why buying Canon 5DS instead of 5DS R

[Victor Engel]

[...]

You cannot get more than the mp value from a formula which is designed to give a lesser value.

Lesser value than what?

You keep talking about different things. Of course, that formula is hidden but (as they say) is based on MTF values. Those values (contrast vs. resolution) are less than 100%, all the time, by definition.

Exactly. The formula is hidden - if there even is a formula, so this discussion is rather pointless.
--
Victor Engel

 

[John W Hall]

Loading…

www.dpreview.com

 

[J A C S]

[...]

You cannot get more than the mp value from a formula which is designed to give a lesser value.

Lesser value than what?

Than the mp count.

You keep talking about different things. Of course, that formula is hidden but (as they say) is based on MTF values. Those values (contrast vs. resolution) are less than 100%, all the time, by definition.

Exactly. The formula is hidden - if there even is a formula, so this discussion is rather pointless.

The description posted earlier makes it clear that it is designed the way I suspect. They do say that it is a weighted MTF average. That poses are natural limit.

 

[Mako2011]

The p-mpix value stands for perceptual pixel count that was probably created in order to put diverse sensor designs on Foveon and Bayer on an equal footing. It's true, it can max out at the number of pixels if there is a one-to-one relationship between the pixel and an area of the subject. I'm suggesting that it's possible to create a camera that uses a single pixel to image more than one area of the subject. In such a case, the p-mpix value exceeds the pixel count.

I don't think that would be the case...as then p-Mpix would have to be redefined/adjusted.

Well, what you gave in your other reply to me is not the definition.

Doesn't matter if to you it's not the definition...it is the definition provided by DxO though. Since it's their number then really, their definition is all that counts. They used to even publish some of the formula specific but no longer do.

I'd characterize it more as an editorial describing implications. I don't know if there is a formal definition, but if you look it up on the dxo website, they state, "P-Mpix is the unit of a sharpness measurement. The number of P-Mpix of a camera/lens combination is equal to the pixel count of a sensor that would give the same sharpness if tested with a perfect theoretical optics, as the camera/lens combination under test."

Exactly...so p-Mpix, by their own definition, can not exceed the camera being used megapixel count. also note: "...Sharpness is expressed in PMpix and is typically between 50% and 100% of the sensor pixel count" A few cases were it fall below 50% but no cases were it can exceed 100%.

I see nothing in that definition that precludes my putative camera design.
--
Victor Engel

 

[Steen Bay]

Right. And p-mpix cannot be higher than the pixel count.

Sure it can (but not with current cameras). Our brains (and I'm including the eyes here) do this sort of thing all the time. What's needed is an additional dimension. The most obvious additional dimension is time. Add a piezoelectric device, or even just a jiggling motor, to the imaging sensor and some fancy circuitry to integrate over time, and you can get more resolution than the pixel density would allow for a static shot.

Of course, then you need more image memory, or a good modeling algorithm to turn the raw data into a good visual model.

You are talking about something different. DXO hides the formula but whatever it is, it is derived from MTF (as they say) and I am pretty sure that it cannot exceed the pixel count.

No. I think it would qualify. I will point out that the same idea has already been implemented (I think)

Yes, on E-M5 II :

http://www.dpreview.com/previews/olympus-om-d-e-m5-ii/4

HR image available in DPR's comparison tool.

and there is a patent http://www.google.com/patents/US5063450 to make the birefringent layers superfluous. The camera need only to expand on this to integrate the image over time.

The p-mpix value stands for perceptual pixel count that was probably created in order to put diverse sensor designs on Foveon and Bayer on an equal footing. It's true, it can max out at the number of pixels if there is a one-to-one relationship between the pixel and an area of the subject. I'm suggesting that it's possible to create a camera that uses a single pixel to image more than one area of the subject. In such a case, the p-mpix value exceeds the pixel count.

In fact, such a camera does exist. The one I have in mind I think was used to create the animations here:
http://www.nps.gov/band/historyculture/animation.htm
It's a one pixel camera. Using laser ranging with rasters one pixel wide, a three-dimensional model was created. Yes, I know it's more complex than that, but it makes my point.

The p-mpix value is intended to be a resolution resulting from a combination of camera and lens. All that's needed is for the camera to do all the extra stuff I'm describing.
--
Victor Engel

 

[gavin]

I think Nikon also cancel the AA filter if I remember correctly but its been a while since I read about the 800s. Normally I think they have one AA filter go one direction and another one go 90deg. They are putting the second one the reverse of the first one to cancel out the AA. However I am only about 50% sure since this is from memory

 

[Victor Engel]

Yes, on E-M5 II :

http://www.dpreview.com/previews/olympus-om-d-e-m5-ii/4

HR image available in DPR's comparison tool.

Thanks. That's not what I had in mind, but it fits right in to the discussion. Just out of curiosity, what is' DXO's score for that high res. mode for that camera with a high quality lens? I couldn't find such a combination on their website. I don't use their data because I think their reporting is bogus, mostly reporting results in higher precision than the data supports.
--
Victor Engel

 

[Steen Bay]

Yes, on E-M5 II :

http://www.dpreview.com/previews/olympus-om-d-e-m5-ii/4

HR image available in DPR's comparison tool.

Thanks. That's not what I had in mind, but it fits right in to the discussion. Just out of curiosity, what is' DXO's score for that high res. mode for that camera with a high quality lens? I couldn't find such a combination on their website. I don't use their data because I think their reporting is bogus, mostly reporting results in higher precision than the data supports.
--
Victor Engel

E-M5 II was just announced recently, don't think it's available yet. Could be interesting if DxO tested it in HR mode (64mp RAW), but don't think it'll happen (no Foveon and X-Trans tests either).

 

[Jeff Fenske]

Looks like the D810 is a serious step backwards from the D800 in terms of phoney details etc. then. Looking at some of the bird photos I was quite appalled by the moiré. All the more reason to snap up a 5Ds when it becomes available before Canon goes for the same daft idea and drops the AA filter in the next iteration. I totally agree with you that at least for the moment obsessive "sharpness freaks" have won the argument. I seem to see excessively sharpened images everywhere.

Wisely said!:

Snap up a 5Ds when it becomes available before Canon goes for the same daft idea and drops the AA filter in the next iteration.

– –

I have come to a similar conclusion, based upon these reasons I shared in another thread http://www.dpreview.com/forums/thread/3801081:

The "unrealistic 'crunchy' appearance" theory presented in this thread is compelling and appreciated, and may be widely known as absolute fact over time.

The two clinchers for me:

(1) The 5Ds R is compromised from birth. Canon decided to use the inferior method of keeping the unused AA filter in the light path, degrading the resolution and dimming light entering the camera. Nikon discarded this method with the D810.

I like the idea of light being able to freely flow into the camera without interference. Putting things in the way that don't do anything is nuts for a high end camera! But it was pointed out that Canon probably didn't know during development that Nikon was going to come out with the freedom edition, AAless D810. But maybe the jaggies and moiré isn't worth the less interfered light transmission, anyway.

There is a strong likelihood that Canon will release a completely AAless 5Ds version in the future, similar to the Nikon D810, unless AAless gets discredited as being undesirable in the meantime.

(2) And DPR was able to duplicate a D800e shot by adding sharpening to the D800 AA filtered version in the green fence photo here: http://www.dpreview.com/reviews/nikon-d800-d800e/27

The buildings photo didn't do as well with sharpening (or oversharpening), but still…. DPR basically proves there is little 'sharpness' advantage through AA cancelation.

(3) A third, less important point for me is the peace of mind from not having to worry about seeing moiré and artifacts once I get home and look at the images more closely on the monitor.

I've read a lot of threads on this subject, including elsewhere, but this discussion was my eureka moment where the 5Ds R no longer makes practical sense to me.

 

[J A C S]

(1) The 5Ds R is compromised from birth. Canon decided to use the inferior method of keeping the unused AA filter in the light path, degrading the resolution and dimming light entering the camera. Nikon discarded this method with the D810.

If you believe that the AA filter should be there (as I do), then the canceled AA filter is better than no AA filter (i.e., more glass) in the first place.

 

[Jeff Fenske]

(1) The 5Ds R is compromised from birth. Canon decided to use the inferior method of keeping the unused AA filter in the light path, degrading the resolution and dimming light entering the camera. Nikon discarded this method with the D810.

If you believe that the AA filter should be there (as I do), then the canceled AA filter is better than no AA filter (i.e., more glass) in the first place.

There are two extra filters in the Canon 5Ds R and the now old Nikon D800e that don't do anything except decrease sharpening and light. I'd rather have nothing in the light path at all, like Nikon did with the D810.

This guy discovered that the D810 is significantly sharper in the mid-frame and edges than the D800e, and loses some sharpness in the center too: https://photographylife.com/is-nikon-d810-sharper-than-d800e

Nikon's promo video states "the D810 has no low-pass filter, and thus delivers sharper and better low-light image quality than even the D800e."Minute-2:15:

 

[J A C S]

(1) The 5Ds R is compromised from birth. Canon decided to use the inferior method of keeping the unused AA filter in the light path, degrading the resolution and dimming light entering the camera. Nikon discarded this method with the D810.

If you believe that the AA filter should be there (as I do), then the canceled AA filter is better than no AA filter (i.e., more glass) in the first place.

There are two extra filters in the Canon 5Ds R and the now old Nikon D800e that don't do anything except decrease sharpening and light.

There is no measurable loss of light by DXO, and sensorgen (based on DXO) actually estimates the D800E to be more efficient than the D810! Well, a different sensor and probably CFA.

Think about the two extra filters as a weak AA filter, which is good. My understanding is that the D810 has glass there, not empty space.

I'd rather have nothing in the light path at all, like Nikon did with the D810.

Not sure that it is nothing. That would affect the focusing, etc.

This guy discovered that the D810 is significantly sharper in the mid-frame and edges than the D800e, and loses some sharpness in the center too: https://photographylife.com/is-nikon-d810-sharper-than-d800e

And aliases more, too.

 

[Its RKM]

now, why canon wouldn't add a feature that photographer could turn off the AA filter if one desired?

They do!

They call it the "A value", or aperture to most of us.

With the 4um pixel the AA filter is pretty much redundant by f/16 and aliasing is near impossible with almost all Canon lenses.

So dial f/# up and down to control the amount of AA filter you desire. ;-)

 

[J A C S]

now, why canon wouldn't add a feature that photographer could turn off the AA filter if one desired?

They do!

They call it the "A value", or aperture to most of us.

With the 4um pixel the AA filter is pretty much redundant by f/16 and aliasing is near impossible with almost all Canon lenses.

Hmm..., see the chart here, same pixel density. And this is WITH an AA filter!

zoom in to the chart

 

[Jeff Fenske]

(1) The 5Ds R is compromised from birth. Canon decided to use the inferior method of keeping the unused AA filter in the light path, degrading the resolution and dimming light entering the camera. Nikon discarded this method with the D810.

If you believe that the AA filter should be there (as I do), then the canceled AA filter is better than no AA filter (i.e., more glass) in the first place.

There are two extra filters in the Canon 5Ds R and the now old Nikon D800e that don't do anything except decrease sharpening and light.

There is no measurable loss of light by DXO, and sensorgen (based on DXO) actually estimates the D800E to be more efficient than the D810! Well, a different sensor and probably CFA.

That's interesting. Maybe the D800e does have a weaker color filter array, as it rates 3 points poorer in the color metamerism index.

Think about the two extra filters as a weak AA filter, which is good. My understanding is that the D810 has glass there, not empty space.

Maybe they can completely undo the affect of the first AA with the second. But if it does work as a very weak AA that would be better than doing nothing but getting in the way of the light path.

Everything I've heard is that both AAs were completely removed in the D810.

The reason that Canon did the cancellation route seems to be to cut costs only. And they probably didn't know Nikon would come out with the D810, until it was too late.

I'd rather have nothing in the light path at all, like Nikon did with the D810.

Not sure that it is nothing. That would affect the focusing, etc.

This guy discovered that the D810 is significantly sharper in the mid-frame and edges than the D800e, and loses some sharpness in the center too: https://photographylife.com/is-nikon-d810-sharper-than-d800e

The light would probably be strongest in the center, so there is less degradation and light falloff there. That just seems to make sense to me. I don't have any expertise in this area. Just trying to find out what's going on.

Nikon totally removed the filters for a reason. And as I point out above, Nikon says the D810 is sharper too.

And aliases more, too.

Right now, I'm strongly leaning toward the idea that a weak AA is the best way to go. The 5Ds will probably have a weak AA, whereas the 5D3 probably has a strong AA, since Canon seemed to tweak it for best video instead of best still imagery.

I'm glad Canon considers the 5Ds to have too many MP for great video, so they are finally making the best camera they can for still photography at low ISO.

 

[Its RKM]

It's a valid point, however in this case, it's the customers who made the choice for Nikon. Most 36 mp users does not want AA filter.

They only think they don't. They're sharpness freaks that think they would rather have a sharp-looking image full of crap than a slightly softer image full of real data.

Nah, they're dynamic range freaks - that other parameter that Canon can't deliver that every non-Canon shooter find useful. ;-)

Really guys, at 50Mp the AA-no-AA filter argument is only valid with the sharpest lenses at their aperture sweet spot. Stop down (diffraction) or up (aberrations) and it disappears - and the loss is no more than having an AA filter in the first place.

SLR's are about flexibility. If you just want to shoot like a robot then buy an iPhone!

 

[Victor Engel]

Right now, I'm strongly leaning toward the idea that a weak AA is the best way to go. The 5Ds will probably have a weak AA, whereas the 5D3 probably has a strong AA, since Canon seemed to tweak it for best video instead of best still imagery.

That's probably more to do with using the full data, rather than every third raster like the 5D Mark II uses.

 

[Lee Jay]

now, why canon wouldn't add a feature that photographer could turn off the AA filter if one desired?

They do!

They call it the "A value", or aperture to most of us.

With the 4um pixel the AA filter is pretty much redundant by f/16 and aliasing is near impossible with almost all Canon lenses.

Hmm..., see the chart here, same pixel density. And this is WITH an AA filter!

But it's at f/2.8, not f/16.

 

[Lee Jay]

It's a valid point, however in this case, it's the customers who made the choice for Nikon. Most 36 mp users does not want AA filter.

They only think they don't. They're sharpness freaks that think they would rather have a sharp-looking image full of crap than a slightly softer image full of real data.

Nah, they're dynamic range freaks - that other parameter that Canon can't deliver that every non-Canon shooter find useful. ;-)

Really guys, at 50Mp the AA-no-AA filter argument is only valid with the sharpest lenses at their aperture sweet spot.

Not true.

For example, a lens like the 18-135STM can provide aliasing at this pixel density, if not for an AA filter.

 

[Jeff Fenske]

It's a valid point, however in this case, it's the customers who made the choice for Nikon. Most 36 mp users does not want AA filter.

They only think they don't. They're sharpness freaks that think they would rather have a sharp-looking image full of crap than a slightly softer image full of real data.

Nah, they're dynamic range freaks - that other parameter that Canon can't deliver that every non-Canon shooter find useful. ;-)

Really guys, at 50Mp the AA-no-AA filter argument is only valid with the sharpest lenses at their aperture sweet spot. Stop down (diffraction) or up (aberrations) and it disappears - and the loss is no more than having an AA filter in the first place.

This shows the affect of diffraction at different f/stops, comparing the D800 and D800e:

Loading…

www.dpreview.com

Have you seen this, how the AAless D800 sharpens up to be similar to the D800e:

http://www.dpreview.com/reviews/nikon-d800-d800e/27

Some theorize that the AAless sharpness results are jaggy and artifacty, and are not natural, similar to what sharpening does.

And if sharpening can achieve almost the same results, if this is what we're going for, the AA versions would be more versatile, and not suffer as much from moiré.

SLR's are about flexibility. If you just want to shoot like a robot then buy an iPhone!

--
Its RKM