Pages: 1 2 [3] 4 5 ... 9   Go Down

Author Topic: Sony A7R, intersting comaprision ...  (Read 44419 times)

Telecaster

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 3686
Re: Sony A7R, intersting comaprision ...
« Reply #40 on: November 19, 2013, 02:36:34 pm »

I'm currently buying some duclos converted m43 f. 095 lenses.  They're not the sharpest lenses in the world but for motion imagery and some stills they do have a unique look, they glow in highlights and the flare like we expect film to flare when tracking across the scene into the sun.

Those are modded Voigtländers, yes? I have the three-lens (unmodded) set...very nice indeed for a film-like look. They remind me of Pentax Super-Taks but with greater clarity. You'll find 'em to be quite sharp when stopped down past f/2.8 too, assuming you actually plan on stopping down at all.

-Dave-
Logged

ErikKaffehr

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 11311
    • Echophoto
Re: Sony A7R, intersting comaprision ...
« Reply #41 on: November 19, 2013, 04:08:21 pm »

Hi,

Although I agree with Ray mostly I feel that there is an advantage of small pixels. Quite true that we get diminishing returns, but large pixels also produce artifacts. Look at the feather images below. They were shot from a fixed distance with 150 mm lens on a: P45+ (39MP MF), a Sony Alpha 99 SLT (24 MP FF) and a Sony Alpha 77 SLT (24 MP APS-C).

The pixel sizes are 6.9, 6, 3.9 microns. If you look carefully you see that the P45+ image on the left has a lot of fake detail while the Sony Alpha 77 on the right has most detail. This doesn't mean that the Alpha 77 outperforms the other two, but it produces less artifacts.

The other images show crops from comparably scaled images from the same cameras, and here it is obvious that the Alpha 77 SLT is lagging the other two significantly. I was not really expecting such a difference so I did reshoot it, with very similar results.

My point is really: The best solution is a large sensor with reasonable small pixels.

Best regards
Erik

It's not the greater number of pixels per se - it's the greater quantity of information. You can subdivide a fixed sensor size into more and more, smaller and smaller pixels, but at some point when the image is oversampled, there is no further gain in information. Maintaining the pixel size and lens performance while increasing the sensor area is still the best route to increased information (detail).

Ray
Logged
Erik Kaffehr
 

BJL

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 6600
Smaller pixels, bigger sensors: where the diminishing returns limit gains
« Reply #42 on: November 19, 2013, 05:02:58 pm »

It's not the greater number of pixels per se - it's the greater quantity of information. You can subdivide a fixed sensor size into more and more, smaller and smaller pixels, but at some point when the image is oversampled, there is no further gain in information. Maintaining the pixel size and lens performance while increasing the sensor area is still the best route to increased information (detail).y
Downsizing of pixels and upsizing of sensors have something in common: if done right, each can produce measurable improvements in image quality, but taken too far, each hits the law of dimishing returns, and only spec. fetishists really benefit from going further.

With pixel downsizing, one limit is the increase in read noise from having more smaller pixels on the same-sized sensor; another is lens resolution. Note however that some photographers will have reason to push sensors to the resolution limits at center of field of the best lenses at their optimal apertures, and we are not there yet.

With sensor upsizing, one limit is the ever smaller increments in IQ versus ever larger increments in size, weight, cost, and the effort and technique needed to realize those IQ gains. Another is the issue of shutter speed, as soon as adequate DOF is an issue. Because once the larger format and so longer lens needs equal DOF, the aperture ratio must go up, and to gather more "information" as Ray puts it, meaning gathering more light means that ISO speed cannot be pushed up enough to keep the shutter speed equal: the aperture is admitting photons at the same rate, so to feed the larger sensor with more photons needs longer exposures. So once high resolution needs imagery impose tight limits on both motion blur and out of focus blur, there is a hard limit on how much light ("information") the sensor can gather, no matter how large the sensor.

As a rough guideline: in situations where DOF and lighting constraints force the use of a higher than base-ISO senstivity in a particular format, a larger format won't help much.
Logged

EinstStein

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 501
Re: Sony A7R, intersting comaprision ...
« Reply #43 on: November 19, 2013, 05:43:47 pm »

Increasing the pixel density can increase the same!ing rate and  reducing the aliasing error, but it also tends to reduce the dynamic range and hence increasing the quantization error. Both have diminishing return either when hitting the lens resolution or hitting the contrast.

Theoretically, we can increase the dynamic range while reduce the pixel density. The oppoortunity is in the vertical dimensions. For example, by building taller vertical transisters (Finn gate) or trench capacitors. Not sure where is the current technology. Back exposure is another trick, this will increase the transparency for a photo to hit the sensor.

I have compared next 6 and 7, every sales at the camera store convinced me that nex 6 performs better than nex 7, and the clue is in the pixel size. This conclusion I is much easier to achieve since they are the same brand, same camera family.

I also did similar comparison between 5Diii and D800, again,it is mainly about the pixel count and the pixel density. Ignoring the other factors just focus on the quantization error (color fidelity and micro contrast) and the aliasing error (spatial frequency response), I don't see either one offer any compelling benefits.
Theoretical, yes increzng pixel density still has room to gain the benefit, if it can keep everything else equal.
ButI don't think it is possible to keep everything else equal. The choice left is to find the different compromise.

Logged

Chris Barrett

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 730
    • www.christopherbarrett.net
Re: Sony A7R, intersting comaprision ...
« Reply #44 on: November 19, 2013, 06:31:11 pm »

Those are modded Voigtländers, yes? I have the three-lens (unmodded) set...very nice indeed for a film-like look. They remind me of Pentax Super-Taks but with greater clarity. You'll find 'em to be quite sharp when stopped down past f/2.8 too, assuming you actually plan on stopping down at all.

-Dave-

I just ordered the 17 for my BMPCC.  I wish they made a 12, that would be a solid set.  Can't decide if I'm gonna get them all.

bcooter

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 1520
Re: Sony A7R, intersting comaprision ...
« Reply #45 on: November 19, 2013, 07:46:44 pm »

Those are modded Voigtländers, yes? I have the three-lens (unmodded) set...very nice indeed for a film-like look. They remind me of Pentax Super-Taks but with greater clarity. You'll find 'em to be quite sharp when stopped down past f/2.8 too, assuming you actually plan on stopping down at all.

-Dave-

Yea Dave, just can't decide if I'm doing the Cine conversion or leave them as standard still lenses, because they are destined for the gh3s mostly for video and I'll probably be pulling focus myself.

I noticed you mentioned you had Voigtländers and your advice is normally solid and and agenda free.

I have two Voigtlanders I bought for my Leica a long time ago, a 35mm and a 50 something.  I never used them for stills, they never looked really sharp until they were stopped down, but they had a look about them and I know this sounds kind of silly but whatever you shot, if the scene was period neutral the image would look . . . historic.  That's hard to explain but we've all seen the look.

Anyway, I'm looking forward to the .95 for a lot of motion imagery and they'll probably make some nice images on the olympus also.

The real upside of the 43 system is the lens base and size.   The downside is the lenses are only good for this format from two makers, but then again Canons really only work on Canons, etc. etc., so I guess that's not a big deal.

Thx for the info earlier.

BC

Logged

EinstStein

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 501
Re: Sony A7R, intersting comaprision ...
« Reply #46 on: November 19, 2013, 10:47:08 pm »

The message from that comparison is, A7R works much worst than A7, it's simply terrible.
Not only that, it's worst than FUji-X (APS-C).
Logged

ErikKaffehr

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 11311
    • Echophoto
Re: Sony A7R, intersting comaprision ...
« Reply #47 on: November 20, 2013, 12:15:33 am »

Hi,

Fuji-X just has a central crop, no corners. The Leica M8, M9, ME and M (240) are you best option for M-mount lenses. The M (240) even sports a sensor developed for Leica.

Lower resolution helps a bit both vignetting and color cast.

I would not buy an A7 or A7r for M-mount wideangles, it is well know it doesn't work well. It is equally well known that it work well with anything else.

Best regards
Erik

The message from that comparison is, A7R works much worst than A7, it's simply terrible.
Not only that, it's worst than FUji-X (APS-C).
Logged
Erik Kaffehr
 

Telecaster

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 3686
Re: Sony A7R, intersting comaprision ...
« Reply #48 on: November 20, 2013, 12:33:45 am »

Yea Dave, just can't decide if I'm doing the Cine conversion or leave them as standard still lenses, because they are destined for the gh3s mostly for video and I'll probably be pulling focus myself.

I had a look at the Duclos site...seems to me the conversion is worthwhile if you can comfortably swing it $$ wise. The lenses are normally about $1k each, so add another $900 or so for the conversion. Standardized filter threads, follow-focus gearing should you ever need it and stepless aperture control with the 25mm (the other two have this as an option out-of-box).

Quote
I have two Voigtlanders I bought for my Leica a long time ago, a 35mm and a 50 something.  I never used them for stills, they never looked really sharp until they were stopped down, but they had a look about them and I know this sounds kind of silly but whatever you shot, if the scene was period neutral the image would look . . . historic.  That's hard to explain but we've all seen the look.

The early Voigts in particular were explicit homages to classic 1950s/60s designs, so there ya go. That look. My SLR 58/1.4 is even branded Tokyo Kogaku (Topcon) rather than Voigtländer to make the inspiration clearer...at least to folks who know from Topcon.   ;)  (Really good lens too, still available under the Voigtländer brand in F mount.)

Quote
Thx for the info earlier.

My pleasure.

(As for M lenses on non-M cameras...I enjoy using some of them on my X-E1 and old Epson R-D1. But I know I wouldn't use them enough to justify the cost of an A7r even if the Sony's sensor got along well with them. I would be attracted to the Sony by a better fleshed-out native lens lineup. We shall see...)

-Dave-
Logged

ErikKaffehr

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 11311
    • Echophoto
Re: Sony A7R, intersting comaprision ...
« Reply #49 on: November 20, 2013, 01:11:27 am »

Hi,

It is well known that the A7/A7r don't play well with most M-mount wide angles because those lenses have large "chief ray angles", which I belive is the correct technical term.

Leica has done a lot to reduce those problems on the M-series. Thin IR filter, no OLP filtering, offset microlenses, specially designed sensor for the M(240) bar coding lenses so camera can make correction for vignetting and lens cast in firmware or pass on to raw converter. Aperture is estimated by camera body, so EXIF has aperture information.

The optical package in fron of the sensor causes astigmatism on "Biogon" type of lenses. The optical package can be taken into account on modern lenses designs.

Best regards
Erik


(As for M lenses on non-M cameras...I enjoy using some of them on my X-E1 and old Epson R-D1. But I know I wouldn't use them enough to justify the cost of an A7r even if the Sony's sensor got along well with them. I would be attracted to the Sony by a better fleshed-out native lens lineup. We shall see...)

-Dave-
Logged
Erik Kaffehr
 

AreBee

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 638
Re: Sony A7R, intersting comaprision ...
« Reply #50 on: November 20, 2013, 05:13:49 pm »

Ray,

Quote
It's not the greater number of pixels per se - it's the greater quantity of information. You can subdivide a fixed sensor size into more and more, smaller and smaller pixels, but at some point when the image is oversampled, there is no further gain in information. Maintaining the pixel size and lens performance while increasing the sensor area is still the best route to increased information (detail).

Thank you. I think I understand the point you make but would appreciate if you will confirm:

For a given sensor area, an increase in the number of pixels increases the level of detail previously captured by the field of view, up until the point of oversampling, whereas increasing the size of a sensor will, all else being equal, change the field of view because the sensor will crop a larger area from the lens image circle.

Is this what you meant, or have I completely misunderstood?
Logged

EricWHiss

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 2639
    • Rolleiflex USA
Re: Sony A7R, intersting comaprision ...
« Reply #51 on: November 20, 2013, 06:01:29 pm »

The pixel sizes are 6.9, 6, 3.9 microns. If you look carefully you see that the P45+ image on the left has a lot of fake detail while the Sony Alpha 77 on the right has most detail. This doesn't mean that the Alpha 77 outperforms the other two, but it produces less artifacts.

You didn't mention aperture used, or i didn't see it, but I've noticed in my own testing, it matters a lot more for smaller pixels as does critical focusing, quality of glass, and post processing algorithms.   I never would have expected this but it seems to me the actual apex of focus is much more pronounced with smaller pixels.  On a test like yours with the feather, very tiny focus adjustments can change your apparent results in less obvious ways as too small an aperture.  And the tiny details - some RAW software just does better at pulling them out.
Logged
Rolleiflex USA

ErikKaffehr

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 11311
    • Echophoto
Re: Sony A7R, intersting comaprision ...
« Reply #52 on: November 20, 2013, 09:34:35 pm »

Hi,

Aperture was f/8.

But you may have misinterpreted the posting. The interesting part is that you have severe aliasing on the P45+ image resulting in cross hatch pattern on the feather, the Alpha 99 image also shows a lot of aliasing so OLP filtering does not really help. The Sony Alpha 77 with it's smaller pixels + OLP filter doesn't show a lot of alising. These images were taken at fixed distance and at 150mm. So the image grows size is larger with decreasing pixel size. The Sony images were thus downscaled using Lanzos (which causes very little aliasing on it's own) to match the P45+ image. The attached crops below are from P45+ and SLT 77 (6.9 my and 3.9 my).

Stopping down would reduce aliasing, due to diffraction, but another P45+ shot at f/16 still had a lot. Anything that reduces contrast reduces aliasing, and the sharper the system is the more aliasing you get.

I made an exposure with a Planar 80/2.8 (at f/8) from the same position and that image also had a lot of aliasing. So it seems you always get it.

This is nothing unexpected, just a demonstration of applied sampling theory. If the lens transfers significant MTF at pixel pitch you will get aliasing. A document from Schneider I have read said that MTF should be below 10% at Nyquist.

The P45+ has Nyquist limit at 72 LP/mm. if you check the MTF curves below the Sonnar has about 30% MTF at 72 lp/mm. The Sony Alpha 77 has Nyquist limit at 128 lp/mm. My MTF graphs don't go that long, but at 120 lp/mm none of the lenses reach more than 5% MTF, so aliasing does not arise.
 



Best regards
Erik


You didn't mention aperture used, or i didn't see it, but I've noticed in my own testing, it matters a lot more for smaller pixels as does critical focusing, quality of glass, and post processing algorithms.   I never would have expected this but it seems to me the actual apex of focus is much more pronounced with smaller pixels.  On a test like yours with the feather, very tiny focus adjustments can change your apparent results in less obvious ways as too small an aperture.  And the tiny details - some RAW software just does better at pulling them out.
« Last Edit: November 20, 2013, 09:49:20 pm by ErikKaffehr »
Logged
Erik Kaffehr
 

EricWHiss

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 2639
    • Rolleiflex USA
Re: Sony A7R, intersting comaprision ...
« Reply #53 on: November 21, 2013, 01:00:05 am »

A lot of my Rollei Schneider lenses are testing in at .49 cycles per pixel @ f/5.6 for MTF 30 (or was it MTF 50? rats have to check this) in the center of the image with the AFi-ii 12 back.   I had to print a 8 foot test chart to make the tests accurate.    This back has 5.2 um pixel pitch.  

I'm wondering if you had better glass for your tests what you'd see.  I also wonder if you shot at a wider aperture properly focused what you'd see.  Lastly, I wonder what role if any fill factor on the sensor would play with all this.
Logged
Rolleiflex USA

ErikKaffehr

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 11311
    • Echophoto
Re: Sony A7R, intersting comaprision ...
« Reply #54 on: November 21, 2013, 01:24:37 am »

Hi Eric,

Best lens I have. Focusing is as good I can optically, next step is focusing bracket. Something to consider is also sharpening, I measure MTF with no sharpening (in this case).

0.49 cycles/pixels at MTF 50 is a lot.

I plotted LWPH for the Sonnar on the MP45+ at different apertures it peaks at f/5.6. It depends on accuracy of focus naturally:


Best regards
Erik


A lot of my Rollei Schneider lenses are testing in at .49 cycles per pixel @ f/5.6 for MTF 30 (or was it MTF 50? rats have to check this) in the center of the image with the AFi-ii 12 back.   I had to print a 8 foot test chart to make the tests accurate.    This back has 5.2 um pixel pitch.  

I'm wondering if you had better glass for your tests what you'd see.  I also wonder if you shot at a wider aperture properly focused what you'd see.  Lastly, I wonder what role if any fill factor on the sensor would play with all this.
« Last Edit: November 21, 2013, 02:11:21 am by ErikKaffehr »
Logged
Erik Kaffehr
 

EricWHiss

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 2639
    • Rolleiflex USA
Re: Sony A7R, intersting comaprision ...
« Reply #55 on: November 21, 2013, 02:21:43 pm »

It is isn't it?
I noticed your test charts showed under sharpening of about 25%.  I add a tiny amount of capture level sharpening in C1 at a small radius and on my tests Imatest reports about 10% under sharpened.   Default sharpening in C1 is too much according to Imatest and with it, I'll get over .5 cy/px on the charts -which isn't possible and only an artifact of the extra sharpening.   

Try something with your P45 feather images - process out with C1 v7 as a .tiff with no noise reduction and 150 sharpening at r 0.7 and T 0, then import into LR and use more sharpening at r1.1 with detail to 100 and threshold at 15 or 20.      This two step approach always seems to work well for me to get the most real detail with my 80mp back. I might not always use those exact values, but approximately. Depends on aperture, subject etc.

btw I'm not doing that 2 step with my lens tests, just on my art prints and art reproduction work.
Logged
Rolleiflex USA

ErikKaffehr

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 11311
    • Echophoto
Re: Sony A7R, intersting comaprision ...
« Reply #56 on: November 21, 2013, 04:05:48 pm »

Hi Eric,

I will try your recepies but not on these pictures. This is just a aliasing study.

Regarding MTF I sometimes measure it without any sharpening. The MTF curves in this case were all measured on an OLP filtered 3.9 micron sensor.

By the way, Imatest always judges sharpening on medium radius and halos. You can sharpen almost any amount on small pixel radii and it will still report undersharpening. Normal sharpening is based on what would be considered normal sharpening for in camera JPEGs.

Best regards
Erik

It is isn't it?
I noticed your test charts showed under sharpening of about 25%.  I add a tiny amount of capture level sharpening in C1 at a small radius and on my tests Imatest reports about 10% under sharpened.   Default sharpening in C1 is too much according to Imatest and with it, I'll get over .5 cy/px on the charts -which isn't possible and only an artifact of the extra sharpening.   

Try something with your P45 feather images - process out with C1 v7 as a .tiff with no noise reduction and 150 sharpening at r 0.7 and T 0, then import into LR and use more sharpening at r1.1 with detail to 100 and threshold at 15 or 20.      This two step approach always seems to work well for me to get the most real detail with my 80mp back. I might not always use those exact values, but approximately. Depends on aperture, subject etc.

btw I'm not doing that 2 step with my lens tests, just on my art prints and art reproduction work.
Logged
Erik Kaffehr
 

EricWHiss

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 2639
    • Rolleiflex USA
Re: Sony A7R, intersting comaprision ...
« Reply #57 on: November 21, 2013, 05:11:49 pm »

I remember asking Norman about the sharpening, but will ask again.  You can quite clearly see the effects of just changing sharpening values in C1 on the output so I do know that it is sensitive to both radius and also amounts.   You can test this yourself.  Takes just a few minutes. 

To me, its not all that important what parameters I am using for sharpening - just as long as I use the same method for all of the things I want to compare.   My comparisons were mostly made to see the difference between the Zeiss and Schneider offerings of the Rollei MF lenses, plus some comparisons to the the LF lenses. 
Logged
Rolleiflex USA

ErikKaffehr

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 11311
    • Echophoto
Re: Sony A7R, intersting comaprision ...
« Reply #58 on: November 22, 2013, 12:36:14 am »

Hi,

What I do is sharpening in LR at radius depending on diffraction. I found that 0.7 works well for both P45+ and the Sonys when shot at medium apertures. For small apertures I go up to perhaps 1.3. Setting detail slider to 100% gives deconvolution, good for microdetail but gives often some outline. The amount I use 45. Don't want to sharpen more.

You could check out SQF in Imatest, it is the integral of  MTF(f) weighted by contrast sensivity of the eye. It is mostly affected by large radii. I tried my sharpening and adding sharpening with radius 3 and amount 15 in PS. That removes undersharpening and adds a lot of pop.

Finally, I have tested my sharpening on a slightly soft P45+ image, and combined with Focus Magic. Incredible result!

Best regards
Erik



It is isn't it?
I noticed your test charts showed under sharpening of about 25%.  I add a tiny amount of capture level sharpening in C1 at a small radius and on my tests Imatest reports about 10% under sharpened.   Default sharpening in C1 is too much according to Imatest and with it, I'll get over .5 cy/px on the charts -which isn't possible and only an artifact of the extra sharpening.   

Try something with your P45 feather images - process out with C1 v7 as a .tiff with no noise reduction and 150 sharpening at r 0.7 and T 0, then import into LR and use more sharpening at r1.1 with detail to 100 and threshold at 15 or 20.      This two step approach always seems to work well for me to get the most real detail with my 80mp back. I might not always use those exact values, but approximately. Depends on aperture, subject etc.

btw I'm not doing that 2 step with my lens tests, just on my art prints and art reproduction work.
Logged
Erik Kaffehr
 

Bart_van_der_Wolf

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 8913
Re: Sony A7R, intersting comaprision ...
« Reply #59 on: November 22, 2013, 04:22:00 am »

Finally, I have tested my sharpening on a slightly soft P45+ image, and combined with Focus Magic. Incredible result!

Hi Erik,

I agree, FocusMagic rocks!

It can also be used with other applications that accept Photoshop Plugins, e.g. Photoline or IrfanView, so no fear for being locked out when switching applications.

Cheers,
Bart
Logged
== If you do what you did, you'll get what you got. ==
Pages: 1 2 [3] 4 5 ... 9   Go Up