p.36 #1 · A thinner sensor stack may be possible after all!
Steve Spencer wrote:
Are you sure that the difference between the A7/A7r and the 5DIII isn't due to the adapter? I think this is a lens where the adapter may need to be shimmed to get optimum performance.
Will a slightly tilted adapter make a difference if you're focusing on the corner? Can you use the tilt function of the lens to account for any adapter tilt for testing purposes?
Mar 23, 2015 at 09:47 AM
Steve Spencer Online • • • • • • • Upload & Sell: On
p.36 #2 · A thinner sensor stack may be possible after all!
Matt Grum wrote:
Will a slightly tilted adapter make a difference if you're focusing on the corner? Can you use the tilt function of the lens to account for any adapter tilt for testing purposes?
It is not so much the tilt in the adapter, but the registration distance. The 24 TSE seems to like a precise registration distance and most adapters are too short.
p.36 #3 · A thinner sensor stack may be possible after all!
Steve Spencer wrote:
Are you sure that the difference between the A7/A7r and the 5DIII isn't due to the adapter? I think this is a lens where the adapter may need to be shimmed to get optimum performance.
Yes. I have spent quite some time testing the adapter (Metabones Mk III), and the lens on the A7r. The outcome is that I had to modify the Metabones adapter to get rid of internal reflections - it's on the A7 thread - covering the inside with matte paint and velours.
I have also measured the adapter at different places and tested different settings, rotating the TS-E to check for a decentered element (the lens had a minor decentering, which was corrected after sending it out for another matter). The finding is that the adapter I have is totally plane. I have also tested multiple focus points with the lens (center focus, focusing on corners) and while there is some field curvature, at f/8-f/11 it is invisible in the huge DOF of the 17mm focal length.
My take away from these tests is that the 17 TS-E is slightly sharper on a Canon body in the extreme corners, fully shifted. Nothing obvious though. There is also noticeable CA on the A7r, which has probably to do with the sensor stack's optical thickness and refraction index. For example I get zero CA in the extreme corners on a 5D MkI, while I see them on the A7r (even resized to 12MP). While minimal, and easily fixable in software, the difference is visible. The optical sensor stack of the 5D MkI is 2mm if I remember correctly, so the glass used (refraction index, I guess) also has an impact on the performance. Not sure if the optical thickness of the sensor stack has been measured on the A7r.
Hope this helps. It will be one of the first lenses I test when I get the A7r back from Kolari; speaking of which it should be delivered to them as we speak
p.36 #4 · A thinner sensor stack may be possible after all!
Steve Spencer wrote:
It is not so much the tilt in the adapter, but the registration distance. The 24 TSE seems to like a precise registration distance and most adapters are too short.
An adapter being too short would mean you can focus past the optimal center focus point the lens was designed for. This could be positive or negative, depending on the direction of field curvature of the lens - pointing toward user, or straight ahead. Obviously if the field curvature looks like a umbrella seen from above, it could prevent the picture from being sharp in the corners.
p.36 #5 · A thinner sensor stack may be possible after all!
beetlephoto wrote:
In theory yes, but experience has shown that the TS-E performs a tiny bit better on the 5DIII for instance, than on the A7/A7r - difference is negligible but present - in the corners. I just hope it is not horrible on the A7rM as it is my favorite lens.
Really? I am using two TSE lenses often on my A7R with MB III adapter - the 24 and the 90 TSE lenses. I see zero difference in corner sharpness compared to the photos taken on my 5D MkII body.
Mar 23, 2015 at 10:20 AM
Steve Spencer Online • • • • • • • Upload & Sell: On
p.36 #6 · A thinner sensor stack may be possible after all!
beetlephoto wrote:
An adapter being too short would mean you can focus past the optimal center focus point the lens was designed for. This could be positive or negative, depending on the direction of field curvature of the lens - pointing toward user, or straight ahead. Obviously if the field curvature looks like a umbrella seen from above, it could prevent the picture from being sharp in the corners.
I think the point that you are missing is that the lens is also designed to be an exact distance from the sensor and being too close for some designs (like some FLE designs) is clearly a bad thing. The 17 TSE has an FLE. What I know is that several people including me (and Nico earlier in this thread) have found that the ZE 21 f/2.8 (which also has an FLE) performs a lot better after you shim the adapter to make it a little thicker. When it is thinner it clearly has more CA and the corners are weaker. It sounds to me like your TSE lenses might improve if you shimmed your adapter to make it a bit thicker. I think it is worth a try.
p.36 #7 · A thinner sensor stack may be possible after all!
Steve Spencer wrote:
I think the point that you are missing is that the lens is also designed to be an exact distance from the sensor and being too close for some designs (like some FLE designs) is clearly a bad thing. The 17 TSE has an FLE. What I know is that several people including me (and Nico earlier in this thread) have found that the ZE 21 f/2.8 (which also has an FLE) performs a lot better after you shim the adapter to make it a little thicker. When it is thinner it clearly has more CA and the corners are weaker. It sounds to me like your TSE lenses might improve if you shimmed your adapter to make it a bit thicker. I think it is worth a try....Show more →
That's a good point, it is worth trying out. I'll measure the thickness of the adapter and see if (thickness of adapter+sony FE theoretical flange distance) matches the Canon EF flange distance. I guess there is also some variation on the FE/EF bodies so maybe the DIY way (shimming the adapter) is a better idea
p.36 #8 · A thinner sensor stack may be possible after all!
retrofocus wrote:
Really? I am using two TSE lenses often on my A7R with MB III adapter - the 24 and the 90 TSE lenses. I see zero difference in corner sharpness compared to the photos taken on my 5D MkII body.
Yes - but you'd have to focus on the pictures to notice the difference. Also - if the downsampled A7r picture matches the 5DMkII picture in terms of sharpness (providing you use a good resizing algorithm), in my opinion it shows that the lens performs a bit better on the 5DII as the 36MP downsampled to 21MP should in theory be sharper than the native 21MP.
p.36 #9 · A thinner sensor stack may be possible after all!
Shimming is obviously faster and easier than grinding adapters which are too thick, but do it carefully. Don't 'overshim' an adapter even for a tiny bit, because you'd loose infinity focus and ,as they say, the cure would be worse than the infirmity,
In my experience it's always better that the adapters are a few cents of mm. too thin, Considering the possible tolerances of bodies and lenses, we need some leeway. I'd say from 0.03 to 0.05 mm. at most, would be a safe bet. Of course manufacturers tend to overdo that philosophy of safety. Isn't rare to find rings which are vastly shorter than required, even among reputed brands.
p.36 #10 · A thinner sensor stack may be possible after all!
artur5 wrote:
Shimming is obviously faster and easier than grinding adapters which are too thick, but do it carefully. Don't 'overshim' an adapter even for a tiny bit, because you'd loose infinity focus and ,as they say, the cure would be worse than the infirmity,
In my experience it's always better that the adapters are a few cents of mm. too thin, Considering the possible tolerances of bodies and lenses, we need some leeway. I'd say from 0.03 to 0.05 mm. at most, would be a safe bet. Of course manufacturers tend to overdo that philosophy of safety. Isn't rare to find rings which are vastly shorter than required, even among reputed brands....Show more →
Yes - I have read that even for the Novoflex adapters. I was lucky as my sample (NEX/LM) was perfect.
I also use the Hawks v5 helicoid adapter for my M lenses, and it is a dream come true: it allows you to adjust infinity precisely via 2 small screws. This combined with the 14x magnifier on the A7r allows to adjust precisely the infinity focus, so that you have a hard stop for infinity. After doing that I checked my Novoflex and it was exactly the same sharpness as the Hawks - probably luck
p.36 #11 · A thinner sensor stack may be possible after all!
beetlephoto wrote:
That's a good point, it is worth trying out. I'll measure the thickness of the adapter and see if (thickness of adapter+sony FE theoretical flange distance) matches the Canon EF flange distance. I guess there is also some variation on the FE/EF bodies so maybe the DIY way (shimming the adapter) is a better idea
I can confirm that in order for the 24 TSEMkII to work on the A7R whether it has the Kolari mod or not the adaptor HAS to be shimmed properly so that the infinity focus is spot on due to the floating element. I was seeing some pretty awful field curvature before I shimmed it and it now is perfect - almost like a new lens and even shifted to the extremes corners are very sharp. Did some test earlier on page 32 in this tread to back this up with examples. - https://docs.google.com/uc?id=0B-0omMe5P91KVWh3MUViQVlhclE&export=download
p.36 #12 · A thinner sensor stack may be possible after all!
Checked and it turns out my 1970's Canon TS 35mm f2.8 also has a floating element, so I should shim the Metabones adapter and see how much it improves. I had posted a few shots with it on the A7M Flickr group (links below, y'all join!) that I can redo after shimming it.
p.36 #13 · A thinner sensor stack may be possible after all!
I was going through the following interview with Samsung, and one of the things they talk about is that the BSI sensor used on NX1 has about 1/3rd the pixel height compared to the traditional front-side illuminated sensors. They talk about how it can reduce corner shading and color bleeding etc:
p.36 #14 · A thinner sensor stack may be possible after all!
nicoimages wrote:
I can confirm that in order for the 24 TSEMkII to work on the A7R whether it has the Kolari mod or not the adaptor HAS to be shimmed properly so that the infinity focus is spot on due to the floating element. I was seeing some pretty awful field curvature before I shimmed it and it now is perfect - almost like a new lens and even shifted to the extremes corners are very sharp. Did some test earlier on page 32 in this tread to back this up with examples. - https://docs.google.com/uc?id=0B-0omMe5P91KVWh3MUViQVlhclE&export=download
Thanks
Nichoals
I'll definitely give it a shot, thanks! My 17 ts-e is already very good but who knows!
p.36 #15 · A thinner sensor stack may be possible after all!
To all the people here who has done the mod and is sharing their experience with us : thanks a lot, folks. This is a very informative post. Kolari should offer you a discount for being such fearless beta-testers... .
Being in a Columbo mood today, there're still a couple of questions which bother me :
-First : Have you you noticed increased sensibility to IR with the modded cameras' (i.e. black synthetic fabrics turning purplish-brown ..).
- Second : Concerning the annoying sensor reflections of the original A7, I expect that there's some ( or a lot of) improvement with the new filter. Any side to side comparison before and after ?
Sory if these two topics have been discussed already, but I don't recall specific comments about them.
p.36 #16 · A thinner sensor stack may be possible after all!
curious80 wrote:
I was going through the following interview with Samsung, and one of the things they talk about is that the BSI sensor used on NX1 has about 1/3rd the pixel height compared to the traditional front-side illuminated sensors. They talk about how it can reduce corner shading and color bleeding etc:
So could it be that the BSI sensors are the answer to the problems with rangefinder lenses on A7x?
Up to this point, sensor glass thickness has strong influence on field curvature and astigmatism, which are more important problem here, of these RF WA lenses. The A7s is already so good at dealing with color cast and vignette. After the mod, with a change in wavelength transmission, it seems that the tiny bit of cyan cast on the stock A7s is gone as well.
p.36 #17 · A thinner sensor stack may be possible after all!
sebboh wrote:
the g28 doesn't have a floating element so the adapter thickness shouldn't make a difference as long as you are choosing the optimal focus point in LV.
Derek, so I did a comparison to see it for myself on the G 21, similarly without floating element. You're right. The optimal infinity setting would give better macro and micro contrast over the whole picture (most notably in the center). However, the difference is only visible under this kind of comparison. I couldn't detect any change in field curvature or increase in astigmatism. I guess the G 28 is just not as good as the G 21.
p.36 #18 · A thinner sensor stack may be possible after all!
hiepphotog wrote:
Derek, so I did a comparison to see it for myself on the G 21, similarly without floating element. You're right. The optimal infinity setting would give better macro and micro contrast over the whole picture (most notably in the center). However, the difference is only visible under this kind of comparison. I couldn't detect any change in field curvature or increase in astigmatism. I guess the G 28 is just not as good as the G 21.
that was my experience as well comparing the two centrally on a NEX, not that the g28 is any slouch in that regard. with the cover glass modification is the g21 also better in the corners stopped down or does the g28 manage to pull ahead there?
p.36 #19 · A thinner sensor stack may be possible after all!
sebboh wrote:
that was my experience as well comparing the two centrally on a NEX, not that the g28 is any slouch in that regard. with the cover glass modification is the g21 also better in the corners stopped down or does the g28 manage to pull ahead there?
Even at f/11, the G 28 is still not as sharp as the G21 in the corners; it still has a bit of fuzziness. With less DOF, the 28 is a bit harder to manage to get sharp corner-to-corner as well.
p.36 #20 · A thinner sensor stack may be possible after all!
charles.K wrote:
Hi Jack, I suspect that having a minimum shutter of 1/125s and above is a lot of the issues even with the wide angles! For myself I know that my technique cannot lapse in holding the A7r steady, otherwise I will notice when I download the RAW files. It is percentage thing, that if I maintain a higher a shutter speed, that is above 1/125sec for 50mm and less, there are very few issues due to shake/vibration when I am holding the cam.
Charles, now that weather cleared up I managed to take a few better test shots, this time in 1/250 - 1/350s range instead of 1/90s. Also, I corrected my own post processing mistake of using de-vignetting option (luminance) in CornerFix. Making long story short, luminance setting carried from ZM Biogon 35 profile caused noise amplification in processed area that give the impression of much higher ISO or heavy noise.
