New Hasselblad XCD 20-35/3.2-4.5

[Doppler9000]

What experience do you personally have to be able to conclude (a) there is atypical sample to sample variance for XCD lenses compared to the lenses of other manufacturers and (b) the degree of variance is "unacceptable"? My understanding both from the testing I have done of the same XCD lenses side by side and what I read is that all manufacturers' lenses have some degree of sample to sample variances. That's a totally different issue from a lens that is decentered and obviously flawed. I have six XCD lenses and in the case of three lenses, I purchased multiple copies and tested them side by side in the field and with a Siemens star chart, including the recent XCD 25mm f/2.5. There were very subtle differences between the copies, but nothing that would show up in print (other than a print of a star chart).

My point wasn’t that Hasselblad’s sample variance is higher than say, Fuji’s. It is that given the price difference of $3,430 (2.4X) on the 20-35mm offerings, the Hasselblad ought to have (much) lower observed variance.

Sample variance is not distinguishable from “decentered and flawed”, as you believe, they are on a continuum. To avoid high-variance products from reaching the market, it is up to the manufacturer to either tighten the tolerances in the manufacturing process or catch the problem in a QC step and either rework or reject the lens.

Of the DPR test lens, drevil said “the corners are B A D”

Steve Hendrix ascribed this to decentering or some other QC problem.

Do you disagree with them?

In my view, what is an acceptable degree of variance depends, in part, on the price of the product. YMMV.

By the way, MGrayson said:

“Sample variation is, IMO, a real problem with recent XCD lenses. dpreview has a decentered zoom.”

“I've seen good and bad copies of both the 25 and 28.”

Can we anticipate you calling him out, as well?

 

[JeRuFo]

My point wasn’t that Hasselblad’s sample variance is higher than say, Fuji’s. It is that given the price difference of $3,430 (2.4X) on the 20-35mm offerings, the Hasselblad ought to have (much) lower observed variance.

Sample variance is not distinguishable from “decentered and flawed”, as you believe, they are on a continuum. To avoid high-variance products from reaching the market, t is up to the manufacturer to either tighten the tolerances in the manufacturing process or catch the problem in a QC step and either rework or reject the lens.

Steve Hendrix ascribed this to decentering or some other QC problem.

Do you disagree with them?

In my view, what is an acceptable degree of variance depends, in part, on the price of the product. YMMV.

I agree that it shouldn't happen. It doesn't sound so difficult to design a test rig to at least weed out the obviously flawed ones. The price itself however says probably more about how much the manufacturer expects to sell and the cost of development and manufacture. Still, at this price point it shouldn't cost a significant amount extra to do it right.

 

[buildbot]

I agree that it shouldn't happen. It doesn't sound so difficult to design a test rig to at least weed out the obviously flawed ones. The price itself however says probably more about how much the manufacturer expects to sell and the cost of development and manufacture. Still, at this price point it shouldn't cost a significant amount extra to do it right.

Sigma & Ziess do this:

https://www.dpreview.com/articles/5410838737/go-inside-sigma-s-factory-to-see-how-lenses-are-made?slide=26

Optical Quality Assurance

Pinhole images from new, but optically decentered lens. Olaf Optical Testing, 2016 After we published a number of posts about copy-to-copy variation, people were quick to say that this company or that needs to 'just' improve quality control. I totally agree but realize most people don’t have a...

www.lensrentals.com

I think it would cost a fair bit extra - It's a scale thing probably. Both Phase One and Hasselblad had/have 3rd parties make their lenses, and those 3rd parties are not going to QC every lens out of their factory like sigma will with their fully integrated assembly line with MTF machines at the end. Though for the phase one aerial lens and possibly new XT lenses, they probably do QC each and every one by hand? For the price you hope they would, but c'est la vie.

 

[hcubell]

My point wasn’t that Hasselblad’s sample variance is higher than say, Fuji’s. It is that given the price difference of $3,430 (2.4X) on the 20-35mm offerings, the Hasselblad ought to have (much) lower observed variance.

Sample variance is not distinguishable from “decentered and flawed”, as you believe, they are on a continuum. To avoid high-variance products from reaching the market, it is up to the manufacturer to either tighten the tolerances in the manufacturing process or catch the problem in a QC step and either rework or reject the lens.

Of the DPR test lens, drevil said “the corners are B A D”

Steve Hendrix ascribed this to decentering or some other QC problem.

Do you disagree with them?

In my view, what is an acceptable degree of variance depends, in part, on the price of the product. YMMV.

By the way, MGrayson said:

“Sample variation is, IMO, a real problem with recent XCD lenses. dpreview has a decentered zoom.”

“I've seen good and bad copies of both the 25 and 28.”

Can we anticipate you calling him out, as well?

Why would I call him out? He said he has SEEN good and bad copies of the 25 and 28. Remember, the thousands of purchasers of XCD lenses who have a "good" copy of an XCD lens do not tend to comment here on that. Only the people who have bad copies. This is a common phenomenon with the internet.

 

[hcubell]

Sigma & Ziess do this:

https://www.dpreview.com/articles/5410838737/go-inside-sigma-s-factory-to-see-how-lenses-are-made?slide=26

Optical Quality Assurance

Pinhole images from new, but optically decentered lens. Olaf Optical Testing, 2016 After we published a number of posts about copy-to-copy variation, people were quick to say that this company or that needs to 'just' improve quality control. I totally agree but realize most people don’t have a...

www.lensrentals.com

I think it would cost a fair bit extra - It's a scale thing probably. Both Phase One and Hasselblad had/have 3rd parties make their lenses, and those 3rd parties are not going to QC every lens out of their factory like sigma will with their fully integrated assembly line with MTF machines at the end. Though for the phase one aerial lens and possibly new XT lenses, they probably do QC each and every one by hand? For the price you hope they would, but c'est la vie.

I vaguely recall reading that Zeiss tested EVERY copy of the Otus line of lenses, and that was a major reason for why they were so expensive.

 

[davidrm]

i never understood zoom lenses in the MF realm anyway

Well, anything that can avoid changing lenses and opening up that dust magnet of a sensor has some strong advantages….

 

[anyone]

I have to say for the multiple thousand euros/ dollars Hasselblad charges I expect proper quality control.

Having said that, I’m completely happy with the three XCD lenses I own.

As for the dust issue, a small rocket blower helps

 

[Doppler9000]

Why would I call him out? He said he has SEEN good and bad copies of the 25 and 28. Remember, the thousands of purchasers of XCD lenses who have a "good" copy of an XCD lens do not tend to comment here on that. Only the people who have bad copies. This is a common phenomenon with the internet.

It seems that you missed parts of what I wrote.

@MGrayson wrote:

“Sample variation is, IMO, a real problem with recent XCD lenses. dpreview has a decentered zoom.”

This is the same lens to which I referred, drawing your ire, and is a lens @MGrayson has not ‘SEEN’, presumably.

I agree that there is a bias toward complaints on line. However, with the caveat that I don’t know how many XCD lenses @MGrayson has seen, I am going to assume it is far fewer than the “thousands” to which you referred.

If true, this would suggest that the prevalence of ‘bad’ XCD 25mm and 28mm lenses is perhaps one or two orders of magnitudes higher than you have implied.

Perhaps @MGrayson could provide some statistical insights.

 

[Steve Hendrix]

Regarding sample variation in lenses. In our experience, there are no manufacturers that are presented on this forum (Phase One, Fujifilm, Hasselblad, Leica, Rodenstock, Schneider, etc.) that have a perfect record with zero sample variation. The cost doesn't seem to have a lot to do with it. We've seen variation with $1,400 lenses and variation with $14,000 lenses. On balance, there's probably fewer instances with the $14,000 category, though there's no guarantees. Any manufacturer that would take the time to run any form of testing on their lenses certainly would need to take into account the cost and production aspects of undertaking that. Even when this is in place, human beings are not perfect, and mistakes can be made. I only had 3 cups of coffee this morning instead of my usual 4, and so on. And as a result, we have seen no perfect record from anyone who manufacturers lenses. A high price doesn't guarantee a perfect copy. It should, but ... reality intrudes. Luxury vehicle are often less reliable than economy vehicles, and so on. Many contrary data points to that mindset.

So, bottom line - it happens. With everyone. You'll never really get to the bottom of how often. The factors are usually more about the difficulty of the lens design as much as the cost of the lens. We quality control test every Rodenstock lens that we sell. Primarily because if there is asymmetry, shifting a lens will make that even more glaring. And sometimes a lens just isn't quite at the level overall that it should be, based on our experience of knowing how those lenses should perform. But more often, it is about symmetry. Other than Rodenstock, we cannot quality control test every single lens that we sell, but we can do so upon request.

As a result of the facts of sample variation, I highly recommend testing your own brand new lens as soon as you get it to make sure that you're happy with it. Never assume it looks good without carefully inspecting the results.


Steve Hendrix/CI

 

[hcubell]

It seems that you missed parts of what I wrote.

@MGrayson wrote:

“Sample variation is, IMO, a real problem with recent XCD lenses. dpreview has a decentered zoom.”

This is the same lens to which I referred, drawing your ire, and is a lens @MGrayson has not ‘SEEN’, presumably.

I agree that there is a bias toward complaints on line. However, with the caveat that I don’t know how many XCD lenses @MGrayson has seen, I am going to assume it is far fewer than the “thousands” to which you referred.

If true, this would suggest that the prevalence of ‘bad’ XCD 25mm and 28mm lenses is perhaps one or two orders of magnitudes higher than you have implied.

Perhaps @MGrayson could provide some statistical insights.

My point is that Mr. Greyson does have personal experience with bad copies of two XCD lenses. I gather you have none. Whenever I see people on internet forums who have no direct personal experience with a product jump in to extrapolate from an extremely small number of complaints about that product to a sweeping generalization about the general image quality of a lens or a manufacturer's overall QC, I discount it as it often seems to be based upon wishful thinking rather than statistically significant objective evidence. If Roger Cicala or Steve Hendrix were to offer up an opinion that in their experience the XCD lenses suffer from poorer QC than lenses from other manufacturers, I would find that credible.
Well, I now see that Mr. Hendrix has spoken.

 

[nazdravanul]

Regarding sample variation in lenses. In our experience, there are no manufacturers that are presented on this forum (Phase One, Fujifilm, Hasselblad, Leica, Rodenstock, Schneider, etc.) that have a perfect record with zero sample variation. The cost doesn't seem to have a lot to do with it. We've seen variation with $1,400 lenses and variation with $14,000 lenses. On balance, there's probably fewer instances with the $14,000 category, though there's no guarantees. Any manufacturer that would take the time to run any form of testing on their lenses certainly would need to take into account the cost and production aspects of undertaking that. Even when this is in place, human beings are not perfect, and mistakes can be made. I only had 3 cups of coffee this morning instead of my usual 4, and so on. And as a result, we have seen no perfect record from anyone who manufacturers lenses. A high price doesn't guarantee a perfect copy. It should, but ... reality intrudes. Luxury vehicle are often less reliable than economy vehicles, and so on. Many contrary data points to that mindset.

So, bottom line - it happens. With everyone. You'll never really get to the bottom of how often. The factors are usually more about the difficulty of the lens design as much as the cost of the lens. We quality control test every Rodenstock lens that we sell. Primarily because if there is asymmetry, shifting a lens will make that even more glaring. And sometimes a lens just isn't quite at the level overall that it should be, based on our experience of knowing how those lenses should perform. But more often, it is about symmetry. Other than Rodenstock, we cannot quality control test every single lens that we sell, but we can do so upon request.

As a result of the facts of sample variation, I highly recommend testing your own brand new lens as soon as you get it to make sure that you're happy with it. Never assume it looks good without carefully inspecting the results.


Steve Hendrix/CI

AFAIK, the only lenses that were individually checked for proper calibration after assembly were the Zeiss Otus lenses. I have yet to find a story a bad Otus sample

 

[Doppler9000]

AFAIK, the only lenses that were individually checked for proper calibration after assembly were the Zeiss Otus lenses. I have yet to find a story a bad Otus sample

It would be great if Cosina made some medium format lenses at some point.

 

[AndereObjektiv]

Otus 1.4/85 ZF.2 works great on 33x44 sensors, only a little vignetting. I think the 100 is also supposed to cover the sensor well.

 

[TechTalk]

The second Albrecht Voss sample is interesting as it shows that, with an appropriate profile (and some skilled post processing), the lens is capable of an impressive rectilinear representation of straight lines. My initial reaction was that there is some wavy moustache-type distortion but it must be an optical illusion because, on inspecting the photograph in Photoshop, all the lines - horizontal and vertical – are suitably true.

Thanks for posting the image with the horizontal and vertical reference grid. I appreciate it.

I also appreciate that GetDPI allows members to freely express diverse viewpoints. Thanks!

Distortion was measured in a review of the XCD 20–35 zoom from Digital Camera World. The distortion is considerably lower than what they measured in the Fuji GF 20–35 zoom.

Hasselblad XCD 20–35 mm f/3.2-4.5 distortion

Fuji GF 20–35 mm f/4 distortion

 

[Doppler9000]

Thanks for posting the image with the horizontal and vertical reference grid. I appreciate it.

I also appreciate that GetDPI allows members to freely express diverse viewpoints. Thanks!

Distortion was measured in a review of the XCD 20–35 zoom from Digital Camera World. The distortion is considerably lower than what they measured in the Fuji GF 20–35 zoom.

Hasselblad XCD 20–35 mm f/3.2-4.5 distortion

Fuji GF 20–35 mm f/4 distortion

Do these charts reflect software correction?

 

[TechTalk]

I would think that it's obvious from the results that it's before any correction. What do you think?

 

[Doppler9000]

I would think that it's obvious from the results. What do you think?

I think you’re being snippy, and I think that being snippy is against the “peace be with you” ethos that you robotically drone on and on about, but I guess we can all have bad days.

 

[TechTalk]

I have my share of bad days — more than I'd like lately. I may get snippy on occasion, but I'm not vindictive — So, Peace be with you.

So do you think those distortion results are before or after software correction?

 

[TechTalk]

For those who don't want to read the entire review linked above, their "verdict" is:

"Hasselblad's first ever ultra-wide is another home run for the XCD lineup. Whether you're looking for the fastest and widest lens possible, you're trying to replace a few wide-angle primes, or you simply want a fantastic tool for all-purpose wide-angle shooting, the XCD 20-35E is what you're looking for."

"I truly can't get over the prime-level performance of this lens when it comes to center sharpness – and remember that the XCD 90V is literally the sharpest lens we've ever tested."

"This is a lens that the X System community has long been dreaming about, and it's always a pleasure when dreams come true."

Regarding distortion they say:

"Speaking of distortion, as you may expect from an ultra-wide, there is some barreling and pincushioning as you move through the range, though nothing too nasty – and nothing that a couple of clicks can't easily fix in post."

 

[MGrayson]

A bit more detail on my personal lens experience. I counted 76 purchases of MF (Contax 645, Mamiya 645, Schneider Kreuznach, Pentax 645, Pentax 67, Leica S, Fuji GF, Hasselblad X) or Leica M/L mount lenses. Of those, only two have been decentered, my first copies of the XCD 25 and 28. Three others (a Mamiya 645 and a Pentax 67, and the Contax 350/4) were simply not great examples. Of the way too many FF, APS-C, and µ43 lenses over the past three decades, only three have been decentered, and they were three of the four Sony lenses I purchased when I started an A7II system.

Now some of these are meant for larger sensors than I've used them on, so decentering may not have been visible even if present. I've used (almost) every one of the MF lenses on either the Fuji GFX 100, Leica S3 (64 MP), or X2D. Now that I think of it, Phocus' adaptive CA removal may make the $100 Pentax 67 400/4 usable wide open. (There's a reason it's $100 for that huge chunk of glass. The 20x more expensive and 3x the weight 400/4 ED-IF is a great lens, but not as good as the lighter - and even more expensive - Zeiss 350/5.6 SA.)

Large enough sample size? I have no idea. Certainly miniscule compared to CI or lensrentals, so I defer to them unreservedly.

Matt