5DSR II?

[Sporgon]

You all can keep guessing at what you think you would or wouldn't see and you're welcome to want to feel good about your purchase of the regular 5DS, but I've seen and used a camera with an AA filter directly next to the same model with the AA removed and the one without the filter is very noticeably clearer, with much better micro contrast and detail, and no moire even on a body with lower pixel density than the 5DS/R.

This is like when digital took over but some people kept holding on to film claiming it had more dynamic range and more accurate colour and digital was a fad. Like, no. That's why Kodak are dead.

So are you suggesting that Canon were completely wrong in producing the 5Ds alongside the 5DsR ?

 

[lexaclarke]

I didn't say that at all.

I think it probably made sense for them to do it at the time because it was their first attempt at a primary body without an AA filter. (As far as I'm aware.) It would be a bit nuts of them to make a fairly significant mechanical change, no matter how good, without giving people an option to keep with what they're used to. Like I said, some people clung on to film for years after it was obvious digital had surpassed it. Some people still swear their film-era lenses are better than modern ones. So there were always going to be some people who wouldn't want to buy a body with a feature like that which the Canon line hadn't had before. There will always be resistance and some people need to be eased into new ideas. So making the regular 5DS made sense for then.

But now people have had that chance to be eased in and other companies have ramped up their use of non-filtered sensors to the point where most new ILC bodies don't have AA filters, right? At least ones in the professional price range. So from this point, releasing pro cameras with an AA filter becomes less necessary and you will eventually get to a point where the amount of people buying the bodies with filters aren't enough to offset the cost of making both versions.

Nikon did this before with the D800, right? They made a regular one and the E model which has the AA filter reversed. And the D800E was way more popular and since then they've just stuck with not having a filter. I don't know about the latest Sonys (mark IIIs and 9) but the A7R definitely sold more than the regular A7, both the originals and the mark II versions.

Canon kind of did a similar thing before with the 1D bodies, if you remember when that line was split. They did the 1D and 1Ds. Then they added a 1Dn and the 1DC. All of them did things slightly differently so people could get the version that had the features they were most comfortable with and needed. But eventually they all got folded into the 1DX models because the people who used to think only resolution matter started feeling like speed did matter after all and the people who thought they only wanted speed decided they did want resolution and the bigger sensor and splitting the models up is more expensive than making one model which takes the best parts of all of them.

So I think making the 5DS and 5DS R made total sense as Canon's first try, but for the second version it will just be the 'R'. I think the 5DS, 1D and 7D cameras will dump the AA filters and the 5D, 6D and 90D will be the ones that keep the filter. So the 5DS (no filter) line is your solution for people who want maximum resolution and the 5D (filter) is the line for people who want a bit of everything. Eventually, like the 1D+1Ds=1DX, technology will get to the point where the 5DS and 5D lines can be combined back into one and we'll have what I guess we could call a 5DX, which would have no filter and high resolution and speed and great video, like how the 1DX does everything the 1D and 1Ds lines did combined.

 

[AlanF]

lexaclarke
+1 to your last two posts.

 

[AlanF]

And even then the 10% 'more' resolution that AlanF claim is moot; if you read the Lens Rentals article, the excellent Zeiss Otus only gave 5% more between the two cameras, and only in the centre of the frame. But also this is shooting a black and white test chart; if I used my camera to shoot black and white test charts I too would choose the 5DSR. ( If I hadn't jumped in front of a bus by then). When shooting in the real world with the Bayer Array method of computing colour I'm not at all convinced that you'll see this '5 - 10%' difference anyway. Certainly those that I know who don't have a personal agenda on the subject, when using the Pentax K5 vs K5s ( or what ever it was called) and the Nikon D800 / D800e and the 5Ds / 5DsR say that there is naff all difference even when printing at full sized output.

However, generally it seems that the sales pitch of "no blurring AA filter" means that the next 5Ds/r will probably not have the AA filter, aka Nikon, Pentax et al

https://www.lensrentals.com/blog/2015/06/canon-5ds-and-5ds-r-initial-resolution-tests/
It's not moot: the 300mm f/2.8 II has a measured better resolution on the 5DSR, and that is a fact that is not made moot because another lens acts differently. (The average value of the Otus was measured to be 8% higher on the 5DSR.) My own experience is with multi-coloured birds, not crows. You can stick with AA-filters, it's your choice, but I don't want them

 

[Frodo]

Just shot three weddings with my 5DsR. No moire. Anywhere.

 

[Sporgon]

And even then the 10% 'more' resolution that AlanF claim is moot; if you read the Lens Rentals article, the excellent Zeiss Otus only gave 5% more between the two cameras, and only in the centre of the frame. But also this is shooting a black and white test chart; if I used my camera to shoot black and white test charts I too would choose the 5DSR. ( If I hadn't jumped in front of a bus by then). When shooting in the real world with the Bayer Array method of computing colour I'm not at all convinced that you'll see this '5 - 10%' difference anyway. Certainly those that I know who don't have a personal agenda on the subject, when using the Pentax K5 vs K5s ( or what ever it was called) and the Nikon D800 / D800e and the 5Ds / 5DsR say that there is naff all difference even when printing at full sized output.

However, generally it seems that the sales pitch of "no blurring AA filter" means that the next 5Ds/r will probably not have the AA filter, aka Nikon, Pentax et al

https://www.lensrentals.com/blog/2015/06/canon-5ds-and-5ds-r-initial-resolution-tests/
It's not moot: the 300mm f/2.8 II has a measured better resolution on the 5DSR, and that is a fact that is not made moot because another lens acts differently. (The average value of the Otus was measured to be 8% higher on the 5DSR.) My own experience is with multi-coloured birds, not crows. You can stick with AA-filters, it's your choice, but I don't want them

Well it is because you are not always achieving 10% extra. As you began one of your posts; if......

You can stick with AA-filters, it's your choice, but I don't want them

Precisely. But stating that a 5Ds owner is missing out on 10% extra resolution in real world shooting is misleading. The SR blurs, then reverses the blur in the way it is set up. I prefer the option to reverse it in post.

If I put say 10 centre crops from identical images I have shot on the 5DsR and 5Ds in dropbox, with three of them being shot on the 5Ds but sharpness optimised in post, you should be able to clearly identify the three from the 5Ds. Care to have a go ?

 

[AlanF]

The SR blurs, then reverses the blur in the way it is set up. I prefer the option to reverse it in post.

If I put say 10 centre crops from identical images I have shot on the 5DsR and 5Ds in dropbox, with three of them being shot on the 5Ds but sharpness optimised in post, you should be able to clearly identify the three from the 5Ds. Care to have a go ?

You cannot reverse blurring in post. What sharpening does in post is to increase acutance (edge sharpness) which appears to the eye as apparent sharpness, it does not restore resolution that has been lost by blurring. Resolution is measured quantitatively by MTFs, which Lensrentals has done for the 5DSR and 5DS. If you don't believe their measurements, then so be it.

 

[privatebydesign]

The SR blurs, then reverses the blur in the way it is set up. I prefer the option to reverse it in post.

If I put say 10 centre crops from identical images I have shot on the 5DsR and 5Ds in dropbox, with three of them being shot on the 5Ds but sharpness optimised in post, you should be able to clearly identify the three from the 5Ds. Care to have a go ?

You cannot reverse blurring in post. What sharpening does in post is to increase acutance (edge sharpness) which appears to the eye as apparent sharpness, it does not restore resolution that has been lost by blurring. Resolution is measured quantitatively by MTFs, which Lensrentals has done for the 5DSR and 5DS. If you don't believe their measurements, then so be it.

Whilst I don't think anybody here would seriously question the measurements from Lensrentals it seems to me we fall into that trap between 'science' or bench testing, and 'real world' imaging. For example does the Lensrental result come from a demosaiced RAW file? If so how was it processed and if different results can be achieved by different processing settings is that not a valid point?

I can see where Sporgon is coming from, from an actual image point of view he personally, with his work, didn't see enough of a difference/improvement in sharpness to risk the moire (and it isn't that difficult to induce), and he is in good company because Kieth over at Northlight came to the same conclusion. Having said that it doesn't mean in a lab setting and with same (not optimal) processing settings then a difference can't be measured.

 

[AlanF]

Lensrentals used Imatest, documented in http://www.imatest.com/docs/sharpness/

The facts are:
1. An AA-filter reduces, but does not entirely eliminate, the occurrence of Moire.
2. The higher the pixel density (more Mpx) the less the importance of Moire.
3. The AA-filter works by blurring by spreading points of light over neighbouring pixels.
4. Sharpening will restore edge sharpness but not restore resolution that is lost.
5. The advantage of absence of an AA-filter will be seen only when there is detail at a fine level that would be lost by by blurring.

I take 25,000+ bird photos a year, and see Moire in only a tiny number of them, mostly with my 5DSR but once or twice with my 5DIV. As I take repeat shots to get the sharpest ones, the Moire in those rare cases may be eliminated in the slightly defocussed ones (it's very difficult to remove in post). You pays your money and takes your choice on AA-filters. To me, a gain of 10-15% in resolution is equivalent to my 400mm lens having the "reach" of a 440-460mm in the absence of a filter, which is a big plus for my bird photography.

 

[Sporgon]

The SR blurs, then reverses the blur in the way it is set up. I prefer the option to reverse it in post.

If I put say 10 centre crops from identical images I have shot on the 5DsR and 5Ds in dropbox, with three of them being shot on the 5Ds but sharpness optimised in post, you should be able to clearly identify the three from the 5Ds. Care to have a go ?

You cannot reverse blurring in post. What sharpening does in post is to increase acutance (edge sharpness) which appears to the eye as apparent sharpness, it does not restore resolution that has been lost by blurring. Resolution is measured quantitatively by MTFs, which Lensrentals has done for the 5DSR and 5DS. If you don't believe their measurements, then so be it.

Whilst I don't think anybody here would seriously question the measurements from Lensrentals it seems to me we fall into that trap between 'science' or bench testing, and 'real world' imaging. For example does the Lensrental result come from a demosaiced RAW file? If so how was it processed and if different results can be achieved by different processing settings is that not a valid point?

I can see where Sporgon is coming from, from an actual image point of view he personally, with his work, didn't see enough of a difference/improvement in sharpness to risk the moire (and it isn't that difficult to induce), and he is in good company because Kieth over at Northlight came to the same conclusion. Having said that it doesn't mean in a lab setting and with same (not optimal) processing settings then a difference can't be measured.

This is pretty much the nub of it, and I'm sure AlanF knows I'm not questioning Lens Rentals results, but I am questioning the broad brush application of these results to the multitude of ways in which these cameras, lenses and subject combinations are used in practice. As I stated at the beginning, I see no significant difference once the 5Ds has been processed - personally. (Even less than "my 135L has a unique signature than no other lens can reproduce" ;D ). My argument isn't really with AlanF's point of view anyway, after all Canon made the SR for him and the S for me; it's the likes of posters who are strongly pro the absence of an AA filter and presume that anyone who wants one does so from a position of ignorance. One such poster can be found in these recent pages

 

[AlanF]

Lively, but polite, discussion is what makes these threads interesting and useful. There are quite a few useful discussions going on at present.

 

[degos]

3. The AA-filter works by blurring by spreading points of light over neighbouring pixels.

Both AA-filtered and non-filtered sensors throw away an infinite amount of data, this including:

1. The pixels landing in spaces between photosites
2. The discrete values of all the pixels landing on each photosite, after being calculated to a single readout value

The non-filtered sensor is dishonest in that it pretends that each photocell knows all the data between itself and its neighbour and that the result was a hard transition between pixel readout values.

The filtered sensor basically says "dunno what was lost so I've graduated the transition to average it out".

Neither has more resolution.

The non-filtered sensor appears sharper because it insists there is a hard transition between neighbouring readout values. But it's a lie.

.

 

[lexaclarke]

That implies that the AA filter has some kind of intelligence behind it. It's not making up a transition between pixels, it's just dispersing the light less accurately.

Sensors with and without AA filters may technically have the same pixel count but the usable resolution you get out of them is different. Just like how putting a less sharp lens on the same body gives you less effective resolution. The pixel count may not change but how much perceivable detail there is absolutely does.

If we were talking about a lens, nobody would claim a softer lens is somehow capable of giving the same detail. If Canon made a lens with a special blurring element inside, everyone would call them crazy. But put a special blurring layer on a sensor and you want to say it's not just the same resolving power, but actually better?

Look, if you like AA filters, okay, cool, good for you I guess. We've had AA filters on cameras for years now and it's not stopped nice images being made. It's certainly good for lower resolutions and video. I wouldn't want to try to film 1080p video without an AA filter. But we're talking about 36mp, 42mp, even 50mp still images. I understand peoples' reluctance to move on in technology but I don't understand, like I literally don't see how it adds up, this idea that an AA filter is better.
AA filters were made as a band-aid to fix a problem. Resolution of digital sensors was too low to produce nice smooth gradients and it got confused by some really fine patterns. So we got weird pixelated curves and false colours. The crude solution to that was to just blur the whole sensor. Now we have sensors with high enough resolution to reproduce those gradients and fine details accurately. More accurately than film was ever able to do and we never had AA filters on film. The problem AA filters were invented to fix is no longer a problem.

I get that there are uses for sensors with AA filters. Like I said in a previous comment, I'm sure the 6D and 90D models will keep AA filters and the 1D, 5D and 7D will adopt them slowly. It's not going to be a total switch overnight. People will continue to have a choice for at least the next what, six years maybe? Maybe ten? But you can't tell me that AA filters are somehow better and more "honest" than a non-filtered sensor. I doubt you would claim that a sharp lens is more "dishonest" than a soft one.