Lens transmission tests

[Royal Winchester]

I am, admittedly, a total gear head.
But, I did some tests today regarding the transmission of a few fast lenses I have.
I mostly shoot with a D500, but I have a Sony A7rii which I use for some vintage lenses I inherited and for full frame wide angle stuff.  I recently acquired the Mitakon 50mm f/0.95 in a Sony FE mount, and it has been criticized for really being a T 1.4, not really 0.95.  So, I thought I'd do an ad-hoc test.  Perhaps others will find it interesting? 

On the Sony I can shoot my Nikkor 50mm f/1.2, Voigtlanger 58mm f/1.4, Rokinon 24mm f/1.4, Mitakon f/0.95, and an old Sears 55mm f/1.4.
On the Nikon I can shoot the first three only.

I started with the 50 f/1.2 on the Sony and exposed the third grey square on an X-rite color card with the lens at 1.4.  Iso 100, shutter 1/50.
I opened it up to 1.2 - no change.  At 2.0 it was down one stop as expected, and at 2.8 it was down 2 etc.
I shot the 58mm at 1.4 - perfectly exposed
I shot the Rokinon at 1.4 - also perfectly exposed as long as I didn't block the light with my body or camera
I shot the Mitakon at 1.4 - also perfectly exposed.  Down one stop exactly at 2.0.  Only up 1/3 of a stop at 0.95 though, instead of the >1 stop you'd hope to gain.  Apparently the transmission is only affected at the widest settings?
I shot the Sears at 1.4 - a full 2/3 of a stop underexposed.

Now, the D500 results.

I put the 50mm f1.2 on and at f1.4 and iso 100, 1/50s was definitely overexposed.  Had to move to 1/80 shutter to get perfect exposure on the square.  So... the D500 is about 2/3 stop more sensitive than the Sony at iso 100 I guess?
Opening to f1.2 gained about 1/3 of a stop.  Looking at the aperture blades, they do barely move from 1.4 to 1.2.  Perhaps vignetting keeps that extra light from reaching the Sony's sensor?
The Voigtlander 58mm f1.4 was also perfectly exposed at f1.4 1/80
the Rokinon 24mm f1.4 was almost 2/3 of a stop underexposed and wasn't right on until back at 1/50 shutter.  That, I have no explanation for...

Anyway - I thought it was interesting!

-Royal

[Erik Lund]

Did you shoot a completely uniform target for all of the lenses? are you aware that most lenses drop in aperture as you focus closer, how did you process and evaluate these 'exposures' RAW or?

Also some of these lenses will have a lot of vignetting especially wide open.

[Bjørn Rørslett]

Shooting at infinity focus is a prerequisite for this kind of testing. Finding a large enough and evenly lit subject might be a problem, though. However it is not required the lens is in focus at the plane of the subject. Thus any diffuse screen that can be illuminated evenly would do if it intercepts completely the field of view of the lens (with additional margin for error).

[Dr Klaus Schmitt]

I do it this way...



Nikkor f4/18mm vs Tokina 3.5/17mm

[Royal Winchester]

I was wondering if my methodology was flawed, and it clearly is.
That's why I love this forum!

Erik:  I shot a color chart, and my exposure measures were camera back using the tightest spot meter I could - in camera.  It wasn't actually based on a photo since I just tried to see what it considered neutral exposure for that third grey square.
The focus range was close, but not exactly the same for all the lenses.  I sat in the same spot and held the camera to my eye and did a spot meter on the target.

I was wondering about vignetting causing some of the symptoms I saw...

Bjorn:  No indeed, I was no where near infinity focus!  Retesting sounds necessary if I want any real information here!

Dr Klaus:  What exactly do you do to generate those curves?

Thanks guys!

[Dr Klaus Schmitt]

I was wondering if my methodology was flawed, and it clearly is.
That's why I love this forum!

Dr Klaus:  What exactly do you do to generate those curves?

Thanks guys!

A UV-VIS-NIR spectrophotometer system optimised to be able to measure optical lenses for their transmission (mainly done to find lenses being able to transmit UV light and to scientifically prove statements about such abilities)

[Bjørn Rørslett]

Thanks Klaus for all the stuff you have published over the years. The curve for the 18/4 AI is precisely what I would expect from practical UV work with that lens.

RW: f-numbers are *always* based upon geometry, not transmission. Thus stating 50mm f/0.95 is actually 50 T 1.4 is not per se conflicting at all. Most lenses have the inevitable light fall-off towards the corners that at the widest apertures can be much more than 1 stop. On-axis, f-numbers and T-numbers tend to be close though, with T values always numerically larger (an f/1.2 lens can, say, be T1.3). Stopping down any lens will make its illumination more even across the projected image circle.

Geometry determines depth of field, bokeh etc., transmission determines exposure. Transmission is wavelength dependent, as seen by the graphs published by Dr. Schmitt. Thus quality of the incident light also influences the final exposure. It is important to keep these aspects of lens performance apart.

[Royal Winchester]

That is fantastic, Klaus.  So nice when science and facts are available.  The www seems to have done the opposite of that in so many ways.
Bjorn, that makes total sense to me, and is what I expected.  It more a question of degree to me.  Surely no system is without loss, but what is reasonable to expect.
If my understanding is correct, the DOF aspect of an f-number is pure geometry.  The transmission, or "see in the dark" fun of a lens is about transmission, and aperture is one factor in the light gathering potential.
Additionally, I seem to find, for the most part, that I can rely on stopping down one stop and halving my shutter speed to basically keep my exposure settings equal.  Or that I can typically rely on the f-stop measurement of a light meter to get a good exposure for ambient or flash.  The question marks popped out of my head when this particular lens was reviewed (not saying by perfectly reliable sources) as being "misleading" because the transmission was so much lower than the speed of the lens implied. 
Perhaps those criticisms are not all that valid. 
I don't have much experience with super fast lenses, so I am not aware if lenses faster than 1.4 tend to have this falloff in transmission from the expected formula.

Thank you guys for your comments!

[Bjørn Rørslett]

Super fast lenses are like any other lens and the aperture as such does not directly influence the relative transmission  However, very fast lenses usually are highly non-symmetric and hence are very likely to show changes in the *effective* aperture as focus goes away from infinity. With a flawed experimental set up one can but expect inaccurate results.

[pluton]

 The transmission *should* always be less than what is suggested by the f/number because of both absorption into the optical glass, and loss through reflection from the optical surfaces. 
 The manufacturers/marketers have been known in the past to occasionally deliberately misstate the f or T values for a given model of lens.  Usually, they designate the optic as being 'faster' than it really is for prestige/marketing reasons.  Often, a lens sold as f/1.2 might have an actual geometric f-stop of f/1.28 or f/1.31, etc.

[Bjørn Rørslett]

They do have an allowed margin of +-5% on specifications, which apparently always is on the - side for focal length and + side for aperture value.

However, don't forget the logarithmic nature of aperture. Usually the T value turns out to be closer than 1/3 EV and often 1/6EV or less from the f-number. Cinematographers need very accurate values because of the requirement to avoid perceptible flickering of the scenes. A still photographer could use either because most scenes - seen isolated from all others - have an exposure margin much wider than this narrow difference f vs T.