MTF - what are they good for???

greenspun.com : LUSENET : Large format photography : One Thread

What do MTF curves mean in the real World? For example, comparing the new Schneider XL 110 at http://schneideroptics.com/large/super_symmar/110xl/mtf2.html with the Schneider Angulon 75 at http://schneideroptics.com/large/super_angulon/75/mtf2.html I see very little difference, if at all, the 75 Angulon has better resolution than the 110 in the corners. Yet everyone seems to agree that the XL 100 is so much sharper than other LF lenses?!?

Another question: Are lp/mm and cycles/mm the same? The curves on the Schneider page are for 5, 10 and 20 lp/mm. Hasselblad publishes MTF with 10, 20, 40 cycles/mm for example see the 100CF at http://www.hasselblad.com/products/lenses/lens_cf/cf100/index.html If lp/mm and cycles/mm are the same, this would suggest that the Hasselblad has twice the resolving power of the Schneider lens? Hard to imagine...

-- Andreas Carl (andreas@physio.unr.edu), February 18, 2000

Answers

I've never used any of the lenses you mention, so I can't say whether any of them are sharper than any other lens. However, in real life we don't judge the overall impression of sharpness by the same methods used in MTF tests. In particular, I believe that we can use higher frequencies than the MTF tests show.

Across manufaturers, and across formats, you are comparing very different things. (a) The manufacturers use different test methods. (b) The 'Blad lenses are quoted for an image height up to 40 mm, the Schneider 110 for up to 144mm. The Schneiders are designed for roughly twice the format, so (for the same enlargement) the acceptable resolution is halved. Well, so goes one argument.

Larger format lenses tend to given lower resolutions on the film, because everything is a trade-off.

Look closely at the numbers. At infinity, 20 lp/mm, the Schneider at f/22 gives more than 60% up to 60mm from the centre. The 'Blad manages 80%, which isn't a huge amount better, given the much smaller format size the lens has to cover.

There is more to the story than mere MTF. The Schneiders you mention cover the same field angle (105 degrees), so the longer one will cover a larger negative. The 110 has slightly worse distortion. And what about the factors they don't mention: coma, chromatic abberation, etc etc.

And finally: there is bound to be variation between samples. Schneider don't claim that these are specifications for their lenses, merely ilustrations, and calculated at that.

-- Alan Gibson (Alan@snibgo.com), February 18, 2000.


Thanks Alan, your points are well taken. I still hope that lp/mm DOES NOT equal cycles/mm. You are right, the Schneider (and this is presumably the top of the line) has over 60% of 20 lp/mm for more than half of its field, but the Hassie has more than 40 cycles/mm for most of its field. Wouldn't this almost entirely negate the advantage of LF over MF (all other things, like chromatic abberations etc. being equal, which as you point out most likely are not equal...still)...

-- Andreas Carl (andreas@physio.unr.edu), February 19, 2000.

I will agree with Alan that these are calculated, not measured. I would disagree with his claim that these curves are not good measures of optical performance.

First, lp/mm is the same as cycles/mm. And yes, the Hassie numbers at 40 lp/mm probably equal the Schneider numbers at 20 lp/mm BUT NOT AT THE SAME f-stop. The Hassie numbers are shown up to f/8. The Schneider numbers at f/22. Physics demands MTF fall off with f-stop (just as "resolution" falls off with f-stop as per Rayleigh criterion). MF lenses are designed to reach their physical limits (diffraction controlled) by f/8 or so since they will be used with moving subjects, and since coverage is not an issue. LF lenses are designed to maximize coverage, and reach diffraction limits by f/16 or so since that is where they will be used.

IF film had perfect MTF (100% at all cycles/mm) then you are right, there would be little advantage to LF over MF. But the image recorded on film is the MTF of the lens TIMES the MTF of the film. For example, most color reversal emulsions still deliver nearly 100% MTF at 20 lp/mm but fall off rapidly between 20 and 50 lp/mm. Take Velvia, the king of MTF. It delivers 100% modulation at 20 lp/mm but only about 60% by 40 lp/mm. So the MF lens must deliver about 1.6 times the modulation at 40 lp/mm just to keep up with LF at 20 lp/mm. And of course granularity is also an issue. So film remains the real issue. Better films help. Provia 100F has finer grain, but its MTF still plummets past 20 lp/mm.

As for your comparison of the 110XL to the 75 SA, I don't think you are reading the curves correctly. At infinity, f/22 and 20 lp/mm, lets look at the corner of a 4x5, about 75mm off axis. For the 110XL, that is about 50% of the maximum image height (50% of 144 mm). I read a tangential MTF of about 45% and a sagittal (radial) of about 62%. For the 75mm, the maximum image height is about 100mm, so we go to 75% of the image height and get a tangential of about 30% and sagittal of about 57%. So it sure looks like the 110 is sharper.

But this is geometry not lens design. Geometry demands that tangential MTF fall off as cos^3 off-axis angle, and sagittal fall off as cos of off-axis angle. The 75mm just cannot do as well in the corners since they are farther off-axis for this shorter focal length. This is also why the Hassie lens looks so good. On 6x6, a 100mm lens is a moderate tele. Image angles are very small, so the curves don't fall off much at all. The same focal length on 4x5 is moderate wide-angle. Image angles are more than twice as large, so the cos and cos^3 fall off is much greater.

So what's this all mean. If you don't need movements and you use the best films available and shoot at optimum apertures for each format, MF comes VERY close to LF in quality. I've been doing some controlled tests with Velvia, matched lenses and high resolution drum scans. For color prints 16x20 or smaller, there is no visible difference between MF and LF. Starting with 20x24 (assuming we use the entire image area) the differences are just barely perceptable. At this size, the MF image has the same fine detail, but at lower contrast. Careful use of USM can minimize the difference, but LF is still slightly better, and grain begins to become apparent in the MF image. By 24x36 the difference is more apparent.

So if you like big prints, like to be able to crop, and like movements, LF still has distinct advantages. If you are not going to make prints larger than 16x20, you can get equivalent results from MF.

-- Glenn C. Kroeger (gkroeger@trinity.edu), February 19, 2000.


My previous email may have sounded too harsh on Alan's comments. I also agree that spatial frequencies higher than 20 lp/mm are important, but I think the reason they aren't published is that they are completely dominated by diffraction effects.

Also, although the Schneider curves are calculated, they DO account for both lateral and longitudinal chromatic aberation via the spectral mixing used to calculate the MFT curves. If Schneider calculated the curves for monochromatic light, they would look better.

-- Glenn C. Kroeger (gkroeger@trinity.edu), February 19, 2000.


" Are lp/mm and cycles/mm the same"

LP/MM does NOT mean lines per MM on an MTF chart.

They mean LINE PAIRS PER MM.

20 lp/mm is 40 lines per mm.

If you would like an in depth explanation of an MTF chart we would be happy to mail you the one Rodenstock prints.

-- Bob Salomon (bobsalomon@mindspring.com), February 19, 2000.



To amplify on Bob's response. On a RESOLUTION chart, lp/mm means a line pairs per mm, each line pair is one black and one white line. MTF is measured using sinusoidal patterns of dark and light. So one cycle/mm would appear to be one dark and one white line, but with a continuous sinusoidal variation of intensity rather than a step function change in intensity.

-- Glenn C. Kroeger (gkroeger@trinity.edu), February 19, 2000.

Moderation questions? read the FAQ