Photons Missing In Action

Started Nov 24, 2009 | Discussions
FKS
FKS Senior Member • Posts: 2,991
Re: Thanks for sharing.

ditto, thanks for taking the time marianne.

the loss of light with the 70-200VR + 1.4TC combination is disappointing. good thing the current set of nikon bodies has good high ISO capabilities.

 FKS's gear list:FKS's gear list
Nikon D3 Nikon 1 V2 Nikon D5200 Fujifilm X-E2 Fujifilm X-T1 +30 more
noirdesir Forum Pro • Posts: 13,586
Re: Part 2: Aperture Control Accuracy

I presume all your measurements are relative to wide-open. If you see a consistent under- or overexposure, could it be that the f-stop wide open isn't exactly the nominal one and that thus the stopped down values are the correct ones and the aperture wide open is the "inaccurate" one?
(And would chipped lenses know that and be able to tell it to the camera?)

Another thought, could it be possible that the ratio between the t-stop and the f-stop varies with aperture, eg, less reflection losses for the light rays passing through the center than for those passing through the outer regions?

ElanR New Member • Posts: 22
Re: Part 1: Lens T-stops

Another thanks for this unique post and all the effort that went into it.

That a lens' light transmission is somewhat different from its aperture rating should not come as a surprise, as that rating is merely a physical ratio that we expect to deliver a certain amount of light, not an actual measure of that light. It also should not be surprising that the tele-zoom lenses with the largest number of elements deliver significantly less light than similarly rated shorter primes, as the extra distance and all the glass the light has to pass through do take their toll. With that said, I am surprised at some of your findings.

More information about the testing equipment used and your methodically would be much appreciated. I'm looking forward to your color-cast test.

OP Marianne Oelund Veteran Member • Posts: 7,777
They're Relative to Wide Open

noirdesir wrote:

I presume all your measurements are relative to wide-open. If you see a consistent under- or overexposure, could it be that the f-stop wide open isn't exactly the nominal one and that thus the stopped down values are the correct ones and the aperture wide open is the "inaccurate" one?

In other words, can the aperture error compensate for the efficiency loss when the lens is stopped down? In some cases yes, but unfortunately that is purely academic, and of little help to the user, as the image exposure will still be thrown off. If a manufacturer were to intentionally do this to "correct" for light loss, they would be doing a disservice to their customers!

Another thought, could it be possible that the ratio between the t-stop and the f-stop varies with aperture, eg, less reflection losses for the light rays passing through the center than for those passing through the outer regions?

Possible, but I have no means of separating that effect from the diaphragm stop-down. In any case, the data do correctly infer the exposure errors which will result for images taken.

noirdesir Forum Pro • Posts: 13,586
Re: They're Relative to Wide Open

Marianne Oelund wrote:

noirdesir wrote:

I presume all your measurements are relative to wide-open. If you see a consistent under- or overexposure, could it be that the f-stop wide open isn't exactly the nominal one and that thus the stopped down values are the correct ones and the aperture wide open is the "inaccurate" one?

In other words, can the aperture error compensate for the efficiency loss when the lens is stopped down? In some cases yes, but unfortunately that is purely academic, and of little help to the user, as the image exposure will still be thrown off. If a manufacturer were to intentionally do this to "correct" for light loss, they would be doing a disservice to their customers!

Just to make sure we are talking about the same thing, I meant that if a f/2 lens actually was an f/2.1 lens and you tried to measure the accuracy of f/2.8 stop you would measure a delta of 0.8 stops between them or in other words conclude that the f/2.8 aperture is 1/5 stop too bright (always assuming naturally that the ratio between f and t-stop is the same for both apertures which is what you assumed already). Now, if the camera new that a lens actually had a f/2.1 aperture wide open it would adjust a nominal f/2 and 1/100 s not to f/2.8 and 1/50 s but to 1/56 s (eg, in aperture-priority mode).

OP Marianne Oelund Veteran Member • Posts: 7,777
Unworkable in Practice

noirdesir wrote:

Just to make sure we are talking about the same thing, I meant that if a f/2 lens actually was an f/2.1 lens and you tried to measure the accuracy of f/2.8 stop you would measure a delta of 0.8 stops between them or in other words conclude that the f/2.8 aperture is 1/5 stop too bright (always assuming naturally that the ratio between f and t-stop is the same for both apertures which is what you assumed already). Now, if the camera knew that a lens actually had a f/2.1 aperture wide open it would adjust a nominal f/2 and 1/100 s not to f/2.8 and 1/50 s but to 1/56 s (eg, in aperture-priority mode).

Using the lens CPU in such a scheme would require that the lens behavior be consistent. However, it is variable from camera to camera - and this is why I am urging users to test their own equipment.

This is all a matter of some very small mechanical misalignments. The movement of the camera's aperture control lever is only about 0.8mm per stop, or less than 0.3mm per 1/3 stop increment.

radddad Veteran Member • Posts: 3,123
Marianne .

First off , I want to thank you for your time and effort in sharing this informative post . Well done , and very entertaining reading .

I see that the 300/2.8vr is NOT in your prime list . This is by-far my favorite lens , and also IMO , the best lens in my line-up . I would be very curious how this lens pans out , especially in the brightness and color rendition area , as it seems to have a quite unique color rendition on photos . Any chance in the future ???
Best regards , and once again , thank you for your time !!!
--
All NIKON

member of WSSA (the Worldwide Squirrel Shooters Association) as member# 159

beru1 Regular Member • Posts: 191
Re: Photons Missing In Action

could you say a bit about your setup, what light source, how you deal with ambient light, exposure times used (if exposure times are not constant, how to know that sensitivity is constant to compare lenses?), how many measurements for each lens (e.g. because of noise variation)? Also, if just one camera is used, how to exclude that it has (shifting) sensitivity/calibration issues?

OP Marianne Oelund Veteran Member • Posts: 7,777
Part 3: Lens color cast
1

Several months ago, I began looking at lens color characteristics, and how they are affected by the type of lighting used. This turned out to be quite interesting, as there definitely are interactions between the lens transmission spectrum, and the source light spectrum. Because of these interactions, I delayed posting the lens color information until I had a chance to study them. I was concerned that there may be some narrow-band absorption contributing to the variations I saw, but fortunately it turns out that this is not the case; lenses have smooth transmission spectra. I will post a few examples later.

Overall, there is an apparent tendency for lenses to slightly exaggerate the color temperature of light sources. For example, warmer sources produce stronger red shift in the lens color cast. For these measurements, I am using the ratios of the three color channel signals, such as red/green ratio or red/blue ratio. The data is taken directly from 14-bit uncompressed NEF files, using a software utility which computes average channel signals from a central 200x200 pixel area of the image, so this is the sensor's view of lens color, with no influence from raw converters, white balance settings, etc.

To summarize the data, I am presenting two values for each lens:

  1. Red/blue skew, given as the percent increase in red/green ratio (or decrease in blue/green ratio) caused by the lens. This value describes how much the light from the lens has been "tilted" toward red and away from blue. For example, the value 2R indicates that the red/green ratio of light from the lens is 2% higher than for the source light. In the rare case where a net blue skew occurs, the letter B is used instead.

  2. Green/magenta skew. This is the percent change in the ratio G/(R+B) caused by the lens. Most lenses have some green shift, which simply means that green light is slightly less attenuated. This is written, for example, as 2G, which indicates a 2% increase in the G/(R+B) ratio. In those few cases where the ratio decreases, the letter M (for magenta) is used instead of G.

The following table summarizes the results. Note that some lenses have not been tested under all light sources, but there is still enough information for each, to see how they compare. For results that are less than 0.5% shift, the letter N (for neutral) is used. Along with the lens data, I give the basic red/green and blue/green ratios for each light source, as measured by the camera's sensor. The source "Daylight CF" is a compact fluorescent lamp, balanced for daylight; it has white balance similar to cloudy day. "Photofld" is a nominal 4800K, 250W photo floodlight, and the other sources should be clear from the names.

After scanning through the data, you will see that there are some notable lenses - some which have nearly neutral balance across many light sources, such as the 35mm f/2D, and others which are strongly shifted, such as the 24-70/2.8G. The 70-200 listed is the VR I version. I have only briefly tested the new VR II, and so far have found it to have about half the red skew of the VR I.

I should also point out how flash lighting is producing widely scattered results. This is obviously a tricky light source to use. Some lenses will behave well with flash, while others are at their worst, or nearly so, with it.

For more information about light sources, look for my new thread "How our cameras see color," which will be posted soon on the D3/D1 forum.

OP Marianne Oelund Veteran Member • Posts: 7,777
Part 3 Addendum: Spectrogram examples
1

As mentioned in my previous post, the lens color spectra are smooth and do not exhibit absorption bands. I've selected a few representative models from the extremes of the color cast range, plus a couple others which are likely to be of interest.

As noted before, the spectral skew is slightly dependent upon the light source, and we see a little more red skew when light sources with high long-wavelength content are used. For these spectrograms, the source was a small halogen lamp, and the spectra do show more attenuation toward the blue end of the spectrum, than we would see with a higher color-temperature source. All of the plots are level, to downward sloping towards blue.

Some of you will need to think backwards in viewing these plots, since I use a frequency-order axis which places the long wavelengths (red) at the left. I've grouped the lenses together by similar type. In each multiple plot, the vertical position is not an indicator of relative lens light transmission; it's just a convenient placement which I have chosen to separate the curves a little. The horizontal axis is labeled with wavelength in nm (nanometers). Due to very low signal levels, data at the spectrum extremes has much lower accuracy, and that is the reason for the somewhat erratic plot ends.

This first group consists of some short primes. With many light sources, the AF 35 f/2 is nearly neutral and would have an almost flat plot, but with this light source, does show a little red shift. Also included are the Zeiss ZF 35 f/2, AF-S 50 f/1.4G and the 45mm PC-E:

The next group includes some interesting normal-range zooms. The 24-70 has a strong red shift, and the 28-70 tends to be neutral to slightly blue with most light sources. The AF 35-135 f/3.5-4.5 has even stronger blue tendency in most cases, but appears essentially flat with this test source:

Finally, we have the two versions of the 70-200 VR. You can see how the spectral response has been made slightly closer to neutral on the new VR II with a modest lift in the cyan-blue range:

mmullen Veteran Member • Posts: 4,285
Re: Part 2: Aperture Control Accuracy

Marianne Oelund wrote:

It is easy to check your own lenses. Make sure you are working under a steady light source (for example, typical household fluorescent lights must not be used). Using manual exposure, take a shot with the lens wide open, and again with the lens stopped down one stop, at half shutter speed. Continue to narrower apertures in this manner, if desired. Then check the image histograms; if they are all essentially identical, your lens is accurate. If you notice significant left-right shifts, then there is an accuracy control problem, which you may want to use exposure compensation to correct.

Marianne, thanks for all the info - interesting stuff.

However, please be aware that the test procedure you provide for checking the accuracy of a lenses aperture mechanism actually is measuring two things at once (the accuracy of the shutter timing at various shutter speeds is the uncontrolled variable).

And shutter mechanisms (while controlled by a quartz timer) are mechanical in nature and depend upon the proper lubrication and lack of wear to expose for the indicated time. I can easily see a worn or improperly adjusted shutter being responsible for much of the error observed in the test designed to measure the accuracy of the aperture mechanism.

And please don't assume that a shutter that is 5% slow at 1/250th second will also be 5% slow at 1/125th second or at 1/500th of a second. Due to the nature of focal plane shutters it doesn't work that way.

-- hide signature --

Mike Mullen

 mmullen's gear list:mmullen's gear list
Canon EOS 40D Canon EOS 7D Canon EF 85mm f/1.8 USM Canon EF 300mm f/2.8L IS USM Canon EF-S 10-22mm F3.5-4.5 USM +6 more
krisw Regular Member • Posts: 201
Re: Part 1: Lens T-stops

go take pictures

Marianne Oelund wrote:

In photography, we generally use the f-stop of a lens to tell us how bright our images will be. Using f-stops to compare the brightness of our various lens options would be perfectly accurate, if it were not for variations in lens transmission efficiency. A convenient way to include the effect of light loss in a lens, is to use T-stops (transmission stops), which are equivalent to the f-stop of a lossless lens that projects the same image brightness.

Some people like to estimate this, by counting the number of air-to-glass transitions in the lens design, and applying a "reasonable" loss factor for each. As it turns out, this can be wildly inaccurate. There is no substitute for actual measurement.

All of these test results are for the green channel only, as that makes up the greatest share of the apparent image brightness. Measurements were made at the center of the image, thus do not include any vignetting effects. Red and blue channel data was also collected, but this will not be discussed until part 3, about lens color cast.

When using autoexposure, the camera's metering takes account of light losses in the lens, so this information is not relevant to metering errors or issues unless you use an external meter. It simply tells you how close your lens comes to its ideal image brightness. You will see that the values span a surprising range, from as high as 96%, down to less than 70%.

For each lens, I give the measured green-channel transmission efficiency in percent, and then the lens T-stop. For example, an f/4 lens with an efficiency of 85% has a T-stop of 4/SQRT(0.85) = 4.34, so its T-stop is written T/4.34. In making these calculations for lenses with nominal f-stops which are not integers, I used the following 3-digit values for the base f-stop: 1.41 for f/1.4; 1.78 for f/1.8; 2.83 for f/2.8; 3.56 for f/3.5.

Prime Lenses

AF DX 10.5mm f/2.8D: 85%, T/3.08
AF 14mm f/2.8D: 74%, T/3.29
AIS 28mm f/2: 83%, T/2.19
Zeiss ZF 35mm f/2: 85%, T/2.17
AF 35mm f/2D: 92%, T/2.09
PC-E 45mm f/2.8: 90%, T/2.99
AIS 50mm f/1.8: 95%, T/1.83
AF 50mm f/1.4D: 91%, T/1.48
AF-S 50mm f/1.4G: 91%, T/1.48
AF Micro 60mm f/2.8D: 83%, T/3.11
Hartblei 65mm f/3.5 T/S: 96%, T/3.62
AF 85mm f/1.8D: 81%, T/1.97
AF Micro 105mm f/2.8D: 82%, T/3.13
AF 135mm f/2D DC: 86%, T/2.16
AF-S 200mm f/2G VR: 79%, T/2.24
AF-S 400mm f/2.8G VR: 76%, T/3.25

Zoom Lenses

AF-S 14-24mm f/2.8G: 90%, T/2.99 (meas. at 20mm)
AF-S 17-35mm f/2.8D: 88%, T/3.01 (meas. at 24mm)
AF-S 24-70mm f/2.8G: 88%, T/3.02 (meas. at 50mm)
AF-S 24-85mm f/3.5-4.5G: 83%, T/4.39 (meas. at 35mm f/4 nominal)
AF-S 28-70mm f/2.8D: 80%, T/3.16 (meas. at 50mm)
AF 35-135mm f/3.5-4.5D: 72%, T/4.73 (meas. at 66mm f/4 nominal)
AF-S 70-200mm f/2.8G VR 1: 70%, T/3.38 (meas. at 200mm)
AF-S 70-200mm f/2.8G VR 1 plus TC-14: 67%, T/4.90
AF-S 80-200mm f/2.8D: 70%, T/3.39 (meas. at 200mm)
Sigma 120-300mm f/2.8: 72%, T/3.34 (meas. at 200mm)
AF-S 200-400mm f/4G VR: 61%, T/5.11 (meas. at 300mm)

I suspect that last entry may cause some shock, but yes it's true, our $6K f/4 lens is only T/5.1. As consolation, it's extremely color accurate, as you will see later.

heavyone Regular Member • Posts: 102
Re: Photons Missing In Action

Don't different frequencies reach the sensor a different times ? . Is that time difference enough to disturbed to quality of the image ? Those photons which are wandering in the lens , do some of them reach the sensor but at a latter time ? I'm just learning about light principals so please excuse me if my question seem out of place. I'm just visualizing light as I see sound. In acoustics different frequencies move at different speeds and this phase difference makes for some weird distortion artifacts that change the percieved sound. Not only that but in electronics negative feedback which is added to the original signal if in high quantities (bigger phase difference) distorts the sound because of the gross phase difference . Does this same phenomenon exist in light principals ? Are the difference in color really differences in the frequency of the photons ? In other words , Frenquency B would be seen by our eyes as Blue, Frenquency A as red and so on. Of course I'm talking about photons.

heavyone Regular Member • Posts: 102
Re: Part 3 Addendum: Spectrogram examples
1

And I thought photography was only about cameras and composition...well, I'm going to have to deal with it now.

BogdanSandulescu
BogdanSandulescu Regular Member • Posts: 251
Re: Photons Missing In Action

I just want to say THX .
--
FujiFilm FinePix S5 Pro
Tokina AT-X M35 Pro DX
http://www.BogdanSandulescu.Ro

 BogdanSandulescu's gear list:BogdanSandulescu's gear list
Canon EOS-1D X Canon EOS-1D X Mark II Canon EF 35mm F1.4L USM Canon EF 50mm f/1.2L USM Canon EF 85mm f/1.2L II USM +3 more
LGO New Member • Posts: 5
Re: Part 1: Lens T-stops
1

krisw wrote:

go take pictures

Understanding this helps you to get better photos.

Thanks Marianne.

 LGO's gear list:LGO's gear list
Nikon D800 Olympus E-M1 Nikon D850 Olympus 40-150mm F2.8 Pro
OP Marianne Oelund Veteran Member • Posts: 7,777
I do not need to make that assumption
1

mmullen wrote:

And please don't assume that a shutter that is 5% slow at 1/250th second will also be 5% slow at 1/125th second or at 1/500th of a second. Due to the nature of focal plane shutters it doesn't work that way.

I am making no assumptions. I have measured the shutter speeds, and their error is far less than the aperture-dependent exposure variations I have found.

You may find the shutter speed measurements here:
http://forums.dpreview.com/forums/read.asp?forum=1021&message=31437081

If you suspect that shutter speed is a significant factor in the aperture-accuracy measurements, you are in for a surprise.

OP Marianne Oelund Veteran Member • Posts: 7,777
That was not called for
7

krisw wrote:

go take pictures

The 600,000 that I took last year are not enough for you?

If you do not value the information that I provide, you may ignore it. There are others who show their appreciation.

HiRez Regular Member • Posts: 458
Re: That was not called for

Marianne Oelund wrote:

krisw wrote:

go take pictures

The 600,000 that I took last year are not enough for you?

You can't be serious about that, can you? 600,000 pictures in one year? That's 3-4 shots per minute, 8 hours per day, every day of the year.

lovEU Veteran Member • Posts: 3,135
Re: That was not called for

HiRez wrote:

Marianne Oelund wrote:

krisw wrote:

go take pictures

The 600,000 that I took last year are not enough for you?

You can't be serious about that, can you? 600,000 pictures in one year? That's 3-4 shots per minute, 8 hours per day, every day of the year.

I don't know why LGO pulled out this old thread and I guess Marianne Oelund didn't realize how old it was. Anyway, it's simple math to get an idea how such an amount of images is taken. By looking at her avatar you'd know about her main genre which is skating. Now, if one is doing say 60 shootings a year in a rink this means 10000 images per event. Since this genre includes taking high speed sequences that's quite realistic, I think.

I don't understand why a long time poster like you would question one of the very well known MVPs around here

Anyway, it would be interesting to learn more about Marianne Oelund's work flow -- it's really beyond my imagination how to handle that amount of images in post (says someone who is struggling with 500 pics actually :-))
--
regards, eric

Keyboard shortcuts:
FForum MMy threads