Thread Rating:
  • 0 Vote(s) - 0 Average
  • 1
  • 2
  • 3
  • 4
  • 5
next PZ lens test report: Pentax SMC DA 35mm f/2.4 AL
#11
[quote name='flektogon' timestamp='1300398989' post='6915']

What are talking about? Those bars are nothing else than graphical representation of the resolution numbers! The only way to tell, which lens has higher resolution then the others is to test/measure their resulution under the exactly same conditions.

[/quote]

The bar's height have a direct relation to the maximum achievable resolving power of the camera used. You do not have to take my word for it, but you really can get a really good impression of a lens' sharpness just by comparing the bar height in relation to the vertical axis, without looking at the measured numbers.



Two examples:

Pentax K10 with Tamron 90mm f2.8 DimMacro:

[Image: mtf.gif]

Canon EOS 350D with same Tamron 90mm f2.8 Di macro:

[Image: mtf.gif]



Canon EOS 350D with Canon EF 100mm f2.8 USM macro:

[Image: mtf.gif]

Canon EOS 50D with Canon EF 100mm f2.8 USM macro:

[Image: mtf.png]



As you can see, the bars show a very similar height both for the 10mp Pentax K10 and the 8mp Canon EOS 350D, both with a Tamron 90mm.

And same with the bars for the 8mp 350D and 15mp EOS 50D, both with a Canon 100mm macro.



The vertical axis is important, it shows the maximum achievable resolution for the sensor. Since for each camera the scale on the vertical bar is different, you can compare results from the K10 tests and the K5 test. Only when a lens starts to resolve about half the maximum achievable reults, you will start to see real differences between sensors with different resolutions.
#12
[quote name='Brightcolours' timestamp='1300400414' post='6916']

The bar's height have a direct relation to the maximum achievable resolving power of the camera used. You do not have to take my word for it, but you really can get a really good impression of a lens' sharpness just by comparing the bar height in relation to the vertical axis, without looking at the measured numbers.



Two examples:

Pentax K10 with Tamron 90mm f2.8 DimMacro:

[Image: mtf.gif]

Canon EOS 350D with same Tamron 90mm f2.8 Di macro:

[Image: mtf.gif]



Canon EOS 350D with Canon EF 100mm f2.8 USM macro:

[Image: mtf.gif]

Canon EOS 50D with Canon EF 100mm f2.8 USM macro:

[Image: mtf.png]



As you can see, the bars show a very similar height both for the 10mp Pentax K10 and the 8mp Canon EOS 350D, both with a Tamron 90mm.

And same with the bars for the 8mp 350D and 15mp EOS 50D, both with a Canon 100mm macro.



The vertical axis is important, it shows the maximum achievable resolution for the sensor. Since for each camera the scale on the vertical bar is different, you can compare results from the K10 tests and the K5 test. Only when a lens starts to resolve about half the maximum achievable reults, you will start to see real differences between sensors with different resolutions.

[/quote]



But this is exactly what I was pointing at. The same lens (if we assume that there were no lens sample variations) mounted on the 10 Mp Pentax camera achieved 2180 pixels (per the image hight) resolution, while mounted on the 8 Mp Canon camera it achieved only 1999 pixels. For sure, poor lenses will show more less the same resolution regardless of the sensor resolution. But a high quality lens will be much more limited by the sensor resolution. And while the sensor requires use of the AA filter (which is still the case even of contemporary 18 Mp sensors), it clearly means that their spacial resolution is still below the Nyquest frequency. Yes, once the future sensors have maybe more than 20 Mp, then their resolving power will not limit even the best lens. How much pixels such sensors would need? I remember, that best lenses from the film era could not resolve more than 100 lp/mm, so let's use that number, though today's lenses, optimized for the APS-C size sensors could have (much) higher resolution. To satisfy the sampling theory, the sensor without a need for the AA filter should have 2 x 100 x 16 (the APS-C sensor hight) pixels in the vertical direction, which is more than 15 Mp for the entire sensor. So, the same lens tested on a camera with a 10 Mp and 16 Mp sensor will show very different resolution.
#13
[quote name='flektogon' timestamp='1300408035' post='6918']

But this is exactly what I was pointing at. The same lens (if we assume that there were no lens sample variations) mounted on the 10 Mp Pentax camera achieved 2180 pixels (per the image hight) resolution, while mounted on the 8 Mp Canon camera it achieved only 1999 pixels. For sure, poor lenses will show more less the same resolution regardless of the sensor resolution. But a high quality lens will be much more limited by the sensor resolution. And while the sensor requires use of the AA filter (which is still the case even of contemporary 18 Mp sensors), it clearly means that their spacial resolution is still below the Nyquest frequency. Yes, once the future sensors have maybe more than 20 Mp, then their resolving power will not limit even the best lens. How much pixels such sensors would need? I remember, that best lenses from the film era could not resolve more than 100 lp/mm, so let's use that number, though today's lenses, optimized for the APS-C size sensors could have (much) higher resolution. To satisfy the sampling theory, the sensor without a need for the AA filter should have 2 x 100 x 16 (the APS-C sensor hight) pixels in the vertical direction, which is more than 15 Mp for the entire sensor. So, the same lens tested on a camera with a 10 Mp and 16 Mp sensor will show very different resolution.

[/quote]

no, you keep focussing on the NUMBERS. Don't do that.... just look at the bars themsleves, in relation to the vertical axis. Not the numbers, but the maximum achievable. The horizontal lines in each graph are meaning the same thing, even if the numbers, for the higher res. sensor, are higher.



So.. the top line is the max achievable. The next line is to what is considered excellent. And so on.Erase the numbers, and just compare the graphs,



As you can see, the bars for the k10 and 350D + tamron then are pretty much the same.

Similar, the graphs of the 350D and 50D are pretty much the same, with the same lens.



Only when the measurements go very low, you will start to see a difference.



And with this knowledge, you can more or less compare the new 35mm f2.4 with the very good discontinued 35mm f2 and the nice 35mm f2.8 macro. Just disregard the numbers on the scale, and only look at the graph itself.
#14
Brightcolor suggestion gives a soft hint but it is not definitive. Anyways there is a lot more to a lens than pure resolution and it is always (well almost always) a lot more expensive to produce a quality FF lens (35f2) than a APSC lens (35f2.4).
#15
You2,



The Pentax 35mm f2.4 has been confirmed by a number of people as covering a full frame (135) image. Despite the DA label it is one of many full frame lenses.
#16
[quote name='Brightcolours' timestamp='1300408508' post='6919']

no, you keep focussing on the NUMBERS. Don't do that.... just look at the bars themsleves, in relation to the vertical axis. Not the numbers, but the maximum achievable. The horizontal lines in each graph are meaning the same thing, even if the numbers, for the higher res. sensor, are higher.



So.. the top line is the max achievable. The next line is to what is considered excellent. And so on.Erase the numbers, and just compare the graphs,

[/quote]



The "Rating Scale" for K-5 is missing from the DAL 35/2.4 test, which is causing some confusion.



For example, this is the one for K-20:

[Image: rating_k10d.gif]



But the vertical axis of the new test is still in the 5 divisions that represent [ex/very good/good/fair/poor].
#17
[quote name='dave9t5' timestamp='1300587505' post='6955']

The "Rating Scale" for K-5 is missing from the DAL 35/2.4 test, which is causing some confusion.



For example, this is the one for K-20:

[Image: rating_k10d.gif]



But the vertical axis of the new test is still in the 5 divisions that represent [ex/very good/good/fair/poor].

[/quote]



In addition, what do those 2350 (K10) and 2700 (K-5) maximum resolution numbers represent? They are not Nyquest frequencies for sure. If they were the maximum resolution numbers, then yes, you could say that the identical (from the resolution point of view) lens would resolve for example 95% of the maximum resolution of a given sensor. So, do we know what are those numbers? And how they were obtained? Using some exteremely sharp lens with almost unlimited resolution?



Thanks.
#18
[quote name='flektogon' timestamp='1300627252' post='6962']

In addition, what do those 2350 (K10) and 2700 (K-5) maximum resolution numbers represent? They are not Nyquest frequencies for sure. If they were the maximum resolution numbers, then yes, you could say that the identical (from the resolution point of view) lens would resolve for example 95% of the maximum resolution of a given sensor. So, do we know what are those numbers? And how they were obtained? Using some exteremely sharp lens with almost unlimited resolution?



Thanks.

[/quote]

Photozone has a way of measuring a bit past the nyquist limit. It does not really matter, though, how photozone gets its maximum achievable value.. as long as they do it the same for each camera.



Th point I made remains the same: you can compare lenses from different bodies, just by looking at the graph itself and ignoring the values. This will give a good enough impression to know how a lens performs.
#19
[quote name='Brightcolours' timestamp='1300638149' post='6973']

Photozone has a way of measuring a bit past the nyquist limit. It does not really matter, though, how photozone gets its maximum achievable value.. as long as they do it the same for each camera.



Th point I made remains the same: you can compare lenses from different bodies, just by looking at the graph itself and ignoring the values. This will give a good enough impression to know how a lens performs.

[/quote]



You canot go beyond the Nyquest frequency. To avoid a "danger" that a given lens could deliver higher than the Nyquest frequency resolution (in the case of image sensor this means to deliver more pixels of resolution then is the number of sensor pixels in that direction), there is an AA filter in front of the sensor. And those 2350 and 2700 numbers represent lower resolution than what corresponds to the Nyquest frequency of 10 and 16 Mp sensors respectively. I woud say that they somehow represent the maximum resolution, which a given sensor INCLUDING the AA filter can deliver. I am just curious, how you can measure such resolution. Unless you use a "perfect" lens.
#20
Not again please. This has been discussed quite a bit.



Please google:

Imatest Nyquist binning
  


Forum Jump:


Users browsing this thread:
2 Guest(s)