Film_Ruled wrote:
Right, but that is *your* set Herb, not mine. At this point, regardless of how the Zeiss and 60 2.8 AF-S that I rented perform, I will keep my obviously unique 50 1.4G. I have pulled in tens of thousands of dollars in income from this lens alone on ad campaigns, it is has more than paid for it self.
The optical formula is unique. Your copy can't be that much better than anyone else's. It's just not possible.
Film_Ruled wrote:
Umm, I said wide open my copy is spectacular, why are people *not* getting that, !!!
And what I don't get is your first reply to me when I stated I wasn't sure about the 50 f/1.4G and D800 combination. What I got from your statement was that all copies should possess the same "spectacular" results as yours.
Now you're saying you got extremely lucky with your copy, which you are pretty much admitting that everyone else has copies that produce the mediocre results compared to yours. Either we have duds, or you have a dud. Or there's a different set of expectations in respect to the lens' sharpness at wide open and when stopped down.
Tanegashima wrote:
You can't get glass good enough for the D800
That camera can outresolve any lens!
No lens can make 50 line pairs per millimeter (or at least, center-to-borders), the D800 sensor has more than 200 pixels per millimeter.
(24*50*2*2)*(36*50*2*2)=34.56 Megapixels.
So, you have to insist in stopping down and getting close (real tests and MTF charts are at infinity, lens performance is usualy better at close distances).
IMHO, Nikon did great and hit the wall down once and for all!
I hope you're joking, because that post is...mildly...inaccurate
In my case, have most of the lenses I want, and will get the last couple soon after the D800 as they only make sense once I have a FF (24 1.4, 45 PC-E, maybe 35 1.4).
You don't need a new computer though. If you don't stitch regularly a computer with 6GB of RAM and any processor from the last couple of years will be plenty. If you do stitch and do operations on large files, then a quad or hexcore + 8-16GB will banish any sluggishness remaining. Of course, if Nikon come out with a DLOesque correction process, that hexcore will be very handy. We're computationally in a much better relative position than we were when the 5D2 came out, so all the handwringing about it is amusing.
i suggest more then computing power, to invest in SSD. I have still 32bit photoshop but thanks to SSD i can open and work on 18-50 D800 photos at the same time, barely notice any slowdown with even 5 history levels.
BigIronCruiser wrote:
This is off topic, but how is the SSD used to improve LR performance?
i dont know lightroom, but every windows benefit from having swap file and system self on the SSD. All system related tasks as well memory management is much faster.
Then you can put LR catalog on the SSD and or all temporary files.
I have two SSDs on my workstation and heavily cached RAID
One SSD is slower, where system is installed
Second SSD is fast, but likely not so reliable, where swap file is, Photoshop Scratchdisc is
on RAID drive are files i am actually working on. Speed is amazing even with files from scanned MF backs.
as said i dont know LR, but windows self will benefit a lot from SSD.
then if you have temp and/or working files on SSD as well even more.
Capture One don't need SSD at all, as the work files and sessions are quite small, but it speeds up listing through directories when you have them on SSD.
I use SSD also for working files before they get exportet to archive.
It is little here, little there and in the end a lot of difference.
People have been wringing their hands in anguish over this exact same question ever since digital cameras made the huge leap from 3 to 6 megapixels...
-1
Respectfully, subjective opinion does not negate objective evidence to the contrary. As I wrote previously:
SloPhoto wrote:
Where is the vote for
"None of the above are terribly important"
Because it is important. Given the small photosite size, diffraction becomes apparent at wider apertures than on a 12 or 24 MP sensor so you can't depend on f/8 to "clean things up".
Leaping from 3 to 6 to is not the same as going from 12 to 36. The optical demands are greater and have to be carefully considered. Just plain physics.
Respectfully, subjective opinion does not negate objective evidence to the contrary. As I wrote previously:
Because it is important. Given the small photosite size, diffraction becomes apparent at wider apertures than on a 12 or 24 MP sensor so you can't depend on f/8 to "clean things up".
Leaping from 3 to 6 to is not the same as going from 12 to 36. The optical demands are greater and have to be carefully considered. Just plain physics.
You keep quoting MTF 50 numbers. What about MTF 10. Peak quality from a bayer pattern sensor is achieved when the sensor out resolves the lens at every aperture and even in fine detail. That is the point when the AA filter is truly unnecessary.
Even in the situation where the lens is outresolved by the sensor, I fail to see the issue with that. It just means you are now getting the very best image possible with that lens. It is not like the camera pops up a warning saying "sorry lens does not have enough resolution, image will not be recorded"
SloPhoto wrote:
Even in the situation where the lens is outresolved by the sensor, I fail to see the issue with that. It just means you are now getting the very best image possible with that lens. It is not like the camera pops up a warning saying "sorry lens does not have enough resolution, image will not be recorded"
Keep this in mind:
MTF[system] = MTF[lens] * MTF [sensor]
So improving the resolution of the lens or sensor will improve the resolution of the photo. In other words, if a 24MP sensor out resolved a particular lens, an image produced with that lens and a higher resolution sensor, say 36MP, would still provide more resolution.
Respectfully, subjective opinion does not negate objective evidence to the contrary. As I wrote previously:
Because it is important. Given the small photosite size, diffraction becomes apparent at wider apertures than on a 12 or 24 MP sensor so you can't depend on f/8 to "clean things up".
Leaping from 3 to 6 to is not the same as going from 12 to 36. The optical demands are greater and have to be carefully considered. Just plain physics.
Ask yourself this:
When lenses massively outresolved sensors, did you say "ugh, there's no point buying this lens because sensors can't keep up"?
There's the far more important tidbit: when total system resolution is increasing linearly with a change in a single component, it's because that component is bottlenecking everything else. As said above, you want sensors to outresolve lenses if you're after the best image quality.
Why does everyone seem so concerned with the processing power needed to PP the D800 files? 16GB of RAM costs under $100, and a $200 processor is more than enough.
have you actually run some numbers through this convolution formula? i did when the D3X came out and it was what convinced me that unless there wasn't a lens worth using on it unless it had very close to the resolution of the sensor on the low end and preferably at least 1.5X. beyond that, even going to 10x had not enough difference to matter but that first 1x to 1.5x better lp/mm than the sensor was really important. since the first order approximation to the convolution is symmetric, it also says that beyond a sensor having 1.5x the resolution of the lens, the gain is imperceptible.
Herb...
afm901 wrote:
Keep this in mind:
MTF[system] = MTF[lens] * MTF [sensor]
So improving the resolution of the lens or sensor will improve the resolution of the photo. In other words, if a 24MP sensor out resolved a particular lens, an image produced with that lens and a higher resolution sensor, say 36MP, would still provide more resolution.
it doesn't do that since Scott keeps posting a convolution formula. that * is the convolution operator, not a multiplication. look at some lens design textbooks. the relationship is highly nonlinear.
Herb...
ausemmao wrote:
There's the far more important tidbit: when total system resolution is increasing linearly with a change in a single component
"The beauty of working in frequency domain is that the response of the entire system (or group of components) can be calculated by multiplying the [MTF] responses of each component."
HerbChong wrote:
it doesn't do that since Scott keeps posting a convolution formula. that * is the convolution operator, not a multiplication. look at some lens design textbooks. the relationship is highly nonlinear.
Herb...
Or if you do software, it is the multiplication operator.
!!!