Forgot Password
Pentax Camera Forums Home
 

Reply
Show Printable Version Search this Thread
01-08-2016, 05:06 AM   #166
Pentaxian
Fogel70's Avatar

Join Date: Sep 2009
Location: Stockholm, Sweden
Photos: Albums
Posts: 3,062
QuoteOriginally posted by Glorfindelrb Quote
The sharpness map is useful mainly for zoom or to find out from which aperture you expect a lens to perform well. If you need more detailed info you have the acutance graph.

70 Ltd on a K3, it's not all green (Sharpness >> P-MPix map): Pentax smc DA 70mm F2.4 Limited mounted on Pentax K-3 : Measurements
Yes, but that only works well for camera/lens combination that give max 12MP, It works for smaller sensor cameras unless the lens is very good,
FI Sigma 18-35/1.8 give 16MP on Canon 760D, on almost all focal lengths and on all apertures up to f/8 (where noise and diffraction probably start to limit resolution) gives 12+ MP.
Compare the Sigma 18-35mm F1.8 DC HSM A Canon on Canon EOS 760D

The acutance graph do not help much more as it's limited the same way.

On FF which can give 20+ MP (some above 30MP), not even on smallest aperture will the resolution be below 12MP, so all graphs will show all green.
Dxomark tests do not give much info on how focal length or apertures perform on best performing cameras and lenses.

01-08-2016, 03:36 PM   #167
Veteran Member
falconeye's Avatar

Join Date: Jan 2008
Location: Munich, Alps, Germany
Photos: Gallery
Posts: 6,871
QuoteOriginally posted by Fogel70 Quote
The Nikon overall sensor score is 93 and the Canon is 86, so there is not a big difference in score.
I agree. OTOH, the total lens score difference of 38 vs. 39 is, at least equally small.

I agree that the individual test graphs are pretty bad too. Mostly because they include a weighting with the accutance of the human eye. Which is ridiculous in a world where people crop images or print large and watch close.

Fortunately, for many lenses (unfortunately not all) DPR republishes DxO tests in a much more meanungful manner. I normally refer to the republished results then it comes to DxO lens tests.
01-09-2016, 04:25 AM   #168
Veteran Member
zoolander's Avatar

Join Date: Dec 2012
Location: Gold Coast
Photos: Albums
Posts: 351
QuoteOriginally posted by Nicolas06 Quote
Of course there Nikon bias until recently Nikon had the cameras with the most MP and contrary to many reviewers that keep reviewing with old camera of 16 or 24MP for a few years, DxO review their lenses on all cameras so you can check what the lens does on your own camera instead of some other camera the reviewer happen to have.

For sure if you use the best camera from each brand, the brand that gona win is the one with the more MP in term of resolution. That's quite obvious.
Okay, that sounds all well and good, but Nikon isn't the brand with the highest pixel count ........at first. Sony make the sensors and put them in their own cameras first. The A65 with 24mp came out in 2011, and the Nikon D3200 24mp six months later. Same with the Sony Arii with 42mp. But now Canon have it with the 5Dsr.

Regarding lens reviewers. The lens reviewers I look at are pretty much reviewing a lens for its absolute resolution, and not based on : "How sharp it is on 16mp or 24mp or 32mp".. I'm talking ABSOLUTE sharpness. You're talking about photozone and DXO who test at the variety of mega pixels, which isn't really testing a lens for its absolute sharpness.

I don't need to know anymore about how sharp a lens is on 16mp or 24mp, that just got confusing. You go researching 100's 100's of lenses and on this body and that body ..... No ! I want absolute resolution ........ How sharp is the lens regardless of camera mega pixels ? Thats where its at ! And that ain't where DXO or Photozone are, they are just muddying the waters with hyperbole and conjecture and "Virtual pixels" .............

There is no such thing as "virtual pixels" ! DXO are clowns, and they're making a fool of people.

You need straight up scientific analysis, and not conjecture.

Whether you use a 16mp or 24mp camera, a good lens tester can extrapolate how sharp the lens is.
01-09-2016, 05:55 AM   #169
Veteran Member
zoolander's Avatar

Join Date: Dec 2012
Location: Gold Coast
Photos: Albums
Posts: 351
QuoteOriginally posted by Fogel70 Quote
As the noise level on the camera is part of the score, then there is definitely something more than lens measurements presented that affect the score.

Which may explain why FI Nikon 85/1.8G on D750 score better than Canon 85/1.2 on 5DS R. Even though Nikon measure 21MP with best aperture at f/1.8 and Canon measure 30MP with best aperture at F/2.0.


QuoteOriginally posted by Fogel70 Quote
As the noise level on the camera is part of the score, then there is definitely something more than lens measurements presented that affect the score.



Which may explain why FI Nikon 85/1.8G on D750 score better than Canon 85/1.2 on 5DS R. Even though Nikon measure 21MP with best aperture at f/1.8 and Canon measure 30MP with best aperture at F/2.0.
Attached Images

Which is the better lens? At Lenstip:


Nikon Nikkor AF-S 85 mm f/1.8G review - Image resolution - LensTip.com


Canon EF 85 mm f/1.8 USM review - Image resolution - LensTip.com

Canon EF 85 mm f/1.2L II USM review - Image resolution - LensTip.com

Which is the better lens? At Ephotozine:

https://www.ephotozine.com/article/nikon-af-s-nikkor-85mm-f-1-8g-lens-review-18810

https://www.ephotozine.com/article/canon-ef-85mm-f-1-8-usm-interchangeable-lens-review-16998

https://www.ephotozine.com/article/canon-ef-85mm-f-1-2l-ii-usm-lens-review-16624

I would argue that the 85mm 1.2L mark ii is better, especially at 1.8.

But when you look at how well Pentax lenses perform, you might think again:

Pentax DA* 55mm 1.4

Pentax smc DA* 55 mm f/1.4 SDM review - Image resolution - LensTip.com

Pentax DA 50mm 1.8:

Pentax smc DA 50 mm f/1.8 review - Image resolution - LensTip.com

The canon was tested with 21mp, the Nikon 24mp at Lenstip. The Pentax 55mm on 14mp and DA 50mm on 16mp. But the bottom line here is that this is absolute resolution, everything is on an even playing field, they've broken things down to see how things really compare.

A 5Dr versus a D750, they're two totally different beasts, and what matters is that you've got a sharp lens, and bodies are usually marginally different in their ISO performance. DXO's widgets are really quite stupid and do not take into account whether a lens is recommended for a set mega pixel density like the 5Dsr.

If I listened to DXO, I'd buy everything Nikon.

01-09-2016, 06:36 AM   #170
Pentaxian
D1N0's Avatar

Join Date: May 2012
Location: ---
Photos: Gallery
Posts: 6,802
I like to watch picture samples to see if I like a lens, not statistics
01-09-2016, 07:16 AM   #171
Pentaxian
Fogel70's Avatar

Join Date: Sep 2009
Location: Stockholm, Sweden
Photos: Albums
Posts: 3,062
On FF Nikon 85/1.8G and Canon 85/1.2 II seems to perform about the same in center in this test, and at the edge the Nikon performs better.

QuoteQuote:
Which is the better lens? At Ephotozine:

https://www.ephotozine.com/article/nikon-af-s-nikkor-85mm-f-1-8g-lens-review-18810On FF Nikon 85/1.8G anf Canon 85/1.2 II show similar result in center, with Nikon the better at the edge.

https://www.ephotozine.com/article/canon-ef-85mm-f-1-8-usm-interchangeable-lens-review-16998

https://www.ephotozine.com/article/canon-ef-85mm-f-1-2l-ii-usm-lens-review-16624

I would argue that the 85mm 1.2L mark ii is better, especially at 1.8.
It's difficult to say which is best as I can't find any numerical values on the graphs. And as Nikon lens was tested on a 12MP camera compared with the Canon lens tested on a 21MP camera.

QuoteQuote:
But when you look at how well Pentax lenses perform, you might think again:

Pentax DA* 55mm 1.4

Pentax smc DA* 55 mm f/1.4 SDM review - Image resolution - LensTip.com

Pentax DA 50mm 1.8:

Pentax smc DA 50 mm f/1.8 review - Image resolution - LensTip.com

The canon was tested with 21mp, the Nikon 24mp at Lenstip. The Pentax 55mm on 14mp and DA 50mm on 16mp. But the bottom line here is that this is absolute resolution, everything is on an even playing field, they've broken things down to see how things really compare.
The Canon and Nikon test was on FF lenses and Pentax was APS-C lenses. APS-C lenses need 1.5x on absolute resolution as FF lenses on FF camera for same resolution in image. These APS-C cameras also has smaller pixels than the FF cameras, so they would score higher lp/mm with the same lens.

QuoteQuote:
A 5Dr versus a D750, they're two totally different beasts, and what matters is that you've got a sharp lens, and bodies are usually marginally different in their ISO performance. DXO's widgets are really quite stupid and do not take into account whether a lens is recommended for a set mega pixel density like the 5Dsr.

If I listened to DXO, I'd buy everything Nikon.
Do you only buy anything if it is the absolute best at everything?

Last edited by Fogel70; 01-09-2016 at 07:32 AM.
01-09-2016, 08:05 AM   #172
Pentaxian
normhead's Avatar

Join Date: Jun 2007
Location: Near Algonquin Park
Photos: Gallery | Albums
Posts: 40,442
I (or someone) once asked Klaus over at photonzone about abolsute testing of all lenses for absolute numbers not body dependant, I believe his answer was that for $150,000 he could buy a machine that would enable him to do that. My guess is, so far, no one has the $150,000 needed to out into this.

Far more of interest to me is a sensor's conversion factor

For example a K-3 is 6000x4000 pixels, but it ongrodices 2700/lwph according to IR> Physically the sensor os capable of 6000, 4000 distinct lines, but practically it's about half that. Yet I'm sure I've seen review of some Canon sensors that achieved a much higher ratio of measure resolution against theoretical. To me, that is as important as the number of pixels. Not the physical theoretical MP, the functional MP, so I'm assuming from he discussion that's what DxO is putting out there.

If a K-3 can only capture 2700 distance lines then it's functional resolution is a little over 7 MP. That's not non-sense that's measurable fact. Even though the K-3 is listed as 24 MP, and 100% efficient 7MP camera could capture the same image. Cameras like the K-3 achieve their images through serious oversampling.

01-09-2016, 08:27 AM   #173
Pentaxian




Join Date: Apr 2007
Location: Romania
Posts: 15,132
Those are some interesting claims, but be careful.
I wasn't aware of IR (Imaging Resource?) using lwph (like Photozone does). Maybe I missed that, could you give us a link? And I'd also like to see the 24MP APS-C camera resolving close(r) to 6000 lwph. Are they even supposed to?

I don't remember IR complaining about its resolving power; on the contrary, they say stuff like "Very high resolution ; excellent per-pixel sharpness (with AA Filter Simulation set to off)".
And the K-3 is definitely capable of reaching nice per-pixel sharpness.
01-09-2016, 08:40 AM   #174
Pentaxian
normhead's Avatar

Join Date: Jun 2007
Location: Near Algonquin Park
Photos: Gallery | Albums
Posts: 40,442
Click on the camera review and go to the Exposure tab



Scroll down the screen until you get to this part.


The nice thing about them providing the data they used to make the determination, is, I can tell, personally I would probably rate the K-3 more like 3000 lw/ph based on the same data. But, as long as the same guy does all the rating and it's consistent I guess it's all meaningful.

QuoteQuote:
And I'd also like to see the 24MP APS-C camera resolving close(r) to 6000 lwph. Are they even supposed to?
Think about the sensor array. It's 6000 pixels, if there was a 1 pixel wide line all black and then another that's one pixel wide line that was all white, and the whole sensor was like that, there'd be 4000 distance lines vertically and 6000 horizontally.

Everything would have to line up perfectly for that to happen, lens characteristics , a test pattern absolutely matched to the sensor lens combination. But that would be the maximum number of lines you could theoretically produce with a K-3. Obviously the functional is going to be quite a bit less than that. But, some companies seem to be better than other at squeezing out resolution, although it does seem to be pretty constant at quick glance, I can't remember which Canon model impressed me with it's low pixel count and high resolution, and IR takes down old camera reviews, so if it's not in the stores, it's not on their site.

Last edited by normhead; 01-09-2016 at 08:54 AM.
01-09-2016, 08:41 AM   #175
Veteran Member




Join Date: Nov 2013
Posts: 4,854
QuoteOriginally posted by zoolander Quote
Okay, that sounds all well and good, but Nikon isn't the brand with the highest pixel count ........at first. Sony make the sensors and put them in their own cameras first. The A65 with 24mp came out in 2011, and the Nikon D3200 24mp six months later. Same with the Sony Arii with 42mp. But now Canon have it with the 5Dsr.

Regarding lens reviewers. The lens reviewers I look at are pretty much reviewing a lens for its absolute resolution, and not based on : "How sharp it is on 16mp or 24mp or 32mp".. I'm talking ABSOLUTE sharpness. You're talking about photozone and DXO who test at the variety of mega pixels, which isn't really testing a lens for its absolute sharpness.

I don't need to know anymore about how sharp a lens is on 16mp or 24mp, that just got confusing. You go researching 100's 100's of lenses and on this body and that body ..... No ! I want absolute resolution ........ How sharp is the lens regardless of camera mega pixels ? Thats where its at ! And that ain't where DXO or Photozone are, they are just muddying the waters with hyperbole and conjecture and "Virtual pixels" .............

There is no such thing as "virtual pixels" ! DXO are clowns, and they're making a fool of people.

You need straight up scientific analysis, and not conjecture.

Whether you use a 16mp or 24mp camera, a good lens tester can extrapolate how sharp the lens is.
99.9% of reviewers review the lens on a given camera by taking picture of some test chart and then using imatest. This procedure is dependant of many factors:
- the number of MP of the sensor of course
- the low pass filter, how strong/weak it is or if there one or not.
- the processing done on the JPEG or RAW by the manufacturer regardless of the theoretical meaning of "RAW" manufacturers apply sharpening, lens corrections and so on (Sony for example).

This is the case of dpreview, photozone, lenstip, slrgear all test lenses on a given camera or set of camera. The only one I know of that use an optical test bench instead independant of any camera is lensrental. But this is the exception.

The issue is that even if the review doesn't say it, it tested the lens on a camera and so the lens was limited by the camera used. The typical reviewer also adapt its rating based on the camera. if the camera can manage 10Mp, then 2350 lw/ph is outstanding and anything that is 2000+ is good, but if the lens is tested on a 24MP without low pass filter it would need to get 3000 lw/ph to get the results. In practice not all lenses handle the increase of Mp as well and some lenses that where reviewed as great on old sensor get a reduced rating on a more modern one. This is the case of FA43 for example.

And this is also forgeting that most reviewer test only 1 sample. The 18-135 on photozone was likely a bad one for example. Sample variation in term of performance can be quite big, in particular toward border/corner. On the review website lens A might be better than lens B but if yourself buy them both you could have you own sample of B being better than your own sample of A. The example is the 20-40. Some review find it better near 20mm other near 40mm other in the middle.

What more? An innexpensive kit lens on a high rez sensor typically does better in absolute than on a lower rez sensor... But will get lower scores by the reviewer because it might not be as near to the top performance the sensor allow. Tha's stupid but that how lenses are reviewed.

You initial point was that DxO was biased for Nikon. Do you know that nowadays the lenses that are mounted on Canon that get the best sharpness score? Well that because now 5Ds is part of the test results. The 35mm f/2 that before had 20MP sharpness score on a 5D mark III just got 34MP on the 5Ds.

An you know what? All theses FF lenses pentax has will get improved score when they will be reviewed with the FF and suddenly people look a bit too fast would think Pentax make much better lens than before.

The other reviewers they still use a given camera to perform their review. What more they don't test on all cameras but on a single one they have arround that is not necessarily the latest and great. Most reviewers use camera that are 24MP or less. This is even more visible if the review is a bit old and then the cameras are 10 or 16MP...

Last edited by Nicolas06; 01-09-2016 at 09:14 AM.
01-09-2016, 08:51 AM   #176
Veteran Member




Join Date: Nov 2013
Posts: 4,854
QuoteOriginally posted by Kunzite Quote
Those are some interesting claims, but be careful.
I wasn't aware of IR (Imaging Resource?) using lwph (like Photozone does). Maybe I missed that, could you give us a link? And I'd also like to see the 24MP APS-C camera resolving close(r) to 6000 lwph. Are they even supposed to?

I don't remember IR complaining about its resolving power; on the contrary, they say stuff like "Very high resolution ; excellent per-pixel sharpness (with AA Filter Simulation set to off)".
And the K-3 is definitely capable of reaching nice per-pixel sharpness.
photozone lw/ph and IR lw/ph are not the same at all. Imaging resource find 1900 lw/ph for K5 JPEG and 2100 from converted raws while photozone find the max lw/ph for a K5 is 2750.

On photozone the best a 24MP APSC sensor get is arround 3700-3800. Basically the max on photozone is 35-40% higher than on imaging resource. One should keep that in mind when comparing results.
01-09-2016, 09:01 AM   #177
Veteran Member




Join Date: Nov 2013
Posts: 4,854
QuoteOriginally posted by normhead Quote
I (or someone) once asked Klaus over at photonzone about abolsute testing of all lenses for absolute numbers not body dependant, I believe his answer was that for $150,000 he could buy a machine that would enable him to do that. My guess is, so far, no one has the $150,000 needed to out into this.

Far more of interest to me is a sensor's conversion factor

For example a K-3 is 6000x4000 pixels, but it ongrodices 2700/lwph according to IR> Physically the sensor os capable of 6000, 4000 distinct lines, but practically it's about half that. Yet I'm sure I've seen review of some Canon sensors that achieved a much higher ratio of measure resolution against theoretical. To me, that is as important as the number of pixels. Not the physical theoretical MP, the functional MP, so I'm assuming from he discussion that's what DxO is putting out there.

If a K-3 can only capture 2700 distance lines then it's functional resolution is a little over 7 MP. That's not non-sense that's measurable fact. Even though the K-3 is listed as 24 MP, and 100% efficient 7MP camera could capture the same image. Cameras like the K-3 achieve their images through serious oversampling.
Signal theory say that if you sample frequency is X, you can't expect to record accurate signal of higher frequency than X/2. This has been proved mathematically. So a K3 with 6000x4000 (rougly) can't record more than 3000 line x 2000 lines accurately.

Additionnaly you have to factor that there is no 24MPixels but 24Mphotosites. The colors are interpolated this is actually what a raw processing software does: interpolating color information. Basically if the lines where red and not B&W, only the red photosites could see it with a red channel real resolution of 3000x2000. The same theorem as before would say you can't record correctly more than 1500x1000 red lines then.

All the improvement you get are from algorithms that make the image look better through lot of complex processing, And that also why pixel shift improve quality so much: information no longer interpolated, all color information is now there, at long at the scene kept perfectly still.

What photozone or maging resource are doing is getting an approximation using imagtest a procedure that is influenced by many factors. if for example you push sharpness setting in camera for Jpegs or in the raw converter, then you'll get noticably better lw/ph results.

When reviewers also speak of center, border and corner sharpness they don't take the measurement at the same place, may not measure the 4 corners or the border in many places and the result may be biased by sharpening algorithm applied directly on the raw by the manufacturer. If you could get more sales by getting better rating just because you adjusted you sharpness result setting differently in camera, wouldn't you want to benefit of it ?

Even if you start from the raw and use the same raw processing software, you have to understand that theses tool are optimized for each camera. It could be that you measure that the default sharpening setting for Camera A has been pushed a bit more than on Camera B, maybe because the raw processing software testers find out that the raw were a bit too soft by default on that camera. That's easy as that.

Last edited by Nicolas06; 01-09-2016 at 09:09 AM.
01-09-2016, 09:09 AM   #178
Pentaxian




Join Date: Apr 2007
Location: Romania
Posts: 15,132
The problem is with the claim that K-3 only resolves about 7MP and it's basically oversampling. Extraordinary claims such as this must be double checked and supported by facts, it doesn't suffice to simply call them "facts".

OK, I missed indeed the 2700 lwph. But that's only slightly below the D7200's 2900 lwph and A77 II's 2850 lwph. Norm, perhaps that's actually a normal value?

Nicolas, IIRC Photozone warns about the variability in lwph measurements - it depends inclusively on the sharpening applied. Same must be true also between a IR-Photozone comparison.
01-09-2016, 09:20 AM   #179
Veteran Member




Join Date: Nov 2013
Posts: 4,854
QuoteOriginally posted by Kunzite Quote
The problem is with the claim that K-3 only resolves about 7MP and it's basically oversampling. Extraordinary claims such as this must be double checked and supported by facts, it doesn't suffice to simply call them "facts".

OK, I missed indeed the 2700 lwph. But that's only slightly below the D7200's 2900 lwph and A77 II's 2850 lwph. Norm, perhaps that's actually a normal value?

Nicolas, IIRC Photozone warns about the variability in lwph measurements - it depends inclusively on the sharpening applied. Same must be true also between a IR-Photozone comparison.
There are fact. a K3 use a bayer filter so you don't get full 24MP. So like all 24MP sensors with a bayer filter you don't really get 24MP but 24 millions photosites. You can interpret this as a perfect 6MP camera or a 24MP interpolated 24MP camera. All manufacturers choosed the second solution.

The lw/lh and interpretation that the K3 behave like a 7MP camera due to having only 2700 lw/ph is wrong. The theorem doesn't change, to accurately measure 2700 lines, you need the double in number of sample so you need at least 5400x3600 that is 19.4MP, not 7MP.

And let's not forget that between what photozone tend to get as measurement and imagining resouce we have 35-40% difference in resolution. In term of MP that the double MP, so it is quite difficult to get any precise conclusion from that.

Last edited by Nicolas06; 01-09-2016 at 09:28 AM.
01-09-2016, 09:31 AM   #180
Pentaxian
normhead's Avatar

Join Date: Jun 2007
Location: Near Algonquin Park
Photos: Gallery | Albums
Posts: 40,442
QuoteOriginally posted by Kunzite Quote
The problem is with the claim that K-3 only resolves about 7MP and it's basically oversampling. Extraordinary claims such as this must be double checked and supported by facts, it doesn't suffice to simply call them "facts".

OK, I missed indeed the 2700 lwph. But that's only slightly below the D7200's 2900 lwph and A77 II's 2850 lwph. Norm, perhaps that's actually a normal value?

Nicolas, IIRC Photozone warns about the variability in lwph measurements - it depends inclusively on the sharpening applied. Same must be true also between a IR-Photozone comparison.
Oh, it definitely is a normal value, and the variation between cameras is actually quite small. But from my perspective, lw/ph is probably a better evaluation of a cameras resolution than MP. Although I'm sure I once saw camera on IR that had an unexpectedly high lw/ph given it's number of MP, so it's definitely not a direct correlation.

And I'm wondering if that's how DxO comes up with their MP number.
Reply

Bookmarks
  • Submit Thread to Facebook Facebook
  • Submit Thread to Twitter Twitter
  • Submit Thread to Digg Digg
Tags - Make this thread easier to find by adding keywords to it!
645z, apples, base, capacity, claims, d810, design, dr, dxo, dxomark, files, ii, iso, k-3, k-5, k-7, lossless, noise, pentax, pentax 645z scores, pentax news, pentax rumors, pixels, score, scores, sensor, technology, unit
Thread Tools Search this Thread
Search this Thread:

Advanced Search


Similar Threads
Thread Thread Starter Forum Replies Last Post
Pentax 645z in DXOMark 101 points..., is it true? Ventzy Pentax Medium Format 32 01-01-2016 12:08 PM
K-S1 scores on DxOMark 6BQ5 Pentax K-S1 & K-S2 34 11-14-2014 07:52 AM
Red Epic Dragon jumps to top of DxOMark Sensor charts with score of 101 ! jogiba Canon, Nikon, Sony, and Other Camera Brands 14 03-10-2014 01:50 PM
Arrggh - Nikon V2 scores lower than V1 or RX100 on dxomark Christine Tham Canon, Nikon, Sony, and Other Camera Brands 36 12-24-2012 04:41 PM
DXOMark scores for the Q posted. . . snostorm Pentax Q 8 12-03-2011 01:47 PM



All times are GMT -7. The time now is 07:05 PM. | See also: NikonForums.com, CanonForums.com part of our network of photo forums!
  • Red (Default)
  • Green
  • Gray
  • Dark
  • Dark Yellow
  • Dark Blue
  • Old Red
  • Old Green
  • Old Gray
  • Dial-Up Style
Hello! It's great to see you back on the forum! Have you considered joining the community?
register
Creating a FREE ACCOUNT takes under a minute, removes ads, and lets you post! [Dismiss]
Top