Just received 5ds and testing VS my CF-39MS in same lens conditions

[BernardLanguillier]

It doesn't have to equal MF lens at all. The point of the exercise is to use the same lens, projecting absolutely the same image (inside a center crop of course), thus leaving the only variable to be the sensor. Introduce another lens, and you are not testing just the sensor anymore, but a combination of lens + sensor. Not that such a comparison would not be valuable in its own right, but than we are talking about a different test.

True, but using a lens that is average may put the more resolving capturing device at a disadvantage since the lens may act as the bottleneck.

I don't think that the Canon 100mm macro is worse per sensor square mm than H lenses though, so there is probably no issue here, but the 100mm is still a pretty average lens relative to its 35mm counterparts according to lenscore.org.

Cheers,
Bernard

[ErikKaffehr]

Hi Bill,

I don't know if this is true any longer. Modern MF lenses can be seriously good, compare for instance the Zeiss 100/2.8 Macro Planar with Hasselblad 120/4. Top Zeiss 100/2.0, below Hasselblad HC 120/.

Now, MF-lenses are more expensive. It could be said that the Otus may be a better comparison.

Anyway, I am thankful for any good data, so Chiek's efforts and willing to share are much appreciated. I

Best regards
Erik

That's right. If only the central area of the MF frame is being evaluated, there is no need for the lens to cover the entire MF frame. Generally, MF lenses resolve fewer line pairs/mm since they have to cover a larger frame area.

Bill

[EricWHiss]

That's exciting Chiek! Thank you for sharing some images and your opinions.

I'm very curious of course to see the comparison with the M/S too and also how the new canon performs when stopped down some.  How is the DR on the new canon? Shadow noise?

[ErikKaffehr]

Thanks,

I would suggest we need to having raw file support to make proper comparisons.

The lens you use is almost certainly a better performer at f/5.6 - f/8 than at f/13, and you would use a larger aperture on the Canon to get the same depth of field compared to the Hasselblad back.

In addition, the 5Ds has a OLP filter, a good thing in my view, but I would say it needs a bit more sharpening,

What I would try to do may be to use JPEG with no (or minimal) sharpening and than try to use a good sharpening tool like "FocusMagic".

Another observation. You would probably use the Hasselblad back with a longer lens, as the two cameras will give different crops. Image diameter on the CF-39 is 60 mm (I believe) while on Canon it is 43 mm.

Anyway, thanks a lot for sharing!

Best regards
Erik

1st test in studio.


this link to download screen shot
https://s3.amazonaws.com/images.us.viewbook.com/59e9d958f6600bea91c80a684b4259ad.jpg?response-content-disposition=attachment&X-Amz-Algorithm=AWS4-HMAC-SHA256&X-Amz-Credential=0SM65CZEDKVWR7WP26G2%2F20150323%2Fus-east-1%2Fs3%2Faws4_request&X-Amz-Date=20150323T125435Z&X-Amz-Expires=3600&X-Amz-SignedHeaders=host&X-Amz-Signature=92bcde69b21ca77bb6b85fb9f1906a4ab1ec69170bdbff6f0a2b62e2b999d43c
Just tested outdoor now. not develop yet.

[Bart_van_der_Wolf]

Yes, that is potentially correct.  But are we sure the center resolving power of the Canon lens (being used) is equal to an MF lens?

We can be, by measuring it.

In general though, and not specifically focusing on (pun intended) the high quality macro lens that the OP used, good quality 35mm FF lenses need to resolve more detail, because the image needs to be enlarged more for a given output size. It's also relatively easier to do so, because the image circle diameter is relatively limited. MF lenses on the other hand need to be corrected for a larger image circle, albeit with longer focal lengths, which constitutes another set of design limitations. Besides, MF images usually require less magnification, so resolution is of lower importance in design trade-offs.

Another factor is that most recent lens designs are specifically made for digital capture. That means different things on MF than on 35mm FF or smaller, due to differences in filterstack thickness/composition and which are part of the optical design parameters for modern lenses.

So it's best to actually measure resolution if one wants to compare performance in an objective way. But we have to assume that the OPs time with the camera is limited, so it may not be possible to conduct all tests one can dream of.

A quick and yet accurate assessment, both visually as well as quantifiably, can be done with my free resolution test chart. Accurate focusing is very important, the rest can be evaluated afterwards based on unsharpened Raw conversions (and Raw converters do also make a difference). The optimal aperture for the EF 100mm f/2.8L Macro IS is at f/4 or f/4.5 (see attachment), which is also in safe territory with regards to per pixel diffraction on smaller sensels.

Cheers,
Bart

[chiek]

HC50-110 set to 100 / 24-70II set to 70
ISO 100 F9, 1/160sec (fastest sync speed to 5ds)

Why f9 ? not F5.6~8 ?
Because of Studio Flash minimum output… Can't down it.







very similar to 39Mpx single shot and 50Mpx 5ds
but definitely win 39Mpx multishot.

You can download 2560x1600, 100% images,
http://chiek.viewbook.com/5ds

[Ken R]

Thx for the tests.

The 5DS shots look like they have some kind of color cast or maybe its a white balance / processing issue? The Hasselblad shots look much more neutral and nice.

[chiek]

The 5DS shots look like they have some kind of color cast or maybe its a white balance / processing issue? The Hasselblad shots look much more neutral and nice.

I don't touch anything.
Only my 30" monitor screenshot.

[Bart_van_der_Wolf]

I don't touch anything.
Only my 30" monitor screenshot.

Hi Chiek,

Thanks for the tests so far.

It's unfortunately hard to compare resolution without knowing if the files were optimally sharpened (or not sharpened at all, so we can try ourselves). They will probably require quite different settings. I assume that the white balance was set with a neutral grey patch sampling.

Cheers,
Bart

[bjanes]

Hi Bill,

I don't know if this is true any longer. Modern MF lenses can be seriously good, compare for instance the Zeiss 100/2.8 Macro Planar with Hasselblad 120/4. Top Zeiss 100/2.0, below Hasselblad HC 120/.

Now, MF-lenses are more expensive. It could be said that the Otus may be a better comparison.

Anyway, I am thankful for any good data, so Chiek's efforts and willing to share are much appreciated. I

Best regards
Erik

Erik,

The MTF graphs are difficult to interpret since the resolutions for the three sets of data are not shown. For MF a lower resolution may have been used since the images would need less magnification to achieve a given print size. As I recall, the resolutions are often appropriate for an 8 x 10 inch print (ISO A4) viewed at arm's length (~34 cm).

Regards,

Bill

[Ken R]

I think the resolution is pretty even and any difference might be more about the lens, depth of field and processing. The default color looks nicer on the hasselblad for sure. At least with this setup. Which is in line from what I see with Canon bodies which really excel in mixed light situations but in the studio it is hard to get neutral color with them. The difference is not huge with the 5DS posted here so thats nice. Would like to see how the Canon 5DS handles more organic shapes like foliage.

[ionut macri]

thanks for sharing your results!
looks like the best photos so far with the Canon 5ds.
the 24-70 seems not up to the task for 50m pixels, for sure the 100L is better suited on the studio at least.

[Bart_van_der_Wolf]

I think the resolution is pretty even and any difference might be more about the lens, depth of field and processing.

I think it's hard to already draw firm conclusions about resolution, because I can still see additional resolution in the Canon 5DS file that has not been used yet. It's also a bit premature to start additional processing based on the crops, because the crops seem to already have been sort of sharpened, so not the best basis for further postprocessing. Therefore, what would be needed is unsharpened files, preferably with some additional info (if not in EXIF), like shooting distance or magnification factor. Of course it would also help to have a good lens test at that shooting distance, although it may be possible to reverse engineer it a bit.

The f/9 diffraction pattern alone, together with the shooting distance, spans at least some 3.2 pixels in diameter in the plane of best focus. And that is without additional lens blur. Given that the diffraction pattern is already responsible for a significant portion of the Capture blur, that means that deconvolution sharpening will make a significant positive difference.

The default color looks nicer on the hasselblad for sure. At least with this setup. Which is in line from what I see with Canon bodies which really excel in mixed light situations but in the studio it is hard to get neutral color with them. The difference is not huge with the 5DS posted here so thats nice. Would like to see how the Canon 5DS handles more organic shapes like foliage.

The color rendering will depend a lot on the color profile used. Given that Canon specifically mentioned the improved color filters, it would also matter a lot which type of rendering (natural/faithful/etc.) is chosen in DPP.

Cheers,
Bart

[orc73]

great! thanks for the test!

I kind of like that you put the same lense on the sensor to compare. Otherwise it would be a lense comparison more so.

That brings us to the fact though, that Canon's weekness will be the lenses. None of the Canon lenses produced me sharp images wide open, and the focus was always a bit random for me(sorry I talk about 5dm2).

Hasselblads truefocus is the only autofocus so far, that delivers me consistent output then any Nikon or Canon. And the lenses are excellent.
This together with the better color and tone rendering was making it a far better system.

Maybe the 5ds will change some of those facts. I wouldn't mind to carry less weight and saving money in the same time

[eronald]

My 5D2 and the one which my friend Norman used, and one which a Canon used for some tests with me all had very random AF.
I think later models -and some earlier ones like the 1ds3 - are much better.

Edmund

great! thanks for the test!

I kind of like that you put the same lense on the sensor to compare. Otherwise it would be a lense comparison more so.

That brings us to the fact though, that Canon's weekness will be the lenses. None of the Canon lenses produced me sharp images wide open, and the focus was always a bit random for me(sorry I talk about 5dm2).

Hasselblads truefocus is the only autofocus so far, that delivers me consistent output then any Nikon or Canon. And the lenses are excellent.
This together with the better color and tone rendering was making it a far better system.

Maybe the 5ds will change some of those facts. I wouldn't mind to carry less weight and saving money in the same time

[Theodoros]

Hi Chiek,

Thanks for the tests so far.

It's unfortunately hard to compare resolution without knowing if the files were optimally sharpened (or not sharpened at all, so we can try ourselves). They will probably require quite different settings. I assume that the white balance was set with a neutral grey patch sampling.

Cheers,
Bart

True... the MF files will benefit further from optimum sharpening....

[Theodoros]

I think the resolution is pretty even and any difference might be more about the lens, depth of field and processing. The default color looks nicer on the hasselblad for sure. At least with this setup. Which is in line from what I see with Canon bodies which really excel in mixed light situations but in the studio it is hard to get neutral color with them. The difference is not huge with the 5DS posted here so thats nice. Would like to see how the Canon 5DS handles more organic shapes like foliage.

You think wrong... Res is with MF by an obvious margin... it's DOF that is different. Res (for MF) will benefit further with optimum sharpening.

[ErikKaffehr]

Hi Bill,

You have a good point about the MTF curves, but all three pairs show 10/20/40 lp/mm. This is the way both Hasselblad and Zeiss present their data and it used to be somewhat of an European standard.

What I can see is that my Hasselblad/Zeiss lenses perform about on the same level as my DSLR lenses near the optical axis. I don't do tests in the corners as I don't have room for large targets.

This compares my Sony Alpha 16-80/3.5-4.5 zoom with my Planar 80/2.8. Both lenses are very sharp at the center. The plots here are lp/mm.



Corners on the Hasselblad lenses are a bit weaker, but they are still almost perfectly sharp in the corners on A2 prints.

This image was shot with a Distagon 50/4 at around f/13 a few days ago:


This is a central crop:


And this one is edge:


I am shooting a Sony Alpha 99 at 24 MP and the Hasselblad with P45+ at 39MP, what I would say is that I don't see a significant difference normally between the two at any distance on A2, size which is what I normally print. At A1 I would say that the advantage of the P45+ is clearly visible. Im am 59, and wear progressive glasses and I am near sighted. A younger man person may have a different viewing experience.

I enclose the Zeiss MTF data for the Distagon 50/4 and for a Distagon 35/2.

I am not a Zeiss freak, but Zeiss has good and comparable MTF data for all of their lenses, so I like to refer to them.

Best regards
Erik

Erik,

The MTF graphs are difficult to interpret since the resolutions for the three sets of data are not shown. For MF a lower resolution may have been used since the images would need less magnification to achieve a given print size. As I recall, the resolutions are often appropriate for an 8 x 10 inch print (ISO A4) viewed at arm's length (~34 cm).

Regards,

Bill

[Theodoros]

My 5D2 and the one which my friend Norman used, and one which a Canon used for some tests with me all had very random AF.
I think later models -and some earlier ones like the 1ds3 - are much better.

Edmund

What AF performance has to do with anything? It is obvious that Chiek has used MF with special care to have focus "dead" on the screw's head in the first images and the "line" of the tool's surface on the rest.... Only DOF is different and the superiority of MF (with the lenses used) is obvious. MF will further benefit (more than FF will) with optimum sharpening. DR is with MF too...

[Slobodan Blagojevic]

... Res is with MF by an obvious margin... Res (for MF) will benefit further with optimum sharpening.

Ok, will refrain from being confrontational, so a couple of genuine questions for anyone who understands those things better than I do:

1. Is it really possible to tell the difference on basically smooth surfaces, without any micro detail?

2. Where is that "obvious margin" obvious?

3. Isn't it the other way around, i.e., that Canon file would benefit more from optimum sharpening? It is Canon that needs to overcome blurring from OLP filter, thus an optimal sharpening would restore sharpness.

The attached file shows the area where I see more surface texture in the Canon file.