MTF50 - How to compare between systems

[Radu B]

I've read on the forum about the Oly 12-50 f/3.5-6.3 ED EZ and most people seem to agree it's an ok lens if you can find it very cheap.

Looking at lenstip review, here if cam be seen that the MTF50 value is ~ 60 wide open.

Instead, for Canon EF 24-105 mm f/4L IS USM , presumably a very good lens, the MTF50 value is ~45

In both cases the value is given in lpmm (line pairs per mm, at the image plane).

I'm explaining it to myself like this: The ff sensor is ~35mm in heigt and the m43 sensor is ~17mm in height. Considering both lenses give a similar field of view at 24mm (canon) / 12mm (oly), but the m43 image is projected on a sensor that has half the height of the ff one, then the MTF50 result of the m43 system should be double the one of the FF system in order to obtain the same resolution (as far as this metric is concerned).

In this case, the Oly 12-40 F2.8 would match and slightly exceed the canon.

Is this correct ?

 

[Stejo]

lpmm describes the amount of detail found within each mm of the sensor's height. It has to do with the resolving power of the lens, the pixel density of the sensor, existence of AA filters etc.

In your example, the FF system resolves 45lpmm. And since it has a sensor height of 24mm, it resolves a total of 1080 line pairs across the entire frame. Or well, it would if it was uniformly sharp across the entire frame, which it isn't, but let's disregard this for now.

The MFT system resolving 60lpmm with a sensor height of 13mm, ends up resolving a total of 780 line pairs.

The MFT lens itself is probably sharper, as most lenses designed for smaller sensors tend to be, aided by the fact that it has to cover a smaller image circle, but the end result is still in favor of the FF system in this case.

In order to match the 45lpmm performance, you'd need a lens resolving 45*(24/13)=83lpmm.

You can easily compare the sharpness between two different systems with the above calculation, but keep in mind that this only tells you the relative performance of the specific lenses on the specific sensors.

If you're interested in the resolving power of the lenses themselves, regardless of system, you need to compare them on the same sensor.

--
https://www.flickr.com/photos/jjsterg/

 

[Great Bustard]

I've read on the forum about the Oly 12-50 f/3.5-6.3 ED EZ and most people seem to agree it's an ok lens if you can find it very cheap.

Looking at lenstip review, here if cam be seen that the MTF50 value is ~ 60 wide open.

Instead, for Canon EF 24-105 mm f/4L IS USM , presumably a very good lens, the MTF50 value is ~45

In both cases the value is given in lpmm (line pairs per mm, at the image plane).

I'm explaining it to myself like this: The ff sensor is ~35mm in heigt and the m43 sensor is ~17mm in height. Considering both lenses give a similar field of view at 24mm (canon) / 12mm (oly), but the m43 image is projected on a sensor that has half the height of the ff one, then the MTF50 result of the m43 system should be double the one of the FF system in order to obtain the same resolution (as far as this metric is concerned).

In this case, the Oly 12-40 F2.8 would match and slightly exceed the canon.

Is this correct ?

The photographically meaningful measure with regards to resolution is lw/ph -- line widths per picture height* on the photo. This is what is measured when testing a lens. Lenstip converts this measure to lp/mm (line pairs per millimeter on the sensor) by taking their lw/ph measurement, dividing it by two, and then dividing it by the sensor height. So, to undo Lenstip's arithmentic, take their lp/mm measurement, multiply by 2, and then multiply by the sensor height to get the lw/ph measurement.

*By "sensor height", we normally mean distance from the center along the diagonal. For example, the "sensor height" for an mFT sensor (17.3mm x 13.0mm -- diagonal = 21.64mm) would be 21.64mm / 2 = 10.82mm, and the "sensor height" for a FF sensor (36mm x 24mm -- diagonal = 43.27mm) would be 43.27mm / 2 = 21.64mm. That said, some may use the actual sensor height for the conversion (13mm for mFT and 24mm for FF), but this will give us the comparative resolution for a FF photo cropped to 4:3.

So, if Lenstip recorded 60 lp/mm for an mFT lens, it would give 60 x 2 x 10.82 = 1298 lw/ph. If Lenstip recorded 45 lp/mm for a FF lens, it would yield 45 x 2 x 21.64 = 1948 lw/ph.

As a side, if a FF lens has half the resolution (lp/mm) as an mFT lens (keeping in mind that Lenstip doesn't actually measure the resolution of the lens, but instead the system resolution, lw/ph, and convert that value to a faux lp/mm), the sensors have the same pixel count and the same AA filter, then the resolution on the photo (lw/ph) will be the same.

As an additional side, I once asked about how the lack of an AA filter affects the MTF-50 score compared to a "normal" AA filter in the PST forum. The best answer I remember getting was that the lack of an AA filter would result in a 20% greater MTF-50 score, all else equal.

In addition, a p% increase in pixel count will result in at most a [sqrt (1 + p%) - 1] percent increase in MTF-50, all else equal. For example, going from 16 MP to 20 MP (25% increase) will yield at most a sqrt (1.25) -1 = 11.8% increase in the MTF-50 score. However, lens aberrations, motion blur, diffraction, and noise can all serve to make for a lesser increase in practice (I use half the maximum possible increase as a rule of thumb).

 

[Radu B]

*By "sensor height", we normally mean distance from the center along the diagonal.

This is new info to me. I know some charts show MTF as distance from the center, but I asumed sensor height is just that (not the diagonal).

For example, the "sensor height" for an mFT sensor (17.3mm x 13.0mm -- diagonal = 21.64mm) would be 21.64mm / 2 = 10.82mm, and the "sensor height" for a FF sensor (36mm x 24mm -- diagonal = 43.27mm) would be 43.27mm / 2 = 21.64mm. That said, some may use the actual sensor height for the conversion (13mm for mFT and 24mm for FF), but this will give us the comparative resolution for a FF photo cropped to 4:3.

So, if Lenstip recorded 60 lp/mm for an mFT lens, it would give 60 x 2 x 10.82 = 1298 lw/ph. If Lenstip recorded 45 lp/mm for a FF lens, it would yield 45 x 2 x 21.64 = 1948 lw/ph.

I will dig a bit deeper into the math behind this to sort out the units in my head, but the result above is in-line with the actual behavior, as far as I've seen.

Thanks for the info !

 

[Radu B]

lpmm describes the amount of detail found within each mm of the sensor's height. It has to do with the resolving power of the lens, the pixel density of the sensor, existence of AA filters etc.

In your example, the FF system resolves 45lpmm. And since it has a sensor height of 24mm, it resolves a total of 1080 line pairs across the entire frame. Or well, it would if it was uniformly sharp across the entire frame, which it isn't, but let's disregard this for now.

The MFT system resolving 60lpmm with a sensor height of 13mm, ends up resolving a total of 780 line pairs.

The MFT lens itself is probably sharper, as most lenses designed for smaller sensors tend to be, aided by the fact that it has to cover a smaller image circle, but the end result is still in favor of the FF system in this case.

In order to match the 45lpmm performance, you'd need a lens resolving 45*(24/13)=83lpmm.

You can easily compare the sharpness between two different systems with the above calculation, but keep in mind that this only tells you the relative performance of the specific lenses on the specific sensors.

If you're interested in the resolving power of the lenses themselves, regardless of system, you need to compare them on the same sensor.

 

[danart]

I've read on the forum about the Oly 12-50 f/3.5-6.3 ED EZ and most people seem to agree it's an ok lens if you can find it very cheap.

Looking at lenstip review, here if cam be seen that the MTF50 value is ~ 60 wide open.

Instead, for Canon EF 24-105 mm f/4L IS USM , presumably a very good lens, the MTF50 value is ~45

In both cases the value is given in lpmm (line pairs per mm, at the image plane).

I'm explaining it to myself like this: The ff sensor is ~35mm in heigt and the m43 sensor is ~17mm in height. Considering both lenses give a similar field of view at 24mm (canon) / 12mm (oly), but the m43 image is projected on a sensor that has half the height of the ff one, then the MTF50 result of the m43 system should be double the one of the FF system in order to obtain the same resolution (as far as this metric is concerned).

In this case, the Oly 12-40 F2.8 would match and slightly exceed the canon.

Is this correct ?

The photographically meaningful measure with regards to resolution is lw/ph -- line widths per picture height* on the photo. This is what is measured when testing a lens. Lenstip converts this measure to lp/mm (line pairs per millimeter on the sensor) by taking their lw/ph measurement, dividing it by two, and then dividing it by the sensor height. So, to undo Lenstip's arithmentic, take their lp/mm measurement, multiply by 2, and then multiply by the sensor height to get the lw/ph measurement.

*By "sensor height", we normally mean distance from the center along the diagonal. For example, the "sensor height" for an mFT sensor (17.3mm x 13.0mm -- diagonal = 21.64mm) would be 21.64mm / 2 = 10.82mm, and the "sensor height" for a FF sensor (36mm x 24mm -- diagonal = 43.27mm) would be 43.27mm / 2 = 21.64mm. That said, some may use the actual sensor height for the conversion (13mm for mFT and 24mm for FF), but this will give us the comparative resolution for a FF photo cropped to 4:3.

So, if Lenstip recorded 60 lp/mm for an mFT lens, it would give 60 x 2 x 10.82 = 1298 lw/ph. If Lenstip recorded 45 lp/mm for a FF lens, it would yield 45 x 2 x 21.64 = 1948 lw/ph.

As a side, if a FF lens has half the resolution (lp/mm) as an mFT lens (keeping in mind that Lenstip doesn't actually measure the resolution of the lens, but instead the system resolution, lw/ph, and convert that value to a faux lp/mm), the sensors have the same pixel count and the same AA filter, then the resolution on the photo (lw/ph) will be the same.

As an additional side, I once asked about how the lack of an AA filter affects the MTF-50 score compared to a "normal" AA filter in the PST forum. The best answer I remember getting was that the lack of an AA filter would result in a 20% greater MTF-50 score, all else equal.

In addition, a p% increase in pixel count will result in at most a [sqrt (1 + p%) - 1] percent increase in MTF-50, all else equal. For example, going from 16 MP to 20 MP (25% increase) will yield at most a sqrt (1.25) -1 = 11.8% increase in the MTF-50 score. However, lens aberrations, motion blur, diffraction, and noise can all serve to make for a lesser increase in practice (I use half the maximum possible increase as a rule of thumb).

Someone pointed to this as how lenstip or more specifically how LW/PH is calculated from lenstips lp/mm measures. This is wrong on both counts - One it is a different testing method. lp/mm (line pairs per millimeter) is a localized spatial frequency measurement taken directly at the sensor plane. Lenstip report lp/mm from their optical bench tests. It’s not scaled to the frame height, and it’s not the same as LW/PH.

Secondly that LW/PH is the diagonal reference to picture height is wrong. The diagonal is the dimension not the height. The standard metric - LW/PH (line widths per picture height) is defined specifically by the image height, i.e. the vertical dimension of the sensor.. It’s been the convention in optics and imaging science for decades. The reason it is height and not dimension is because it accounts for different formats - if using the diagonal or dimension of the sensor, it skews results between two formats - picture height is the constant across formats which normalises results between them.

In other words, you take lenstips results as presented and tested on the specific sensor and you cannot reliably translate lenstips lp/mm to LW/PH or another sensor makeup - other than retesting the lens on a different sensor. Lenstip themselves have retested lenses for different cameras because simply applying a basic formula would lead to an incorrect and misleading measure.

 

[James Stirling]

I've read on the forum about the Oly 12-50 f/3.5-6.3 ED EZ and most people seem to agree it's an ok lens if you can find it very cheap.

Looking at lenstip review, here if cam be seen that the MTF50 value is ~ 60 wide open.

Instead, for Canon EF 24-105 mm f/4L IS USM , presumably a very good lens, the MTF50 value is ~45

In both cases the value is given in lpmm (line pairs per mm, at the image plane).

I'm explaining it to myself like this: The ff sensor is ~35mm in heigt and the m43 sensor is ~17mm in height. Considering both lenses give a similar field of view at 24mm (canon) / 12mm (oly), but the m43 image is projected on a sensor that has half the height of the ff one, then the MTF50 result of the m43 system should be double the one of the FF system in order to obtain the same resolution (as far as this metric is concerned).

In this case, the Oly 12-40 F2.8 would match and slightly exceed the canon.

Is this correct ?

The photographically meaningful measure with regards to resolution is lw/ph -- line widths per picture height* on the photo. This is what is measured when testing a lens. Lenstip converts this measure to lp/mm (line pairs per millimeter on the sensor) by taking their lw/ph measurement, dividing it by two, and then dividing it by the sensor height. So, to undo Lenstip's arithmentic, take their lp/mm measurement, multiply by 2, and then multiply by the sensor height to get the lw/ph measurement.

*By "sensor height", we normally mean distance from the center along the diagonal. For example, the "sensor height" for an mFT sensor (17.3mm x 13.0mm -- diagonal = 21.64mm) would be 21.64mm / 2 = 10.82mm, and the "sensor height" for a FF sensor (36mm x 24mm -- diagonal = 43.27mm) would be 43.27mm / 2 = 21.64mm. That said, some may use the actual sensor height for the conversion (13mm for mFT and 24mm for FF), but this will give us the comparative resolution for a FF photo cropped to 4:3.

So, if Lenstip recorded 60 lp/mm for an mFT lens, it would give 60 x 2 x 10.82 = 1298 lw/ph. If Lenstip recorded 45 lp/mm for a FF lens, it would yield 45 x 2 x 21.64 = 1948 lw/ph.

As a side, if a FF lens has half the resolution (lp/mm) as an mFT lens (keeping in mind that Lenstip doesn't actually measure the resolution of the lens, but instead the system resolution, lw/ph, and convert that value to a faux lp/mm), the sensors have the same pixel count and the same AA filter, then the resolution on the photo (lw/ph) will be the same.

As an additional side, I once asked about how the lack of an AA filter affects the MTF-50 score compared to a "normal" AA filter in the PST forum. The best answer I remember getting was that the lack of an AA filter would result in a 20% greater MTF-50 score, all else equal.

In addition, a p% increase in pixel count will result in at most a [sqrt (1 + p%) - 1] percent increase in MTF-50, all else equal. For example, going from 16 MP to 20 MP (25% increase) will yield at most a sqrt (1.25) -1 = 11.8% increase in the MTF-50 score. However, lens aberrations, motion blur, diffraction, and noise can all serve to make for a lesser increase in practice (I use half the maximum possible increase as a rule of thumb).

Someone pointed to this as how lenstip or more specifically how LW/PH is calculated from lenstips lp/mm measures. This is wrong on both counts - One it is a different testing method. lp/mm (line pairs per millimeter) is a localized spatial frequency measurement taken directly at the sensor plane. Lenstip report lp/mm from their optical bench tests. It’s not scaled to the frame height, and it’s not the same as LW/PH.

Secondly that LW/PH is the diagonal reference to picture height is wrong. The diagonal is the dimension not the height. The standard metric - LW/PH (line widths per picture height) is defined specifically by the image height, i.e. the vertical dimension of the sensor.. It’s been the convention in optics and imaging science for decades. The reason it is height and not dimension is because it accounts for different formats - if using the diagonal or dimension of the sensor, it skews results between two formats - picture height is the constant across formats which normalises results between them.

In other words, you take lenstips results as presented and tested on the specific sensor and you cannot reliably translate lenstips lp/mm to LW/PH or another sensor makeup - other than retesting the lens on a different sensor. Lenstip themselves have retested lenses for different cameras because simply applying a basic formula would lead to an incorrect and misleading measure.

Loading…

www.dpreview.com