A very impressive lens.
No 35 mm lenses I know of are sharpest at F2.8.
Which 50mm on the DX3? Are we reading the same numbers? The 200mm peaks at 4076 center, while the best 50mm peaks at 4008. And what about border/corner resolution? Or did you look at only DX?
Since we are really past the limit of the measuring system (sensor), the numbers may not be precise for either of these lenses. Small differences are not conclusive, particularly given the issues with testing long vs. normal lenses. BTW, what do you think is PZ's margin of error? Copy variation?
I guess we have to agree on what "sharpest" means vs. aperture and the image field. Certainly I think you can agree that super-teles are among the sharpest lenses from any manufacturer. To say if they are sharpest or not needs better data.
In whose mind? If they are among the sharpest lenses (over the entire field and wider f-stops), then who is highly exaggerating?
It also hard and expensive to make a super sharp 50mm. That's why the Otus is $4k.
I think what gives me the impression are the MTF charts.
PZ measures the system response of hardware + software (it demosaics and sharpens images), so I am not sure how relevant their results are when it comes to the sharpness of a lens.
Most of the cameras used for testing also have AA filters.
This might be a significant concern if
Which ends us being the same restriction as PZ.
Yes, and differing pixel pitches, hence the suggestion to compare lenses tested on the same camera. The key imho is to try to assess the physical characteristics of the hardware separately from the processing - since the processing can produce arbitrary results.
They do to my way of thinking. Plus PZ demosaics the image data, effectively performing an upsizing operation on the raw information. It's a bit like trying to assess the quality of the meat used in a Big Mac by eating the whole, dressed up burger drenched in pickles and sauces to (their) taste. It's not a coincidence that they sometimes get impossible values (e.g. lw/ph > actual number of rows on sensor).
Not quite, Lenstip attempts to isolate the hardware. PZ = Big Mac. Lenstip = mostly the beef.
Another thing is that these tests are done at relatively short shooting distance and do not reliably describe infinity performance. This heavily penalizes lenses that are great at infinity, but not great at MFD (case in point: the current Zeiss Planar 50/1.4). It also penalizes some wide angle lenses that have field curvature at short distances.
Compared to the other sources of variation, this is small potatoes. PZ data corresponds to real world experience as well (or as poorly) as the other sites. MTF50 itself is only a simple proxy for what most people look for in a lens.
They still use dcraw to demosaic -- IIRC, the non-demosaic options for Imatest were not available when LT started testing. They would have to invalidate or recalculate the prior results to switch methods.
Thanks for the link. I found Roger's articles are always worth reading.
Thanks, I was always wondering about that.
Ok, I can tell you are smart guy who understands the compromises involved, so if MacDonald's is your thing go right ahead.
Many thanx for that info
I'm not asking any such thing. For numeric result lens tests of super-teles, we have very limited data. That data neither busts or confirms the myth; it remains plausible.
If you have a better source of data for the lenses in question, by all means bring it to our attention.
Our? Yours you mean, since anybody who has been paying attention already knows what my criteria for better lens testing sites are: more beef, less sauce and pickles. Or put another way: stick with sites that try to minimize processing and stay away from those whose MTF50 readings exceed the laws of physics.
And those sites have actually tested some super-tele lenses? That's why I asked if you know of some relevant data. Of course relevance is not important when you only want to rant on your pet peeve.
Now now Erik, no need to get nasty and shoot the messenger. Physics is physics and the D3x's sensor only has 4044 rows.
With the criteria mentioned anyone could look up the data themselves. But you are making me work, so here for instance is the Canon EF200 f/2L which by any account is comparable to the Nikon you were referring to: MTF50 of 45 lp/mm on the 21.5MP 1DsMkIII in the center, or 2160 lw/ph. Similar results on the EF300 f/2.8L
