely related to the aperture of the microlenses themselves.

[Akira]

That makes no sense.
The sensor only knows how many photons fall on it.
The claim that it is saturated at f/1.2 and can not accept any more light is crazy.
It may be true with a certain light level, but the image would be completely overexposed.
The same could be true with any aperture given bright enough light.

I'd still interested to make sure if the f0.95 Z Neo-Noct allows 1+ stop faster shutter speed than an f1.4 lens, say the f1.4 F Neo-Noct.  That should give the answer to the matter of the speed of on-sensor microlenses at the same time.

Some say that the aperture index of Kamlan 50/1.1 lens is inaccurate, which should affect the accuracy of the measurement of reciprocity.  The aperture index of Neo-Noct should be way more accurate.

If the f0.95 speed doesn't allow the expected faster shutter speed, the wide opening won't be too beneficial for the astro-photography, for example.

[Erik Lund]

I'd still interested to make sure if the f0.95 Z Neo-Noct allows 1+ stop faster shutter speed than an f1.4 lens, say the f1.4 F Neo-Noct.  That should give the answer to the matter of the speed of on-sensor microlenses at the same time.

Some say that the aperture index of Kamlan 50/1.1 lens is inaccurate, which should affect the accuracy of the measurement of reciprocity.  The aperture index of Neo-Noct should be way more accurate.

If the f0.95 speed doesn't allow the expected faster shutter speed, the wide opening won't be too beneficial for the astro-photography, for example.

Difference between f/0.95 and f/1.4 is 1 1/6 of a stop - between f/1 and f/1.4 is 1 stop.

That the micro-lenses on the sensor should have any impact on the difference between individual lenses is IMHO not valid in real life shooting - This I have never seen on Leica M shooting f/0.95, f/1, f/1.1 and f/1.2 together with f/1.4 - No uneven exposure compensation needed at all!

What has substantial impact is variance in coatings, glass quality, amount of glass elements, quality of grinding/casting of lens elements and optical formula, all have impact on vignetting and the final amount of light hitting the sensor or film.
T-stop comes up here as the ultimate solution for the lens itself. The re-housing of Noct-Nikkor for 'Cinema lenses' is quite common there it is labeled 58mm T1.3
That a camera body should be labeled Nikon Z7 T1.2 is not making any sense to me at all IMHO.
Also I have never ever seen these fast lenses not following the aperture values and also never seen any substantial difference between ISO values in Leica M or Nikon digital bodies   

[Birna Rørslett]

The relative exposure values for my Repro-Nikkor 85mm f/1 is f/1: 0 EV, f/1.4: +1 EV, f/2: +2 EV and so forth on each and every Nikon camera I have used it with. No additional adjustments are required.

[Akira]

Difference between f/0.95 and f/1.4 is 1 1/6 of a stop - between f/1 and f/1.4 is 1 stop.

That the micro-lenses on the sensor should have any impact on the difference between individual lenses is IMHO not valid in real life shooting - This I have never seen on Leica M shooting f/0.95, f/1, f/1.1 and f/1.2 together with f/1.4 - No uneven exposure compensation needed at all!

What has substantial impact is variance in coatings, glass quality, amount of glass elements, quality of grinding/casting of lens elements and optical formula, all have impact on vignetting and the final amount of light hitting the sensor or film.
T-stop comes up here as the ultimate solution for the lens itself. The re-housing of Noct-Nikkor for 'Cinema lenses' is quite common there it is labeled 58mm T1.3
That a camera body should be labeled Nikon Z7 T1.2 is not making any sense to me at all IMHO.
Also I have never ever seen these fast lenses not following the aperture values and also never seen any substantial difference between ISO values in Leica M or Nikon digital bodies 

Thank you, Erik, for the confirmation from your real-life experience!  So, the issue was an urban myth!

The relative exposure values for my Repro-Nikkor 85mm f/1 is f/1: 0 EV, f/1.4: +1 EV, f/2: +2 EV and so forth on each and every Nikon camera I have used it with. No additional adjustments are required.

When I initially posted my question, I thought about the Repro Nikkor.  However, the "f1.0" value of the 85mm Repro is an imaginary one when it would be focused at infinity.  So, I intentionally ignored it.  The same should go with Heligons: one can use it only at very close distance and the effective aperture should not be that astonishingly fast.  Also, Heligons don't have aperture blades, which makes it impossible to compare the shutter speeds at various aperture settings.

[Birna Rørslett]

No more imaginary than any lens

[Roland Vink]

I'd still interested to make sure if the f0.95 Z Neo-Noct allows 1+ stop faster shutter speed than an f1.4 lens, say the f1.4 F Neo-Noct.  That should give the answer to the matter of the speed of on-sensor microlenses at the same time.

I noticed that many lenses seem to under-expose when shot wide open (I especially noticed this when shooting slide films). Mechanical vignetting is highest wide open, so even if the centre of the image is exposed correctly, the outer regions could be under-exposed.

Closing down one stop will reduce the aperture in the centre. However, in lenses with high mechanical vignetting (especially zooms and fast lenses) the corners are hardly even "touched" by closing the aperture one stop, they were effectively stopped down already. So the average exposure is less than one stop different from wide open. This would account for why the shutter speed is often less than one stop different between wide open and closed down one stop, we don't need to blame micro-lenses for that

[pluton]

A sensor is a detector that will give a brighter image output if brighter light is presented to it.
Whether I open up the lens aperture from f/1.4 to f/0.95 OR keep the aperture at f/1.4 and dial up the brightness of the light by 1 1/6 stops shouldn't matter. 
The light intensity presented to the sensor has increased by 1 1/6 stops in either case.

[Toby]

My question was directed to anyone who has the hands-on experience with the Neo-Noct, but I would appreciate if you would perform a test to make sure of this aspect of the optical wonder.

Theoretically and practically, the relationship between the aperture setting and the shutter speed for the same exposure value is constant.  For example, 1/2000sec.@f1.4=1/1000@2.0=1/500@2.8, etc.  That is obvious.  But, because the physical opening (the speed) of the microlenses image sensor is said to be f1.2, the same relationship won't apply when the attached lens is faster than f1.2.

You may want to refer to this review of a budget 50mm/f1.1 lens by Christopher Frost:

https://www.youtube.com/watch?v=TRt2mNFxdDI

This is certainly not the case when I use my Canon f0.95. It is f0.95 as read by the camera.

[chambeshi]

Interview in Chinese with the designers of the 58mm f0.95S Nikkor. This is part II following previous about design of the Z-Mount
Google Translate renders distinctly garbled prose but it's readable

https://translate.googleusercontent.com/translate_c?depth=1&pto=aue&rurl=translate.google.com&sl=zh-CN&sp=nmt4&tl=en&u=https://info.xitek.com/allpage/attitudes/202005/06-345705.html&usg=ALkJrhhw_-0hQpbZZZ05rrIA-X5d3pW0aw

EDIT: this link works much better
https://info.xitek.com/attitudes/202005/06-345702.html#artPos

some western photo media sites have also picked up this interview - https://www.digitalcameraworld.com/news/the-monster-nikon-z-58mm-f095-s-noct-could-have-been-even-bigger

[Erik Lund]

Very interesting insight, A big ring motor SWM, wow yes that would have made it huge