I stumbled upon an MTF chart, which looks wrong. It's for Kowa LM5NCL lens: https://www.kowa-lenses.com/en/lm5ncl-mp-industrial-lens-c-mount?number=10097. Here is the chart in question:

enter image description here

It shows the worst performance at the center TS 0.0000 MM, progressively improving as you move towards the edge TS 4.0000 MM. Is it possible? The data sheet claims a believable resolution of 100lp/mm center and 40lp/mm corner. Did someone mislabel the curves?


This post is related more to machine vision than still photography, therefore the chart may be a bit confusing. Photography substack was the only one with frequent posts about MTF, so I chose to post here. I'm suggesting that the labels should be swapped in this manner:

  • TS 0.0000 MM --> TS 4.0000 MM
  • TS 2.4000 MM --> TS 3.2000 MM
  • TS 3.2000 MM --> TS 2.4000 MM
  • TS 4.0000 MM --> TS 0.0000 MM

Here is a graph of MTF of LN4NCL lens of similar design, but with a shorter FL, to illustrate correct labelling:

enter image description here

  • \$\begingroup\$ It seems to show the best performance at or near the center, which is charted in the upper left hand corner of the chart. Perfect MTF is 1.0 (100%) where all contrast is preserved through the lens. Unlike most MTF charts that move from center to edge at a a constant spatial frequency as the line is drawn from left to right, this one shows the decreasing MTF measurement as the spatial frequency is increased. \$\endgroup\$
    – Michael C
    Aug 10, 2022 at 6:41
  • \$\begingroup\$ @MichaelC Note that the center has the worst MTF and the edge has the best MTF. Pay attention to TS labels. \$\endgroup\$ Aug 10, 2022 at 16:10

1 Answer 1


It's an unusual chart, but I'm not reading it the same as you are. The individual lines are labeled Tangential (Meridional) and Sagittal rather than solid/dashed lines as typical. The lines are color coded for distance from center rather than that being the X axis as typical. And the X axis represents cycles (LP) per mm rather than that being the color's meaning as typical.

What I see is that the center reaches 100LP/mm with no deviation of the T/S lines; but at an extremely low contrast (MTF) of only about 1-2%. The 4mm lines also reach 100LP/mm but at 30% contrast for tangential lines and 55% contrast for sagittal lines. So you have a very sharp center (no deviation) that is able to resolve a very high frequency with a low contrast. And a near edge (80% diagonal) that can resolve a high frequency of sagittal lines better than it can resolve tangential lines; but seemingly both to a higher level of contrast than the center can.

A few things to note...

  • The chart is for infinity, which means it is almost certainly calculated using their design software (as almost all MTF charts are). Which means it is theoretical, and not comparable to someone else's MTF charts (because they use different software with different assumptions and parameters).
  • The chart is for a monochromatic wavelength of 660nm which is near the limits of visible light/near IR. Might be useful for a monochromatic machine vision camera, not so much for a more typical color image camera.
  • Test results for a single wavelength will generate very different results compared to test results for white/visible light.
  • Modern 1/1.8 sensor designs are reaching 540 pixels/mm; so the nyquist/usable resolution limit is somewhere near 270LP/mm.
  • I suspect there is no T/S line spread for the center of the lens on any of the Kowa MTF charts because only one set of lines are used at that position (there isn't really any difference between T/S oriented lines at 0.0000; just rotate/roll the lens and they swap/interchange).

So, is the chart mislabeled? I doubt it since the charts are auto/software generated/labeled. Almost all of the other Kowa MTF charts I have looked at are color coded as R>Y>G>B from center towards the edge; as does the example in the added note. That would indicate not only mislabeling, but also miscoloring; which I think is that much less likely (relabeling doesn't fix the out of sequence colors). And I did find another MTF chart that uses the same B>G>R>Y (the 4.4mm LM4NCM), which seems to indicate that the 4.4/4.5mm charts are just older and using another convention. Also, the blue line on the 4.5mm chart does not show a T/S spread; which indicates that it IS the center of the lens.

Is the chart believable? I would say probably, as much as any other theoretical MTF chart is.

Is the chart useful? I would say not so much from the perspective of still "photography;" or rather, the lens is not... I suspect it is a specialized design for a specific machine vision purpose; or it's an artifact of the method used for correcting an UWA.

  • \$\begingroup\$ So, is the chart mislabeled? I doubt it - have a look at the note I added to my question. \$\endgroup\$ Aug 10, 2022 at 17:07
  • \$\begingroup\$ @PaulJurczak, I updated my answer with a few more thoughts/specifics after looking at a number of other charts. Basically, it seems the lens center never shows a T/S line separation; probably because only one set of lines are used/fit at 0.0000mm.... which would indicate the blue line IS center on that chart. \$\endgroup\$ Aug 10, 2022 at 18:35
  • \$\begingroup\$ If blue line IS center than MTF at 100lp/mm at the center is practically zero, contrary to stated resolution of 100lp/mm at the center. If yellow lines are center than claimed resolution of 100lp/mm at the center makes sense with avg. MTF of 0.43. \$\endgroup\$ Aug 11, 2022 at 4:27
  • \$\begingroup\$ The % contrast is very low for the center at 100cycles... lower than I would expect useful, even for machine vision. But if the blue line is edge instead; why is there no T/S spread, and where would the 60 cycles stated resolution come from? \$\endgroup\$ Aug 11, 2022 at 13:40
  • \$\begingroup\$ IDK where they are getting the stated resolutions from; I can't make sense of it from the charts. The stated center resolution for the 4.4mm LM4NCM is "120" and that's not even on the supplied charts. It also doesn't say at what level of contrast they are stating the resolution for; it doesn't seem to be the standard MTF50. I did note that the "wavelength range" is visible light+ (380-780nm) and the chart if for single wavelength. I think that there just isn't enough information provided... perhaps the stated resolutions are some kind of composite number or calculated differently/separately. \$\endgroup\$ Aug 11, 2022 at 13:43

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Not the answer you're looking for? Browse other questions tagged or ask your own question.