Pages: 1 2 3 [4] 5 6 ... 11   Go Down

Author Topic: Canon EOS 5Ds and 5DsR - 50.3MP with and without AA filters  (Read 78364 times)

Phil Indeblanc

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 2017
Re: Canon EOS 5Ds and 5DsR - 50.3MP with and without AA filters
« Reply #60 on: February 02, 2015, 01:18:21 am »

I wish Canon worked with Foveon and gave that thing a boost! Sad that tech isn't being pushed hard to at least 30-40mp. They claim it now, but I heard from folks who use it it is a far from the claim.
Maybe if Canon worked with it, they could unlock some limits?
Logged
If you buy a camera, you're a photographer...

ErikKaffehr

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 11311
    • Echophoto
Re: Canon EOS 5Ds and 5DsR - 50.3MP with and without AA filters
« Reply #61 on: February 02, 2015, 01:44:49 am »

Hi,

I guess it would be priced as a 5D camera and not like a 1D camera, but in all probability more expensive than the 5DIII. Than of course there is the subtle question of competition.

If Nikon introduces 50MP in a D4s at D4 prices Canon can keep price up, if Nikon will have a D900, the competition landscape may be different,

Best regards
Erik

I heard it would be $8K! I hope not.
If it is a out-house sensor, wouldn't the price also be closer to the numbers we are used to..sure a bit higher as it is 50 vs 36mp, but I don't think they should pass roughly $4K price point(?)
The D810 is under $3k now. How much a difference is the extra mp going to make in price? Given the pixel quality is equal.
Logged
Erik Kaffehr
 

torger

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 3267
Re: Canon EOS 5Ds and 5DsR - 50.3MP with and without AA filters
« Reply #62 on: February 02, 2015, 06:18:04 am »

Yes I've tried to debate that it's better with AA filter, but 99% of photographers think it's a feature, end of discussion. They don't know about false detail and false colors around small details, and only look for large surfaces of colorful moiré, which indeed is quite rare. I'm quite sure the same thing as with the D800 will happen, the first model will come with both configurations, but the AA-filter version will sell so poorly that the followup will come in only one model.

When you know a bit too much about demosaicing and how artifacts look being without AA filter does not come up as a good proposition.

However, this sensor will through better area coverage with microlenses and relatively fuzzier lenses alias less than my Hassy... it shall be interesting to see some A/B comparisons between the two models by someone who knows what they're doing. AA filters are weak to not take away too much detail so they don't solve the aliasing problem completely either, so in the end one might end up with the conclusion that you could just go with the R version anyway.

For my Hassy I know it certainly wouldn't hurt with an AA filter, on the other hand as a software developer and hobbyist I find the problem of rendering from an aliased source as good as possible an interesting problem. If you're using RawTherapee you should use DCB demosaicer instead of Amaze, the latter is a tiny bit better at detail and a little bit more robust but DCB is better at hiding false color aliasing artifacts.
« Last Edit: February 02, 2015, 06:20:21 am by torger »
Logged

hjulenissen

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 2051
Re: Canon EOS 5Ds and 5DsR - 50.3MP with and without AA filters
« Reply #63 on: February 02, 2015, 07:58:55 am »

Yes I've tried to debate that it's better with AA filter, but 99% of photographers think it's a feature, end of discussion. They don't know about false detail and false colors around small details, and only look for large surfaces of colorful moiré, which indeed is quite rare.
Some rumors suggest that this camera will have more color-selective CFA, in line with the landscape/studio/... pitch.

I wonder if more selective filters will exaggerate the problems with having no AA filter. To some extent, if the color filters are very wide/overlapping, this could introduce some "bleeding" prefiltering?

On the other hand, as long as sensel density develop faster than lens quality (or photographer technique), we might reach a point where AA filters are not needed for most of us.

I believe that the value of no AA filtering is highest when proper deconvolution cannot recover sharp pixel transitions (i.e. low light), and when color is of little importance. So is it the natural-light ISO12800 B&W photographers who crave no AA the most?

-h
Logged

dwswager

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 1375
Re: Canon EOS 5Ds and 5DsR - 50.3MP with and without AA filters
« Reply #64 on: February 02, 2015, 09:44:57 am »

Hi,

I guess it would be priced as a 5D camera and not like a 1D camera, but in all probability more expensive than the 5DIII. Than of course there is the subtle question of competition.

If Nikon introduces 50MP in a D4s at D4 prices Canon can keep price up, if Nikon will have a D900, the competition landscape may be different,

Best regards
Erik


I wouldn't expect the D5 to be no more than 24MP, 36MP the absolute max.  Basically 10fps would be the cutoff and I just don't see them able to process and push 50MPs worth of data around fast enough to support 10fps at 50MP.  There should and I would expect to see a higher MP variant.

Me, I'd have a D5, D5s (sports) and a D5L (landscape) and leave the D810 area alone for awhile.  But Nikon's marketing is so bizarre who knows?  Because I would have the D810, D750 and D610 and the D7x00 when it appears all in the pro style body and leave the Dial interface strictly for the consumer market.
Logged

Phil Indeblanc

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 2017
Re: Canon EOS 5Ds and 5DsR - 50.3MP with and without AA filters
« Reply #65 on: February 02, 2015, 12:02:34 pm »

I've tested AA vs non-AA and there is a pretty significant difference. color bleed/contamination or not.
The files are much sharper, and have a more 3D quality to them. Even from the DSLR, they are more MFDb looking. Sure if you shoot lots of fashion clothing and fabrics close up or not, you will get moire.
But for many if not most, at least in my experience, I have rarely come across the issue using a MFDb without a AA, as most are, nor a DSLR without a AA.

I wonder if they could ever make this a filter like addition over the sensor, or  maybe even over the lens for those that really want to address the issues?

Anyway, I'm very happy to see this. What size are the photons on a FF 50mp sensor? (Wow)
Logged
If you buy a camera, you're a photographer...

NancyP

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 2513
Re: Canon EOS 5Ds and 5DsR - 50.3MP with and without AA filters
« Reply #66 on: February 02, 2015, 12:51:27 pm »

I doubt that it is a Foveon sensor, but certainly a top ISO of 6400 (or less) would fit a Foveon sensor - the Sigma iterations are good to 400 for best color.
Logged

jduncan

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 434
Re: Canon EOS 5Ds and 5DsR - 50.3MP with and without AA filters
« Reply #67 on: February 02, 2015, 04:41:09 pm »

I wouldn't expect the D5 to be no more than 24MP, 36MP the absolute max.  Basically 10fps would be the cutoff and I just don't see them able to process and push 50MPs worth of data around fast enough to support 10fps at 50MP.  There should and I would expect to see a higher MP variant.

Me, I'd have a D5, D5s (sports) and a D5L (landscape) and leave the D810 area alone for awhile.  But Nikon's marketing is so bizarre who knows?  Because I would have the D810, D750 and D610 and the D7x00 when it appears all in the pro style body and leave the Dial interface strictly for the consumer market.

I believe what the poster mean is a successor  for the D3x not for the  D4s. So a D4x @5fps will be have a high price, and Canon could base the price the camera accordingly.
As I nikon user it continues to annoy me that  Nikon still on the same  51 with  15 cross type autofocus system that they had on the original D3. So the Canon auto focus will make it very competitive.

I am surprised that the rumor does not talk about 4K. Canon does not have the grip they use to have with videographers so I will think that they will  make a jump forward.  This is particularly telling since the 1Dc dropped in price.

Best regards
Logged
english is not my first language, an I k

eronald

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 6642
    • My gallery on Instagram
Re: Canon EOS 5Ds and 5DsR - 50.3MP with and without AA filters
« Reply #68 on: February 02, 2015, 11:51:19 pm »

Torger,

 I would have thought that if one wishes to physically filter rb one destroys a lot more of the useful data in the higher resolving g.
 Also the spatial frequencies in the diagonal direction are dilated by another factor of sqrt(2) or so, which means one would need to filter ... even more brutally.
 And even if one does filter brutally there is still a differential effect on rb vs g due to the slope of the filter modulus, leading to color aliasing  ...
 This is purely a construct of my mind of course, feel free to set me right.

Edmund

Yes I've tried to debate that it's better with AA filter, but 99% of photographers think it's a feature, end of discussion. They don't know about false detail and false colors around small details, and only look for large surfaces of colorful moiré, which indeed is quite rare. I'm quite sure the same thing as with the D800 will happen, the first model will come with both configurations, but the AA-filter version will sell so poorly that the followup will come in only one model.

When you know a bit too much about demosaicing and how artifacts look being without AA filter does not come up as a good proposition.

However, this sensor will through better area coverage with microlenses and relatively fuzzier lenses alias less than my Hassy... it shall be interesting to see some A/B comparisons between the two models by someone who knows what they're doing. AA filters are weak to not take away too much detail so they don't solve the aliasing problem completely either, so in the end one might end up with the conclusion that you could just go with the R version anyway.

For my Hassy I know it certainly wouldn't hurt with an AA filter, on the other hand as a software developer and hobbyist I find the problem of rendering from an aliased source as good as possible an interesting problem. If you're using RawTherapee you should use DCB demosaicer instead of Amaze, the latter is a tiny bit better at detail and a little bit more robust but DCB is better at hiding false color aliasing artifacts.
Logged
If you appreciate my blog posts help me by following on https://instagram.com/edmundronald

torger

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 3267
Re: Canon EOS 5Ds and 5DsR - 50.3MP with and without AA filters
« Reply #69 on: February 03, 2015, 01:26:53 am »

No you're right I suppose, AA-filtering in cameras is not about solving the problem, it's about reducing it to sufficient level without losing too much resolution. It is possible to get moire showing with a AA-filtered camera, it's just a whole lot harder.

As you don't solve the problem completely you could always argue that well, why not skip it all-together? That will depend on how aliased it will be in practice, as said earlier it will probably be less aliased than my Hassy back with tech lenses as the 5DsR has microlenses and less sharp lenses (but on the other hand you more often shoot it at apertures f/1.4 and such so diffraction won't save you), but I don't think it will be on a negligible level... we just have to wait and see I guess...

Torger,

 I would have thought that if one wishes to physically filter rb one destroys a lot more of the useful data in the higher resolving g.
 Also the spatial frequencies in the diagonal direction are dilated by another factor of sqrt(2) or so, which means one would need to filter ... even more brutally.
 And even if one does filter brutally there is still a differential effect on rb vs g due to the slope of the filter modulus, leading to color aliasing  ...
 This is purely a construct of my mind of course, feel free to set me right.

Edmund

Logged

Phil Indeblanc

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 2017
Re: Canon EOS 5Ds and 5DsR - 50.3MP with and without AA filters
« Reply #70 on: February 03, 2015, 01:51:57 am »

First I bought this...Then I read this..... :-\

(Does anyone remember the Circuit City ad) :-)
http://www.sonyalpharumors.com/cp-camera-rumors-so-far-a7000-a7rii-a3xxx-and-a99ii/

from site...
Quote
I even got info via Tamron sources about a new A-mount Full Frame camera with close to 50MP
« Last Edit: February 03, 2015, 01:53:28 am by Phil Indeblanc »
Logged
If you buy a camera, you're a photographer...

davidgp

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 758
    • davidgp fotografia
Re: Canon EOS 5Ds and 5DsR - 50.3MP with and without AA filters
« Reply #71 on: February 03, 2015, 02:20:47 am »

I doubt that it is a Foveon sensor, but certainly a top ISO of 6400 (or less) would fit a Foveon sensor - the Sigma iterations are good to 400 for best color.

Maybe I'm mistaken here, but AA filters are there because moire is a problem in bayern type sensors, right? Foveon is not a bayern type sensor, so it will make not too much sense here to have a version of the camera with an AA filter on it...

Phil Indeblanc

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 2017
Re: Canon EOS 5Ds and 5DsR - 50.3MP with and without AA filters
« Reply #72 on: February 03, 2015, 03:00:50 am »

I think people are just wishing for Foveon. It would the ultimate in what DSLR's have to offer for now. With a 50mp, and a bit later Sony will have its 50mp, and there is still great things to gain from it, but not like what Foveon at a true 50mp size could do.

Was the Foveon limited in funding to further develop or was it a technical physics stop? Sorry, I didn't read all the technical back and forth, so perhaps it was mentioned. Maybe a simple answer is out there :-)
Logged
If you buy a camera, you're a photographer...

ErikKaffehr

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 11311
    • Echophoto
Re: Canon EOS 5Ds and 5DsR - 50.3MP with and without AA filters
« Reply #73 on: February 03, 2015, 04:00:05 am »

Hi,

Color moiré is a bayer issue but moiré in general is a sampling problem.

Color moiré is usually much more obvious than the monochrome version.

Best regards
Erik



Maybe I'm mistaken here, but AA filters are there because moire is a problem in bayern type sensors, right? Foveon is not a bayern type sensor, so it will make not too much sense here to have a version of the camera with an AA filter on it...
Logged
Erik Kaffehr
 

torger

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 3267
Re: Canon EOS 5Ds and 5DsR - 50.3MP with and without AA filters
« Reply #74 on: February 03, 2015, 04:29:15 am »

Personally I don't believe in Foveon. With current technology I rather have a higher density AA-filtered bayer sensor, better quantum efficiency it seems (and higher full well capacity), and better color filters as you can't really design color filters as precisely with Foveon. That is if you back out from pixel peep global color would be better on a bayer with color fidelity-optimized CFA, dynamic range is higher and and overall noise is lower.
« Last Edit: February 03, 2015, 04:40:11 am by torger »
Logged

Guillermo Luijk

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 2005
    • http://www.guillermoluijk.com
Re: Re: Re: Canon EOS 5Ds and 5DsR - 50.3MP with and without AA filters
« Reply #75 on: February 03, 2015, 05:04:33 am »

Maybe I'm mistaken here, but AA filters are there because moire is a problem in bayern type sensors, right? Foveon is not a bayern type sensor, so it will make not too much sense here to have a version of the camera with an AA filter on it...
Yes you are mistaken. Foveon sensors can produce moiré as a result of aliasing if proper AA filtering is not done  It will be a 'monochrome' moiré though, much less visible since there is no colour interpolation. This makes eliminating the AA filter less risky on Foveon sensors.

Bart_van_der_Wolf

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 8913
Re: Canon EOS 5Ds and 5DsR - 50.3MP with and without AA filters
« Reply #76 on: February 03, 2015, 06:39:43 am »

Was the Foveon limited in funding to further develop or was it a technical physics stop?

While the Foveon is a quite an interesting technical achievement, it is not without some serious problems for high quality imaging. One of the major obstacles is that the color separation by penetration depth in silicon, delivers an almost monochrome rendering of the scene. It requires huge color separation factors to split the data into R/G/B planes. That separation process comes with an increase of noise, and the fill factor of the sensels is already low to begin with. Another problem is that penetration depth of silicon varies with the angle of incidence, so larger sensors will require even more heavy lifting in postprocessing to somewhat mitigate the adverse color pollution cause by oblique rays.

Besides that, as mentioned, without AA-filter the Foveon will also suffer from luminance moiré. Because R/G/B is sampled at approx. the same position, false color aliasing artifacts are less of an issue.

Also, sampling 3 color planes per sensel will also triple the uncompressed file size compared to Bayer CFA Raw data. That will require more processing power in camera (and more battery capacity), more storage space, and takes longer to record (lower number of images per second). It will also be slow in Raw conversion, and there may be little Raw converter support other than from the manufacturer themselves.

Canon have patents for quite different technology that also allows to capture R/G/B per sensel, but they are not likely to use that any time soon (complex/expensive to produce and proprietary Raw conversion needed), if ever. The Bayer CFA still has lots of potential, and it seems (if the rumors are true) that accurate color is the preferred approach for the 5Ds (s is for studio). That will be possible if more selective band-pass filters are used (hence the modest ISO claims), the opposite of the hugely overlapping Foveon color planes. Maybe a future addition of a 4th color (e.g. Yellow) in the filter array can help to get even better color, especially if sensel pitch keeps shrinking.

Cheers,
Bart
Logged
== If you do what you did, you'll get what you got. ==

Chris Livsey

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 807
Re: Canon EOS 5Ds and 5DsR - 50.3MP with and without AA filters
« Reply #77 on: February 03, 2015, 07:47:55 am »

One of the major obstacles is that the color separation by penetration depth in silicon, delivers an almost monochrome rendering of the scene. It requires huge color separation factors to split the data into R/G/B planes. That separation process comes with an increase of noise, and the fill factor of the sensels is already low to begin with. Another problem is that penetration depth of silicon varies with the angle of incidence, so larger sensors will require even more heavy lifting in postprocessing to somewhat mitigate the adverse color pollution cause by oblique rays.


Interesting that of course this models the coating of colour negative film:
Assume a 100 iso film. Then the emulsions have to be far faster than 100 to achieve this final speed. The absorber dyes and antihalation are going to eat up speed, and third, turbidity of couplers and overlying layers are going to eat up speed due to internal reflection and absorption of light.
Assume that the blue layer, on top, has to be 200 speed, the green layer coming second, has to be 300, and the red layer on the bottom has to be 400 to account for these losses. This means 3 different emulsions with graded sensitivity, grain size, halide content and etc. And, being on the bottom, the cyan (red) layer will be the slowest to develop due to grain size, halide ratio, and diffusion effects.
No wonder the chip makers are struggling.


Logged

davidgp

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 758
    • davidgp fotografia
Re: Canon EOS 5Ds and 5DsR - 50.3MP with and without AA filters
« Reply #78 on: February 03, 2015, 10:53:42 am »

Many thanks for the clarification Erik, I was only thinking in color moiré, but right, sampling can produce moiré...

Hi,

Color moiré is a bayer issue but moiré in general is a sampling problem.

Color moiré is usually much more obvious than the monochrome version.

Best regards
Erik



davidgp

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 758
    • davidgp fotografia
Re: Re: Re: Canon EOS 5Ds and 5DsR - 50.3MP with and without AA filters
« Reply #79 on: February 03, 2015, 10:55:48 am »

Yes you are mistaken. Foveon sensors can produce moiré as a result of aliasing if proper AA filtering is not done  It will be a 'monochrome' moiré though, much less visible since there is no colour interpolation. This makes eliminating the AA filter less risky on Foveon sensors.

Yes, as I commented to Erik, I should have thought of monochrome moire due to sampling... not only color moiré due to the bayern or somilar pattern...  my bad...
Pages: 1 2 3 [4] 5 6 ... 11   Go Up