T-stop measures actual light transmission in a lens. Therefore, different f/2.8 lenses can have different T-stops. T-stops are usually only given for cine lenses. For photography lenses, is there any practical way to measure the light transmission in a lens?
The only way to get an absolute measurement is to have a lightsource of known luminance (which requires use of a calibrated luminance meter) and then measure image of that lightsource as projected by the lens.
You can get a relative value for an unknown lens by shooting a uniform lightsource with two lenses, one of which with a known the T-stop (several websites provide this value e.g. DXO mark). You can then work out the T-stop by comparing brightness of these two images.
Both lenses should be set to the same F-stop and focused to infinity. The accuracy of your results will depend upon a) how accurately the F-stop is stated for the lenses (manufacturers round the values, usually in their favour) and how linear the sensor response is. If you avoid the top end of the exposure range, shoot RAW and use something like RawAnalyzer, the linearity should be good enough.
Finally I wouldn't place too much emphasis on T-stop performance unless you are severely light-limited; there are often other factors which are far more critical.
First, what is not required for such measurement?
calibrated light source (because transmission is about relative, not absolute intensity)
calibrated light meter (because, again, transmission is about comparing light intensity, not measuring absolute values)
We will, however, need a reliable constant intensity light source and an image sensor that reports linear light intensity, which, luckily is your regular camera sensor, because that's how cmos physics work.
This should be enough for comparing T stop values or transmission in lens - but remember to verify the manufacturer declared f-stop (aperture), it is sometimes misleading.
Do we need a lens with previously known (calibrated) transmission to determine the absolute T stop value or transmission of our unknown lens? No - because we can always compare it to "no lens" configuration, which is by definition, 100% transmission.
This case is slightly different and more interesting, as this time we will be using the collimated light source (otherwise we could not control the amount of light entering our lensless sensor)
narrow collimated beam * ) = = = = = = = = = = = = = = = [ ] ^ ^ ^ light source..: :..collimating lens :...camera without lens
* ) = = = = = = = = = = = = = = = O[ ] ^ ^ lens under test...: :..camera
T-stop can be measured for a lens by shining a precisely regulated light source through a lens in a controlled testing environment with no other extraneous sources of light and measuring by a calibrated meter on the other end. Unless you have a sophisticated optical lab there is no real way to precisely measure the actual transmission through a lens.
The most comprehensive database I have found for lenses tested in a controlled lab is at DxO Mark. Here is the page for the Canon EF 50mm f/1.2 L. One of the measurements listed on the Scores tab is Transmission. On the Scores tab you can even select from as many different camera bodies as they tested with the particular lens. Click on the Measurements tab and then Transmission to see the results of the test on a chart with another showing the measured difference between the aperture setting and the actual transmission. You can also select up to three different lenses and compare them.
Maybe a more practical way to determine the T-Stop of a lens?:
Arrange a uniformly lit white surface big enough to cover the FOV of a lens.
Use your camera with fixed shutter speed (M-Mode), ISO, WB etcerera to make an exposure of the surface without a lens.
Put the lens on the same camera with the same settings, and make a second exposure of the surface.
Compare these two images and use software with the ability to change exposure in post.
Brighten up the second exposure until it matches the first one.
Note the EV value you needed and the shutter speed used.
Use formula N=sqrt(2EV*t) with N is t/stop Number, t is duration Time of shutter speed — the camera settings.
alternatively make the exposures in camera (A-Mode) being same and calculate both with the noted shutter speeds chosen by the camera and take the difference of both values.
Who can confirm this?