What do those MTF-50 numbers really mean?

[noreast]

Hello,
Different sites report different MTF-50 numbers for the same lens. For example, Digital Photography Review reports MTF-50 well below 1500 LW/PH at the center for the Nikkor AF-S DX 35mm f/1.8 G while Photozone reports 2027, both readings at F1.8.

The DPR sets the Nyquist Frequency (the theoretical limit of the sensor's resolution) of the D300 sensor (their test camera) at little below 1500. According to these numbers, this lens should be soft at the center at F1.8 on a D300 according to DPR, and out-resolve the sensor according to Photozone.

As a measure a physical phenomenon, shouldn't the MTF-50 test give consistent readings (excluding manufacturing variations between test lenses)?

If I see a review for lens A at site X and see a review for lens B at site Y, how can I compare the MTF-50 numbers of the two lenses and draw a conclusion which lens is more adequately sharp for my camera?

[McQ]

I would expect there to be some variation between lens tests, but 1500 versus 2030 is a big difference. The difference is likely due to varying methodology or definitions as opposed to anything else. For this reason, one should only really compare relative values performed by the same person/website if they want accurate comparisons between lenses. Pay much less attention to the absolute MTF numbers.

But that still doesn't answer the question: why the discrepancy? I would check each of their definitions for MTF-50. In particular, what do they mean by "picture height" in the line widths per picture height metric (LW/PH) ? For example, is it the height of the sensor or is it the height of an equivalent 35mm sensor -- and is it in the vertical or horizontal direction. Further, is the LW/PH being measured at the direct center of the frame, or is it somewhat offset? Is LW/PH measured using the RAW file, and is "standard sharpening" applied prior to this measurement? Hopefully all of this will give you an idea of all the reasons for variation. It'd be nice if there was a quick and easy number that correlates with lens quality.

But...it might be something simple, like a mistake in either of their tests or calculations. That's something that shouldn't be overlooked.

[noreast]

Hello McQ, thanks for the reply.

Is LW/PH measured using the RAW file, and is "standard sharpening" applied prior to this measurement?

If MTF-50 tests are performed on images extracted from a camera, how can someone claim that a lens over-resolved the sensor? Not only does DPR tell you that a lens outresolves the sensor, they also tell you by how much. Can these numbers be a mere extrapolation?

For example, is it the height of the sensor or is it the height of an equivalent 35mm sensor

Does the multiplication factor of the, say APC-S, sensor apply to MTF-50 numbers?

For this reason, one should only really compare relative values performed by the same person/website if they want accurate comparisons between lenses. Pay much less attention to the absolute MTF numbers

True. Unfortunately, I have not yet found such a person/website . If one is trying to make a decision to buy on of two different lenses, usually reviews for the two lenses do not appear on the same website. It would really be great if, as you mentioned, a standard metric was adopted.

[McQ]

If MTF-50 tests are performed on images extracted from a camera, how can someone claim that a lens over-resolved the sensor? Not only does DPR tell you that a lens outresolves the sensor, they also tell you by how much. Can these numbers be a mere extrapolation?

I'm going to reserve any speculation about why this might be the case. There's just too many variables. But yes...it is quite odd. Photozone does use Norman's Imatest for their lens quality testing, which produces great results if the tests are performed correctly.

Does the multiplication factor of the, say APC-S, sensor apply to MTF-50 numbers?

Strictly speaking, no it should not. But it becomes less clear when you're testing a 35mm lens on a cropped camera. One could argue that you're not being fair to a 35mm lens by only looking at LW/PH for a cropped sensor -- that is, if comparing the lens itself is the primary goal. In this case though it's a Nikon DX lens.