Home · Register · Join Upload & Sell

Moderated by: Fred Miranda
Username  

  New fredmiranda.com Mobile Site
  New Feature: SMS Notification alert
  New Feature: Buy & Sell Watchlist
  

FM Forums | Canon Forum | Join Upload & Sell

1       2       3              10      
11
       12       13       14       end
  

Archive 2017 · Article posted on Dpreview re: Canons lack of innovation

  
 
alundeb
Offline
• • • • •
Upload & Sell: Off
p.11 #1 · p.11 #1 · Article posted on Dpreview re: Canons lack of innovation


Holger wrote:
Dpreview studio scene shows less aliasing artefacts in some areas where others (D7100, has no AA filter or D810, A7rii) show it.


Those comparisons also show that the 24 MP X-Trans files from the X-Pro 2 show less fine detail than 24 MP sensors with AA filters and much less detail than 24 MP files without AA filter but Bayer array.



Sep 18, 2017 at 03:05 AM
Holger
Offline
• • • •
Upload & Sell: Off
p.11 #2 · p.11 #2 · Article posted on Dpreview re: Canons lack of innovation


Wilbus wrote:
As a m43 user I think you are right with most everything you say here. Only way to make smaller lenses is with smaller sensors (most of the time). Sony is living proof of this. You "save" a few hundred grams in the body but the lenses are as heavy or heavier then Nikon and Canon counterparts. APS-C and m43 can live with a bit smaller lenses. For me the E-M1 feels great in my hand, much better then the recent Nikon D810, D5 etc bodies I have tried but I also suspect this is a lot to do with
...Show more

The nice thing is abount the Sony FF mirrroless, you can use smaller lenses, too, if you like, you don't need to go with f2.8 zooms or f1.4 lenses.
Take the 28/2, 35/2.8, 55/1.8, 85/1.8 or the f4 zooms (like the new 12-24f4). You can use many small manual focus lenses (some with excellent quality, like the Loxia 21mm/f2.8, or good compromises like the Voigtländer lenses). Having the larger sensor still lets your 35/2.8 perform like what you get on MFT with a 17mm/1.4 lens in terms of AOF, DOF and S/N ratio, for example. What you won't get is something similar to a Oly 300/f4. A 600mm lens with f8 (for similar DOF and S/N ratio) will probably not be made by Sony. If you are happy with less DOF control, mft provides the smaller package. But using mft with larger aperture lenses to reduce DOF and your weight and size will grow to similar values.



Sep 18, 2017 at 03:12 AM
Holger
Offline
• • • •
Upload & Sell: Off
p.11 #3 · p.11 #3 · Article posted on Dpreview re: Canons lack of innovation


alundeb wrote:
Those comparisons also show that the 24 MP X-Trans files from the X-Pro 2 show less fine detail than 24 MP sensors with AA filters and much less detail than 24 MP files without AA filter but Bayer array.


I wouldn't go that far and be more careful in stating this. When I had the D7100 I tried out the xt2 just for fun and the differences were negligible (strongly raw converter dependent, lens dependent, color etc.). Printed out the difference was not detectable in most cases anyway. Oftentimes people misinterpret false detail due to alisaing with true details. Our D750 surely didn't look better, so your claim is too general in my opinion (strength of AA filter, direction, plays a role, too).



Sep 18, 2017 at 03:23 AM
Wilbus
Offline
• • • •
Upload & Sell: Off
p.11 #4 · p.11 #4 · Article posted on Dpreview re: Canons lack of innovation


Holger wrote:
The nice thing is abount the Sony FF mirrroless, you can use smaller lenses, too, if you like, you don't need to go with f2.8 zooms or f1.4 lenses.
Take the 28/2, 35/2.8, 55/1.8, 85/1.8 or the f4 zooms (like the new 12-24f4). You can use many small manual focus lenses (some with excellent quality, like the Loxia 21mm/f2.8, or good compromises like the Voigtländer lenses). Having the larger sensor still lets your 35/2.8 perform like what you get on MFT with a 17mm/1.4 lens in terms of AOF, DOF and S/N ratio, for example. What you won't get is something
...Show more

So true, it's also a reason I am hoping a camera company will release a FF mirrorless. I have gotten used to having to work a little to throw the background out of focus. Not only with very fast glass but also to chose a location where I get more background separation from the beginning.

You can use slower F2.8 lenses rather then the faster 1.4 lenses as long as those 2.8 lenses are still good (many of them are). What does bother me with some of the slower glass for many systems though is they are never built like the "pro" glass. Take Olympus PRO series for example, they are all built like tanks, weather sealed, metal, high end and quality through and through. Even though you may find something slower and still with good image quality in a FF brand, those lenses are rarely weather sealed and when they are, they are still built like "crap" compared to the high end offerings in m43.

This is why one can never really compare lenses to lenses either.

But yes, all in all one can get a small system with a FF mirrorless body.

And yes, I did say I am waiting for a camera company to make an FF mirrorless. No, Sony is not a camera company although they create beautiful sensors but I haven't been able to get along with their Sony A7's. Compare the A7's with an Olympus, Nikon or Canon and you know that those companies have a far longer history in ergonomics. Sony releases a lot of beta products (that have gotten a lot better). They also failed to release lenses for the first few years, which has also gotten a lot better but it also proves that they don't quite understand what makes a camera system.



Sep 18, 2017 at 03:25 AM
alundeb
Offline
• • • • •
Upload & Sell: Off
p.11 #5 · p.11 #5 · Article posted on Dpreview re: Canons lack of innovation


Holger wrote:
, so your claim is too general in my opinion


I was only commenting on the comparison tool you referred to, where it is very clear that also real detail is missed. It is probably the raw converter that has killed all fine detail, and that's what they have to do to avoid artifacts in the details.



Sep 18, 2017 at 03:33 AM
Mike Veltri
Offline
• • •
Upload & Sell: Off
p.11 #6 · p.11 #6 · Article posted on Dpreview re: Canons lack of innovation


Funny that when I opened this tread to have a look, Malwarebytes pops up a warning that it has blocked the site?



Sep 18, 2017 at 03:52 AM
melcat
Offline
• • • •
Upload & Sell: Off
p.11 #7 · p.11 #7 · Article posted on Dpreview re: Canons lack of innovation


Holger wrote:
I would not go so far as to call it errant development or scam. There is a valid idea behind backed by science (it doesn't need to be fully random).


It isn't random at all. The X-trans array is a regular repeating array, and as such it *is* subject to aliiasing unless an antialiasing filter is placed in front of it. Since it actually has a lower spatial frequency in some channels than a standard Bayer array, the X-trans array is *more* subject to aliasing than a Bayer filter when neither has an AA array.

Clearly someone at Fuji didn't understand the maths, and I'm afraid as far as I'm concerned so long as X-trans is on the market I wouldn't go near any Fuji camera product. Who knows what moronic things they might do with, say, raw compression? (I note that their new medium format mirrorless system does *not* use X-trans; perhaps the company really does understand their mistake, but have snookered themselves with marketing of the APS-C systems.)

Dpreview studio scene shows less aliasing artefacts in some areas where others (D7100, has no AA filter or D810, A7rii) show it.

The aliasing will occur at different frequencies depending on the channel, and it's harder to construct a test which shows it and easier to find examples where it doesn't. Dpreview aren't exactly the gold standard - who can forget the tests using JPEGs and the "equivalent aperture" crap?

I found it quite disappointing when the 5DSr appeared - Canon caving in to marketing over engineering.



Sep 18, 2017 at 04:18 AM
charlyw
Offline
• • •
Upload & Sell: Off
p.11 #8 · p.11 #8 · Article posted on Dpreview re: Canons lack of innovation


melcat wrote:
I found it quite disappointing when the 5DSr appeared - Canon caving in to marketing over engineering.


Yes, that was quite a disappointment - but at least they didn't put all of their eggs into one basket, they have produced the 5Ds which is clearly the choice for anyone not buying into the AA-less craze. And a craze it is because at 50Mp Bayer resolution we are still only a fifth of the way to be really in a position to do without an AA filter.

Optical properties of visible light allow to calculate the smallest detail size that can be resolved with a perfect lens at any given aperture, that formula is only depending on aperture and the wavelength of the light - and the resolution limits are 100Mp for APS-C and roughly 250Mp for a full frame camera. Then even a hand crafted, absolute perfect lens at f/2.8 (that's the assumption I put into that calculation, faster lenses could outresolve but they would need to do that at a wider aperture and the likelihood of that ever happening within financial viability are pretty close to 0) can no longer lead to aliasing on a Bayer sensor, on the x-trans the resolution limits need to be higher by a factor of 2 due to the incompetently chosen spatial placement of the red and blue pixels...



Sep 18, 2017 at 04:33 AM
alundeb
Offline
• • • • •
Upload & Sell: Off
p.11 #9 · p.11 #9 · Article posted on Dpreview re: Canons lack of innovation


charlyw wrote:
Optical properties of visible light allow to calculate the smallest detail size that can be resolved with a perfect lens at any given aperture, that formula is only depending on aperture and the wavelength of the light - and the resolution limits are 100Mp for APS-C and roughly 250Mp for a full frame camera. Then even a hand crafted, absolute perfect lens at f/2.8 (that's the assumption I put into that calculation, faster lenses could outresolve but they would need to do that at a wider aperture and the likelihood of that ever happening within financial viability are pretty close
...Show more

I can demonstrate aliasing at F/4 on a 360 MP equivalent sensor (Pentax Q) in the center of a lens with 42 mm image circle. Still, for practical purposes, above 200 MP on FF wil be sufficient to get less aliasing than with 50 MP and anti-aliasing filter. Even at 120 MP it will become diificult to advocate the use of an AA filter.



Sep 18, 2017 at 05:23 AM
dehowie
Offline
• • •
Upload & Sell: Off
p.11 #10 · p.11 #10 · Article posted on Dpreview re: Canons lack of innovation


Seriously that article was the biggest load of crap i have read in years.




Sep 18, 2017 at 05:57 AM
charlyw
Offline
• • •
Upload & Sell: Off
p.11 #11 · p.11 #11 · Article posted on Dpreview re: Canons lack of innovation


alundeb wrote:
I can demonstrate aliasing at F/4 on a 360 MP equivalent sensor (Pentax Q) in the center of a lens with 42 mm image circle.


Hm, that should be physically impossible according to the optical laws of physics - just look at the size of the airy disks at f/4 compared to the pixel pitch of the sensor.



Sep 18, 2017 at 06:08 AM
alundeb
Offline
• • • • •
Upload & Sell: Off
p.11 #12 · p.11 #12 · Article posted on Dpreview re: Canons lack of innovation


charlyw wrote:
Hm, that should be physically impossible according to the optical laws of physics - just look at the size of the airy disks at f/4 compared to the pixel pitch of the sensor.


It is not the size of the airy disc that sets the diffraction limit for resolved detail. It is the distance between two adjacent airy discs where the two tops are still slightly higher than the intersection. When this difference equals 9% MTF, it is the definition of the Rayleigh criterion.

See under "diffraction":

http://www.normankoren.com/Tutorials/MTF1A.html

At F/4 the Rayleigh limit is 408 lp/mm



Sep 18, 2017 at 06:11 AM
Holger
Offline
• • • •
Upload & Sell: Off
p.11 #13 · p.11 #13 · Article posted on Dpreview re: Canons lack of innovation


melcat wrote:
It isn't random at all. The X-trans array is a regular repeating array, and as such it *is* subject to aliiasing unless an antialiasing filter is placed in front of it. Since it actually has a lower spatial frequency in some channels than a standard Bayer array, the X-trans array is *more* subject to aliasing than a Bayer filter when neither has an AA array.

True. It is, however, possible to construct quasi-random CFAs, which allow to reduce aliasing artefacts by reducing the overlap between chrominance and luminance signal. There are lots of papers investigating that in the scientific literature.

Clearly someone at Fuji didn't understand the maths, and I'm afraid as far as I'm concerned so long as X-trans is on the market I wouldn't go near any Fuji camera product. Who knows what moronic things they might do with, say, raw compression? (I note that their new medium format mirrorless system does *not* use X-trans; perhaps the company really does understand their mistake, but have snookered themselves with marketing of the APS-C systems.)

I am not that religiously against x-trans, but I am not a user or fan. The disadvantages of the Fuji-implementation make it less suited for me . But nevertheless I like the look many generate with it. As long as the clients are happy, one shouldn't predict the doom of photography.

Dpreview studio scene shows less aliasing artefacts in some areas where others (D7100, has no AA filter or D810, A7rii) show it. The aliasing will occur at different frequencies depending on the channel, and it's harder to construct a test which shows it and easier to find examples where it doesn't. Dpreview aren't exactly the gold standard - who can forget the tests using JPEGs and the "equivalent aperture" crap?

I don't know what you mean with equivalent aperture crap. In general equivalence is something I find very useful and a concept important in many fields of science.





Sep 18, 2017 at 06:28 AM
charlyw
Offline
• • •
Upload & Sell: Off
p.11 #14 · p.11 #14 · Article posted on Dpreview re: Canons lack of innovation


Holger wrote:
True. It is, however, possible to construct quasi-random CFAs, which allow to reduce aliasing artefacts by reducing the overlap between chrominance and luminance signal. There are lots of papers investigating that in the scientific literature.

And then any image enhancement in demosaicing is prevented. Sorry, you can't look at one stage without including the other. Current demosaicing algorithms do include a lot of reconstruction of edges and the like because the regular Bayer filter array gives them the data to do so. With a pseudo random CFA these possibilities will diminish even more. Look at the complexity in the current x-trans demosaicing which even though a lot of the drawbacks of traditional bayer filters still apply (because of the highly regular nature of the CFA) not enough remains to apply the image enhancing algorithms. If something truly random will be used then the demosaicing will be erratic and will have severe problems even reconstructing a color image without discontinuities... Sure, the noise will look nice but any edges in the subjects will look mottled as if the mother of all moth swarms had had a go at the edges...

Holger wrote:
But nevertheless I like the look many generate with it.

That's not because of the x-trans CFA but rather despite the CFA...



Sep 18, 2017 at 06:45 AM
DAphoto77
Offline
• •
Upload & Sell: Off
p.11 #15 · p.11 #15 · Article posted on Dpreview re: Canons lack of innovation


I think you all are arguing about " how many angels can dance on the head of a pin" A lot of these posts argue about peripheral technicalities that are probably financially not feasible in an actual declining competitive market. The financially fittest will survive in a milieu that is slowly being decimated by cell phone photo picture technology. I believe that financial success is determined by accountants and market research above technical superiority. Canon has the best chance of survival I think.. The the financially supportive low end customer base has moved to cell phones.









Sep 18, 2017 at 07:56 AM
Holger
Offline
• • • •
Upload & Sell: Off
p.11 #16 · p.11 #16 · Article posted on Dpreview re: Canons lack of innovation


charlyw wrote:
And then any image enhancement in demosaicing is prevented. Sorry, you can't look at one stage without including the other. Current demosaicing algorithms do include a lot of reconstruction of edges and the like because the regular Bayer filter array gives them the data to do so. With a pseudo random CFA these possibilities will diminish even more. Look at the complexity in the current x-trans demosaicing which even though a lot of the drawbacks of traditional bayer filters still apply (because of the highly regular nature of the CFA) not enough remains to apply the image enhancing algorithms. If
...Show more
Science progresses all the time and the newer algorithms proposed for demosaicing in the literature for quasi-random CFAs seem to be comparable in output quality. They are applied to all kinds of CFAs (including Bayer and compared to well-known algorithms) and don't produce erratic results. Right now I can't see why you shouldn't be able to use such a sensor and include image enhancements if required. Whether it makes sense financially is an other issue.



Sep 18, 2017 at 08:58 AM
charlyw
Offline
• • •
Upload & Sell: Off
p.11 #17 · p.11 #17 · Article posted on Dpreview re: Canons lack of innovation


Holger wrote:
Science progresses all the time and the newer algorithms proposed for demosaicing in the literature for quasi-random CFAs seem to be comparable in output quality. They are applied to all kinds of CFAs (including Bayer and compared to well-known algorithms) and don't produce erratic results. Right now I can't see why you shouldn't be able to use such a sensor and include image enhancements if required. Whether it makes sense financially is an other issue.


These sensor types are a solution to a non existent problem. Only Fuji chose to make it an issue by ineptly deciding on a worse CFA (again)... The problem will go away soon enough as resolutions rise.

The only thing people will have to say goodbye to is excessive cropping which is advocated by idiots every time the resolution rises...




Sep 18, 2017 at 09:09 AM
gdanmitchell
Offline
• • • • • • •
Upload & Sell: Off
p.11 #18 · p.11 #18 · Article posted on Dpreview re: Canons lack of innovation


charlyw wrote:
[Indeed there are plenty of examples and all show the flaws very clearly. None of them do point out any benefits. I have read a lot of these publications and I am versed in the art of demosaicing, being a software developer myself. x-trans is a chimera with aspirations that don't work out!


Two basic observations/comments.

1. As a photographer who has used various formats, cameras with and without AA filtering, and the x-trans system... I don't think that the x-trans approach really produces the advantages that Fujifilm hoped/claimed. I can produce excellent images from my x-trans files, but I can produce equally excellent images from non-x-trans files from cameras w/o AA filtering. (To be clear, I'm not dissing Fujifilm. I like their cameras a lot and I rely on them — and they produce excellent prints. I'm just saying that, for me, x-trans turned out to be an interesting theory but not an interesting reality. Note that they did not carry it forward to their miniMF camera.)

2. While "engineering" and others kinds of technical analysis of photographic technology are often interesting and useful — I engage in them, too — in the end what really matters is looking at photographs, preferably in the form of high quality prints. Here I find that quite a bit of what generates very heated debates on photographic forums like this one turns out to be pretty much irrelevant in photographs. For example, after shooting many tens of thousands of frames with several AA-filterless cameras, I see no issue at all with aliasing — in fact, I saw more examples with older cameras that did have AA filtering!



Sep 18, 2017 at 09:43 AM
charlyw
Offline
• • •
Upload & Sell: Off
p.11 #19 · p.11 #19 · Article posted on Dpreview re: Canons lack of innovation


gdanmitchell wrote:
I see no issue at all with aliasing — in fact, I saw more examples with older cameras that did have AA filtering!


OTOH I see problems with aliasing in loads of photographs from others - because I am usually looking at macro shots which are taken under extreme care to yield as sharp an image as possible, optimum aperture, tripod, subject plant clamped for more stability and so forth. And there the problems with aliasing are visible even in the reduced size images shown by those that use cameras without AA filtering. Those that do use cameras with aliasing normally even on average present the sharper and more natural looking images - because they don't need to apply as much processing to get rid of the moiree in the eyes of insects or their wings, they don't need to process the spider webs to get rid of colored edges... AA filter removal for me only means: Heaps more work to process the images!



Sep 18, 2017 at 10:07 AM
gdanmitchell
Offline
• • • • • • •
Upload & Sell: Off
p.11 #20 · p.11 #20 · Article posted on Dpreview re: Canons lack of innovation


charlyw wrote:
OTOH I see problems with aliasing in loads of photographs from others - because I am usually looking at macro shots which are taken under extreme care to yield as sharp an image as possible, optimum aperture, tripod, subject plant clamped for more stability and so forth. And there the problems with aliasing are visible even in the reduced size images shown by those that use cameras without AA filtering. Those that do use cameras with aliasing normally even on average present the sharper and more natural looking images - because they don't need to apply as much processing to
...Show more

Aliasing can occur with essentially any camera using digital sampling methods — it is essentially a compromise between keeping the data you usually want in your image versus eliminating high frequency data near the half-sampling rate frequency.

Now, I can't argue that there could be specialized types of photography — for example, you cite insect photography — in which aliasing might be more of an issue and in which camera AA filtering might be the better solution. I don't do that kind of photography, so I can't say.

Most of us are more likely to see it when, for example, the image contains high frequency lines. I've encountered it when photographing buildings with corrugated walls or certain roof patterns. However, in my experience...

- I encountered my most problematic examples of this on photographs made with cameras that had AA filtering! (That matches your experience described above.)

- Quite often the aliasing isn't actually produced in-camera, but is rather introduced in post when downsizing for web display.

- While I've seen the test examples of aliasing (for example, shots of certain kinds of test targets), in my actual photography I virtually never encounter it. On the extraordinarily rare occasions when I have — with AA filtering and non-AA-filtering cameras — I have always found a solution in post.

- Before I got my 5DsR I experimented with some files containing highly subjects that should have produced aliasing effects. They didn't.

- Note that virtually all (or is it now simply all?) digital MF systems have eliminated the AA filters.

None of this is meant to suggest that aliasing cannot happen with AA-free cameras. But one would think that if it were actually a major issue — or even a step backwards from AA-filtering cameras — that lots of photographers, especially the often-picky MF photographers — would be raising a stink.

My own sense, based on using the 5DsR mostly, is that the ultimate image quality difference between cameras applying AA filtering and otherwise-identical cameras that don't (5Ds), once the image has been taken through optimized post-processing, is virtually nil. Let's say that the 5Ds is a tiny bit softer than the 5DsR — in the end the photographer is likely to apply different and probably slightly greater magnitude sharpening to the 5Ds file than to the 5DsR file, producing prints that are so similar that one probably can't tell which is which.

In any case, it certainly seems that AA filtering is on the way out.



Sep 18, 2017 at 10:41 AM
1       2       3              10      
11
       12       13       14       end




FM Forums | Canon Forum | Join Upload & Sell

1       2       3              10      
11
       12       13       14       end
    
 

You are not logged in. Login or Register

Username       Or Reset password



This site is protected by reCAPTCHA and the Google Privacy Policy and Terms of Service apply.