• Log in with Facebook Log in with Twitter
Photo

Comparision of MTF values old vs new tests


  • Please log in to reply
10 replies to this topic

#1 planck

planck

    Newbie

  • Members
  • Pip
  • 5 posts

Posted 20 December 2016 - 12:43 AM

Hi guys,

 

I might have a silly question, might be already asked. I was looking to the resolution test of Tamron AF 17-35mm f/2.8-4 done with D200. MTF is around 2200 in the center and 1900 in the border and it is written that the max (theoretical I believe) is 2320. Then I looked for Nikkor AF-S 20mm f/1.8 G where the max is 4000, measured are around 3900 and 3300. When I normalized the measured values to the max of their own, it seems that MTF of both lenses are the same. I have a D810 with the Tamron and considering the 20mm f/1.8 but if the resolutions will be the same I won't bother for a faster lens. But the price tags and dxo test do not say so. What am I missing?

 

 



#2 mst

mst

    Advanced Member

  • Moderators
  • 1,942 posts
  • LocationWesterwald, Germany

Posted 20 December 2016 - 09:14 AM

Welcome on board :) There are two reasons why your assumptions will not lead to a valid result. First of all, a comparison across several test systems is an invalid approach in itself. It's tempting to 'normalize' the results, but bound to fail because so many factors influence a test system's maximum resolution, among those the sensor characteristics and the raw converter used in the analysis workflow. Short summary: the MTF results are not comparable across different systems (where every test camera we use is a 'system' in this context).

 

In addition, you're comparing apples and oranges anyway :) The AF-S 20/1.8 was tested on the D3x, so a full frame camera. The Tamron review is fairly dated and based on the D200, a DX camera with a smaller sensor. So border and corner values in those reviews refer to very different spots in the image circle of the lenses.

 

I can tell you, however, that the Nikkor does provide better sharpness (and way less CAs), and in addition opens up a few additional creative options with its fast f/1.8 aperture and the corresponding shallow depth of field. On the other hand, you lose the flexibility of the zoom.

 

Give me a little time, and I may be able to provide a few numbers so you can compare yourself :)

 

-- Markus


Editor (Nikon, Leica, Samsung reviews)
photozone.de

#3 Brightcolours

Brightcolours

    Advanced Member

  • Members
  • PipPipPip
  • 4,562 posts
  • LocationThe Netherlands

Posted 20 December 2016 - 09:31 AM

You are missing:

 

  • The Tamrob was tested on a 10MP APS-C camera. The Nikkor on a 24mp FF camera. From the Tamron you only know that its performance is sufficient on a low MP APS-C camera. It is like comparing the center test from a 15mp FF camera to a 24mp FF camera. A lens can seem sharp on lower resolution, but on higher resolution it may show its limits.
  • The Tamron was trested many years ago. Different RAW converter, different version of test software, it all makes comparison tricky.
  • The Tamron has been tested on APS-C. The lens' weak border/corner performance on FF is not shown.
  • The Nikkor is faster, you can use it with bigger apertures.
  • Besides all that, the Nikkor has better AF, better coatings, better CA performance.

 

The only thing you know from the Tamron is that it was quite an impressive lens way back when 10mp APS-C was the norm.



#4 planck

planck

    Newbie

  • Members
  • Pip
  • 5 posts

Posted 20 December 2016 - 03:47 PM

Thanks for the inputs and detailed explanation. But still I would like to have some insight to compare :) So can we say that the values after the normalization is the best that old lens can have with the new cameras and converters? They will probably lower especially in the border but cannot be higher. Would this be true?



#5 mst

mst

    Advanced Member

  • Moderators
  • 1,942 posts
  • LocationWesterwald, Germany

Posted 20 December 2016 - 04:04 PM

No :) Please read both messages above again.

 

But to sum it up again: normalization per se will not work. Reasons explained above.

 

Plus: border on DX and border on FX are two very different spots, you simply cannot compare them.

 

-- Markus


  • JoJu likes this
Editor (Nikon, Leica, Samsung reviews)
photozone.de

#6 planck

planck

    Newbie

  • Members
  • Pip
  • 5 posts

Posted 21 December 2016 - 12:53 PM

Sorry I understand that it will not work. But maybe I wasn't clear enough. I will go step by step so you can say at which point I start to get lost :)

i) when you measure the MTF with the same system that you use to test Nikon 20mm f1.8, the max value you can get is 4000.

ii) for the old setup, the max was 2320. 

iii) the measured MTF with the old setup is 2200. it makes 3793 in the new setup. 

iv) since the new setup is much more sensitive and precise, the new MTF will more likely be below 3793.

v) in the dx sensors, borders are much closer to the center wrt the ff sensors.

vi) MTF decreases gradually when you go from the center to the border.

vii) therefore as a general fact MTF of the borders will be lower when it is measured with ff sensor. 



#7 chrismiller

chrismiller

    Advanced Member

  • Members
  • PipPipPip
  • 205 posts
  • LocationCambridge, UK

Posted 21 December 2016 - 03:43 PM

point iii) has a few assumption that doesn't hold out - its like saying my vw golf with 100hp can do 140mph so a Ferrari with 500hp will be able to do 700mph.

 

Interestingly the pixel density of old tests done on an 10Mpix APC sensor are similar to the density on the 24Mpix full frame sensor (24/1.5/1.5 = 10.6), so if you were feeling brave and ready to ignore lots of factors such as photodiode geometry, sensor glass thickness & and other unknowns and dodge the wrath of the rest of photozone you could maybe do a comparison of center values (ignore borders as the APC edge is a different part of the lens to FF). The 17-35 center number is 2200, so scaling for a large sensor gives 3300lp/ph. So way worst than the 20mm 1.8 at 3900lp/ph, and the center is usually the good bit. Frame corners are always worse.

 

I think the adage of marry your lenses and date the bodies rings true here. If the Tamron is resolving 3300lp/ph over the whole of the D810 sensor (unlikely) you are only getting 14.5Mpix of information, hardly worth having a 36MPix body I would say.



#8 mst

mst

    Advanced Member

  • Moderators
  • 1,942 posts
  • LocationWesterwald, Germany

Posted 29 December 2016 - 02:44 PM

FWIW, I just published the FX review of the Tamron lens. So there is a more solid foundation for comparisons now :)

-- Markus
Editor (Nikon, Leica, Samsung reviews)
photozone.de

#9 planck

planck

    Newbie

  • Members
  • Pip
  • 5 posts

Posted 29 December 2016 - 05:47 PM

Thanks a lot Markus, I guess I triggered this review!  :) The results are as I was expecting. Clearly I need to replace the Tamron. But it seems I do not have many options. I take landscapes and architectural photos mostly with ND filters (this is some sort of the best of 2016). So I need a lens under 20mm with filter thread. Nikkor AF-S 16-35mm f/4 is not a good optic, Zeisses and Nikkor AF-S 17-35mm f/2.8D are way too expensive. So I end up with 20mm f/1.8. Would you have any other recommendation?



#10 Brightcolours

Brightcolours

    Advanced Member

  • Members
  • PipPipPip
  • 4,562 posts
  • LocationThe Netherlands

Posted 29 December 2016 - 05:56 PM

Thanks a lot Markus, I guess I triggered this review!  :) The results are as I was expecting. Clearly I need to replace the Tamron. But it seems I do not have many options. I take landscapes and architectural photos mostly with ND filters (this is some sort of the best of 2016). So I need a lens under 20mm with filter thread. Nikkor AF-S 16-35mm f/4 is not a good optic, Zeisses and Nikkor AF-S 17-35mm f/2.8D are way too expensive. So I end up with 20mm f/1.8. Would you have any other recommendation?

If you stop down the lens to f8 in your usage, the small manual focus Voigtlander 20mm f3.5 SL II does is pretty good job for landscapes and buildings. Downsides: focussing close by astigmatism ruins the corners, and almost no software offers correction profiles for it (at least in relation to Canon FF, may be different for Nikon FF?).

 

A popular lens for Nikon is the cheaper 18-35mm, it has much better corners than the 16-35mm f2 VR. 

http://www.photozone...rafs18353545ff 



#11 Rover

Rover

    Advanced Member

  • Members
  • PipPipPip
  • 1,430 posts
  • LocationRussia

Posted 29 December 2016 - 10:24 PM

The old 18-35 is junk, while the new (AF-S) one is pretty good. Be sure to get the new one if you decide to walk that path.






0 user(s) are reading this topic

0 members, 0 guests, 0 anonymous users



© by photozone.de