DR and RAW bit depth on Canon's forthcoming...

[wickidwombat]

How do all the high end medium format digital backs get such good DR? do they read at a higher true bit rate though sampling before the AD converter?

[torger]

How do all the high end medium format digital backs get such good DR? do they read at a higher true bit rate though sampling before the AD converter?

The good DR of medium format is a bit of a myth. They do have 16 bit ADC, but that's meaningless since the noise levels are too high, 14 bit would yield the same result. I've also heard the exposure metering is different from DSLRs, underexposing more so you get more highlight headroom so you can push more in Lightroom which some mistakingly think means that there is more dynamic range.

And then there is what you compare with. If you compare with Canon 5D mark II, yes, there is better dynamic range, because the MF sensors are usually very nice random noise, not patterned noise like on the 5D. But compare with a Sony Exmor sensor and the MF sensor has a little bit less DR. All MF vs DSLR tests I've seen has been with Canon cameras, and they always lose. But they would lose against a Nikon or a Sony with a Sony Exmor sensor too.

Then there is the larger sensor area, there is some DR gain from that since you can gather more photons (less shot noise), but that gain is not very large.

Throughout the years many DSLRs have been released with less than perfect DR, while MF sensors has at least since 2003 had good DR. But that's only on base ISO of course, those CCDs are not good at high ISO. DSLRs have always aimed for more all-around performance including high ISO. Anyway for several years if you have got MF you know that base ISO performance will be great, which has helped established the belief that MF is generally superior, although some DSLRs today mave even have greater DR.

[jrista]

I wouldn't be surprised if modern DSLR's, particularly Sony Exmor, had better DR. The pixel densities (pixel pitch) of 60mp FF MF, or 40mp cropped MF, are about the same as a Canon 5D II. From what I understand, the read noise on an MF sensor is just about as bad as on the 5D II, however if they start wiht 16 bits, and lose two stops to read noise, they would still end up with a solid 13-14 stops of DR, which is still about 2 stops better than any Canon sensor. They might lose out just a bit to a Sony Exmor at that point, as Sony consistently pulls off about 13.8 stops of DR.

[Flake]

I have a comment, who here has ever managed to test or has seen a test on a sensor from any camera?  I've never ever seen a test on a bare sensor, and I truly believe it can't be done in any case.  All the tests I've ever seen have been in camera, together with the image processor, and there's the rub, because the Digic, Expeed, Bionz device has at least as much an influence as the sensor if not more.  The Bionz device is only 12 bit, the expeed has poor throughput so is switchable to allow 12 / 14 bit, only the Digic is full 14 bit.

According to one Canon interview (which I can't put my finger on) there's two stops more of DR to be wrung from the image processor than the sensor, before physics says there's a limit.

Cameras are more than sensors (though reading some threads you'd never believe it) Other components have a massive influence on image quality, and it's a mistake to overlook their importance.

[jrista]

According to one Canon interview (which I can't put my finger on) there's two stops more of DR to be wrung from the image processor than the sensor, before physics says there's a limit.

Absolutely...its not exactly the sensor thats the problem, its the ADC. I believe that actually resides in the Digic image processing chip (I'm not 100% certain of that, I've heard conflicting things, and I don't know what patents of Canon's to look at to figure out which is true), rather than in the sensor (at least for current Canon sensors.) If Canon has changed their approach, and put the ADC on-sensor with their new generation of sensors, and even partially taken the Sony Exmor approach of parallelizing ADC, then we could very likely see an improvement. I don't know enough about CMOS electronics to understand exactly why the ADC burns shadow detail and injects so much fixed-pattern noise. There are also other noise sources, such as thermal, which can be introduced by hot components. That would become more of a problem with high speed cameras, however it can be addressed with proper cooling devices, passive or active.

Cameras are more than sensors (though reading some threads you'd never believe it) Other components have a massive influence on image quality, and it's a mistake to overlook their importance.

Sure, its not just the sensor, its the whole image processing pipeline. I guess I've been using terms loosely, although in the case of Sony Exmor sensors, they definitely pack column-parallel ADC on-chip, and however they designed them, they introduce very little noise (about a third of what the 7D's introduces.) Once the analog signal is converted into digital, its just 1's and 0's, and your not going to see any further introduction of noise...at worst, you might digitally amplify noise (like increasing exposure in post), but with RAW that is obviously not going to happen in-camera.

[KeithR]

the fact is you can't get out of a 7D image (and yes I have one) wht you can from a D7000 at ISO100.

And it amazes me that people get so frothed up about the fact that at one ISO setting the D7000 can do something that the 7D can't do quite as well...

In over 30,000 actuations, my 7D hasn't once been at 100 ISO; and at any other ISO it compares just fine.

Frankly, I'd expect a near two year newer sensor to be much, much more impressive, in direct comparison to the 7D, than it actually is.

Most importantly of all is the fact that "properly" exposed, my 7D has not let me down once. Literally not once.

People who get all worked up about the "magic" D7000/Pentax K5 sensor seem primarily to do so on the basis that it will allow them to stuff up their exposures, secure in the knowledge that the "ISOless" nature of this sensor will allow them to recover at conversion/in post.

Bugger that - I'll just get it right in camera, thanks..!

[jrista]

As someone who also uses my 7D for landscapes, I use ISO 100 as well. I want that maximum DR, and having about two stops chopped off by crappy electronic noise is something I think many of the other landscape photographers who use a 7D is an entirely valid talking point. It should also be noted that for many things, like wildlife and bird photography, the key subject is usually only a part of the frame, and it other parts are blocked or blown it usually doesn't matter as much. When it comes to landscapes (which the 7D is entirely valid for, when you don't have $3000 to spend on a better performing full-frame camera), the whole entire image area matters.

You need every scrap of DR you can get your hands on. Far more often than not, you have to compromise on image quality by using graduated neutral density filters, which almost always create some kind of undesirable outcome to one degree or another (i.e. black mountain peaks, visible separation between sky and land where land or trees protrude through the flat horizon, etc.) When its obviously possible to NOT lose some two stops of DR to read noise because its been done by the competition, its entirely valid to ask for the same improvement and expect something be done by Canon to provide the same benefit to its customers...especially for the prices they seem to be asking for new gear (which is greater than inflation would indicate, if $3500 for a 5D III body is actually true.)

[Curmudgeon]

It's too bad the swamping of the servers sort of dried up this thread, although maybe the subject has now been covered in reasonable depth. I've been looking back over the posts, and I have to say this is one of the most informative and civilized Internet discussions I've ever encountered.

We all know about the excesses and personality aberrations that online forums seem to encourage, but this forum in general and this thread in particular strike me as an example of the positive possibilities of the Internet. I posted several questions myself over the course of the thread and within 24 hours had incisive answers from knowledgeable respondents. You can't get that off Wikipedia unless you already know a lot about the subject you're querying. Have there been disagreements between posters? Of course. Knowledgeable people frequently disagree on specific points. But for the most part the discussion has been carried out without giving or taking the kind of personal offense that has led me to shy away from participating in these discussions.

As a film era fogey trying to expand my grasp of what has become an ever more technical medium, this is exactly what I hoped to get when I registered for this forum. Not only have I increased my understanding of the subject in an academic sense, but I've acquired some practical, useful knowedge about what I can reasonably expect from the 5D3. When (if) it arrives I'll have a better idea of whether it's time for a new body or in the short term my precious photographic dollars would be better spent on, say, a full lash-up of Lee GNDs.  (I've noticed that a lot of people get a little sheepish about the geekiness of participating in these discussions, and I know, I know, it's the photographer not the equipment, but, ... it seems to me that the medium has indeed become so technical--and expensive--that making informed purchases plays at least a modest part in improving the quality of one's photographic output.)

In any event, I'd like to offer a little word of appreciation to jrista, dtaylor, LetTheRightLensIn, Mt Spokane, mkln, Kernuak, Tijn, Flake, Flake, torger, Marsu42 and all the others who made this a stimulating and useful discussion. And Hello to Aglet, another noobie who got sucked off the forum sidelines and into the fray by the issue of dynamic range in Canon equipment.

And while I'm at it, amen to jrista's previous post:

"You need every scrap of DR you can get your hands on. Far more often than not, you have to compromise on image quality by using graduated neutral density filters, which almost always create some kind of undesirable outcome to one degree or another (i.e. black mountain peaks, visible separation between sky and land where land or trees protrude through the flat horizon, etc.) When its obviously possible to NOT lose some two stops of DR to read noise because its been done by the competition, its entirely valid to ask for the same improvement and expect something be done by Canon to provide the same benefit to its customers...especially for the prices they seem to be asking for new gear (which is greater than inflation would indicate, if $3500 for a 5D III body is actually true.)"

There are three significant problems with NDs.

1) As jrista points out, it's hard not to leave obvious traces of them in the photograph--especially if, like me, mountain scenery is a favorite subject.

2) A good set is considerable added weight, especially if you're an aging backpacker who's already made the concession to artistic integrity of schlepping a full frame DSLR and several lens.

3) Shooting landscapes is not like covering the sidelines of the NBA or NFL, but it's amazing how often the best shots involve serendipity and "shooting from the hip." One day 40 years ago I waited until the rain stopped to take out the trash--and walked into what remains the most spectacular sunset I've ever seen. Cloud textures in infinite variety and god rays up the ying yang. I dropped the can, ran inside and grabbed my tripod and old Mamiya SLR and fired off the half roll of film in the camera. I'd never heard of a neutral density filter, but I got the shot. The scene was gone in two minutes. If I'd been fumbling around mounting, testing and aligning NDs on my 5D2 I'd have gotten a well-exposed shot of the afterglow.

[Kernuak]

The 7D isn't really suited to landscape work, for more reasons than DR, it can do it, but it doesn't produce the best results. Just like the 5D MkII and wildlife, it can do it, but it isn't its forte. I orignally got my 7D to replace the 40D, but I kept the 40D, following an instance where I missed a shot through only having one body. The plan was to use it for landscape and wildlife, just like I had with the 40D. I had seen some deficiencies with the 40D for landscapes, but as soon as I got the 7D, everything was magnified. L zooms start to look soft on crop cameras due to diffraction and of course, the diffraction limited aperture on the 7D is wider than on the 40D due to the pixel density and interference due to Airey discs. This coupled with the higher resolution show up the softer images more, particularly in the corners (even though the worst performing parts of a lens are cropped out by the sensor). According to DXO, the 7D has more DR than the 5D MkII, but I don't actually find to usable, particularly before one of the early firmware updates. You tend to get more colour casts on the 7D and some of my early shots were distinctly pink in the highlight areas.
I much prefer to get it as right as possible in camera and that is why I choose to use grad filters. Many of the landscapes I shoot have upwards of 5 stops difference between the foreground and sky, so even with a 3 stop improvement in DR, I won't be selling off my Lee grad filters any time soon.

[jrista]

The 7D isn't really suited to landscape work, for more reasons than DR, it can do it, but it doesn't produce the best results.

Hmm, really? I...I would have never known...

(Just a random sampling. Take a skilled photographer, the 10-22 EF-S, and maybe a few GND filters, and the 7D can do wonders!)
http://500px.com/photo/1157203
http://500px.com/photo/3718508
http://500px.com/photo/3131439
http://500px.com/photo/1929326
http://500px.com/photo/3981585

Sadly, you can clearly tell most of those have used a GND filter or two. But thats exactly my point about DR.

Just like the 5D MkII and wildlife, it can do it, but it isn't its forte. I orignally got my 7D to replace the 40D, but I kept the 40D, following an instance where I missed a shot through only having one body. The plan was to use it for landscape and wildlife, just like I had with the 40D. I had seen some deficiencies with the 40D for landscapes, but as soon as I got the 7D, everything was magnified.

The reason the 5D II isn't great for wildlife is because its limited by AF. It doesn't really have better DR than the 7D though (I think the difference between the two is a small fraction of a stop), and at ISO 100, the two cameras are little different from each other (except for the fact that the 5D II exhibits less noise and has a larger frame.) Conversely, the 7D HAS a highly capable AF system, but that doesn't preclude it from being used for other purposes, like landscapes. If the 5D III gets a nice AF system, it will be every bit as capable for wildlife as for landscapes, limited against the 7D only in the aspect of reach.

L zooms start to look soft on crop cameras due to diffraction and of course, the diffraction limited aperture on the 7D is wider than on the 40D due to the pixel density and interference due to Airey discs. This coupled with the higher resolution show up the softer images more, particularly in the corners (even though the worst performing parts of a lens are cropped out by the sensor).

I think this is an area where most people misunderstand diffraction. On the 7D, vs. the 40D, diffraction only sets in sooner...simply because the 7D is recording at a higher resolution. The 7D is recording more information at a finer degree, and the only reason...at 100% crop...it looks "worse" than the 40D is simply because the 40D is incapable of capturing enough detail for an even comparison. Normalize the image size with the 40D, and the 7D will always win. You can NEVER get WORSE IQ with more resolution...you can only approach a point where you get diminishing returns for significant increases in cost (and image files that are so huge and contain so many pixels that you don't need all the extra data anyway, even IF your upscaling.) Scale any pair of images from the 7D and 40D down to say 900 pixels long edge, and all that "softness" you see when pixel peeping at 100% will become a total non-issue. Print at 8x10 or 13x19, and the problems you perceive at 100% will also cease to be an issue (well, probably less than 13x19, since the 40D can't actually achieve that natively...maybe 11x16.) You can't compare IQ when viewing a 100%, because you are literally comparing apples to oranges.

According to DXO, the 7D has more DR than the 5D MkII, but I don't actually find to usable, particularly before one of the early firmware updates. You tend to get more colour casts on the 7D and some of my early shots were distinctly pink in the highlight areas.
I much prefer to get it as right as possible in camera and that is why I choose to use grad filters. Many of the landscapes I shoot have upwards of 5 stops difference between the foreground and sky, so even with a 3 stop improvement in DR, I won't be selling off my Lee grad filters any time soon.

Sure, I'm an advocate of get it right in-camera, too. And "usable DR" is a tough thing to nail down, so experiences differ on that front. I usually have to use my Lee .6 GND, sometimes the .9, and its only in extreme cases where I need to correct more than about 3 stops of contrast different to capture a landscape the way I want to (I don't usually want to capture an unlimited amount of detail in the shadows, I'm happy leaving them dark most of the time.) I think my assertions here about DR play MORE to the whole "get it right in camera" crowd than anyone else though. With more DR, you have more room within which to work...in camera...out in the field, than if you have less DR. With more DR, you have to spend less time fiddling around with physical filtration to get a good shot, meaning that when that moment...that extremely SHORT moment...hits, when light and shadow is absolutely perfect, and your simply agape in awe, you have more time to think about composition, exposure, etc. and get that superb shot...WITHOUT any funky shading because of the need to use graduated filters.

[jrista]

It's too bad the swamping of the servers sort of dried up this thread, although maybe the subject has now been covered in reasonable depth. I've been looking back over the posts, and I have to say this is one of the most informative and civilized Internet discussions I've ever encountered.

We all know about the excesses and personality aberrations that online forums seem to encourage, but this forum in general and this thread in particular strike me as an example of the positive possibilities of the Internet. I posted several questions myself over the course of the thread and within 24 hours had incisive answers from knowledgeable respondents. You can't get that off Wikipedia unless you already know a lot about the subject you're querying. Have there been disagreements between posters? Of course. Knowledgeable people frequently disagree on specific points. But for the most part the discussion has been carried out without giving or taking the kind of personal offense that has led me to shy away from participating in these discussions.

As a film era fogey trying to expand my grasp of what has become an ever more technical medium, this is exactly what I hoped to get when I registered for this forum. Not only have I increased my understanding of the subject in an academic sense, but I've acquired some practical, useful knowedge about what I can reasonably expect from the 5D3. When (if) it arrives I'll have a better idea of whether it's time for a new body or in the short term my precious photographic dollars would be better spent on, say, a full lash-up of Lee GNDs.  (I've noticed that a lot of people get a little sheepish about the geekiness of participating in these discussions, and I know, I know, it's the photographer not the equipment, but, ... it seems to me that the medium has indeed become so technical--and expensive--that making informed purchases plays at least a modest part in improving the quality of one's photographic output.)

In any event, I'd like to offer a little word of appreciation to jrista, dtaylor, LetTheRightLensIn, Mt Spokane, mkln, Kernuak, Tijn, Flake, Flake, torger, Marsu42 and all the others who made this a stimulating and useful discussion. And Hello to Aglet, another noobie who got sucked off the forum sidelines and into the fray by the issue of dynamic range in Canon equipment.

Certainly! I always enjoy a good debate, especially when they don't degrade into uselessness like so many things on the net.

[Kernuak]

The 7D isn't really suited to landscape work, for more reasons than DR, it can do it, but it doesn't produce the best results.

Hmm, really? I...I would have never known...

(Just a random sampling. Take a skilled photographer, the 10-22 EF-S, and maybe a few GND filters, and the 7D can do wonders!)
http://500px.com/photo/1157203
http://500px.com/photo/3718508
http://500px.com/photo/3131439
http://500px.com/photo/1929326
http://500px.com/photo/3981585

Sadly, you can clearly tell most of those have used a GND filter or two. But thats exactly my point about DR.

You don't need a 10-22 to get good results, in fact for many of the landscapes around here, that would be far too wide. The difference is, when you're using longer focal lengths, you need narrower apertures, so diffraction becomes more of an issue. While they're good images, they're too heavily processed for my tastes. That's probably the main difference between your experience and mine, as the more processing work, the greater the chance that noise starts to show. Also, I judge based on 50%, as that is more representative of what is visible in an average print and may have a greater tolerance for noise in an image anyway.

Just like the 5D MkII and wildlife, it can do it, but it isn't its forte. I orignally got my 7D to replace the 40D, but I kept the 40D, following an instance where I missed a shot through only having one body. The plan was to use it for landscape and wildlife, just like I had with the 40D. I had seen some deficiencies with the 40D for landscapes, but as soon as I got the 7D, everything was magnified.

The reason the 5D II isn't great for wildlife is because its limited by AF.

Precisely. However, when it gets dark, I find the AF on the 5D MkII more reliable...as long as I use the centre point.

It doesn't really have better DR than the 7D though (I think the difference between the two is a small fraction of a stop).

I didn't say that the 5D has more DR than the 7D (DXO suggests the opposite), but in terms of what [i'I[/i] find usable, it is better, in fact, the cut off at both ends is slightly different between the two. It also has a different tonality, that I prefer.

Conversely, the 7D HAS a highly capable AF system, but that doesn't preclude it from being used for other purposes, like landscapes. If the 5D III gets a nice AF system, it will be every bit as capable for wildlife as for landscapes, limited against the 7D only in the aspect of reach.

Likewise, I didn't say that it was preculded from other purposes, on the contrary, I used it for landscapes for around 6+ months before I got the 5D.

L zooms start to look soft on crop cameras due to diffraction and of course, the diffraction limited aperture on the 7D is wider than on the 40D due to the pixel density and interference due to Airey discs. This coupled with the higher resolution show up the softer images more, particularly in the corners (even though the worst performing parts of a lens are cropped out by the sensor).

I think this is an area where most people misunderstand diffraction. On the 7D, vs. the 40D, diffraction only sets in sooner...simply because the 7D is recording at a higher resolution. The 7D is recording more information at a finer degree, and the only reason...at 100% crop...it looks "worse" than the 40D is simply because the 40D is incapable of capturing enough detail for an even comparison. Normalize the image size with the 40D, and the 7D will always win. You can NEVER get WORSE IQ with more resolution...you can only approach a point where you get diminishing returns for significant increases in cost (and image files that are so huge and contain so many pixels that you don't need all the extra data anyway, even IF your upscaling.) Scale any pair of images from the 7D and 40D down to say 900 pixels long edge, and all that "softness" you see when pixel peeping at 100% will become a total non-issue. Print at 8x10 or 13x19, and the problems you perceive at 100% will also cease to be an issue (well, probably less than 13x19, since the 40D can't actually achieve that natively...maybe 11x16.) You can't compare IQ when viewing a 100%, because you are literally comparing apples to oranges. [/quote]


If that was the only reason, then why is DLA quoted in terms of pixel size? I agree, that a large part of the increased softness is due to looking at a smaller area and more detail, but there are also relationships between the pixel:Airy disk size ratio. When using the 17-40 and 24-105 though, I certainly noticed soft corners (although I also see it on the 5D because of the lack of sweet spot effect from crop sensors).

According to DXO, the 7D has more DR than the 5D MkII, but I don't actually find it usable, particularly before one of the early firmware updates. You tend to get more colour casts on the 7D and some of my early shots were distinctly pink in the highlight areas.
I much prefer to get it as right as possible in camera and that is why I choose to use grad filters. Many of the landscapes I shoot have upwards of 5 stops difference between the foreground and sky, so even with a 3 stop improvement in DR, I won't be selling off my Lee grad filters any time soon.

Sure, I'm an advocate of get it right in-camera, too. And "usable DR" is a tough thing to nail down, so experiences differ on that front. I usually have to use my Lee .6 GND, sometimes the .9, and its only in extreme cases where I need to correct more than about 3 stops of contrast different to capture a landscape the way I want to (I don't usually want to capture an unlimited amount of detail in the shadows, I'm happy leaving them dark most of the time.) I think my assertions here about DR play MORE to the whole "get it right in camera" crowd than anyone else though. With more DR, you have more room within which to work...in camera...out in the field, than if you have less DR. With more DR, you have to spend less time fiddling around with physical filtration to get a good shot, meaning that when that moment...that extremely SHORT moment...hits, when light and shadow is absolutely perfect, and your simply agape in awe, you have more time to think about composition, exposure, etc. and get that superb shot...WITHOUT any funky shading because of the need to use graduated filters.

For the shooting that I do, it is variable on what level of filtration I need. When I photograph sunsets, it tends to be with quite dramatic lighting, rather than pastel shades that others prefer. Some such as this one didn't need any grads at all (taken with the 40D a few years ago).

Glowing Waves by Kernuak, on Flickr
Others I need up to 5-6 stops, so end up having to combine filters, although for sake of IQ, I prefer not to. Use of Lee filters has enabled me to use more with less IQ degradation though, in comparison to my old Hi-Tech filters. Also, the 24mm MkII helps .

Picnic Table Sunset by Kernuak, on Flickr

Anyway, I have some images to process before it's bedtime.

[jrista]

If that was the only reason, then why is DLA quoted in terms of pixel size? I agree, that a large part of the increased softness is due to looking at a smaller area and more detail, but there are also relationships between the pixel:Airy disk size ratio. When using the 17-40 and 24-105 though, I certainly noticed soft corners (although I also see it on the 5D because of the lack of sweet spot effect from crop sensors).

The misconception is in the (incorrect) idea that once the airy disk has grown larger than about a 2x2 pixel area, IQ DROPS. That is ABSOLUTELY NOT the case. If you take a camera with a pixel pitch of 6um, and another camera with a pixel pitch of 3um. You can fit four times as many 3um pixels into the total area of a single 6um pixel. Diffraction will set in at around f/11 for the 6um camera, where as it will probably set in around f/6 for the 3um camera. Note the pixel sizes there...at f/6, the airy disk has grown large enough to become apparent. The 3um sensor is now diffraction limited, in that as you continue to stop down the aperture, diffraction will interfere more and more with detail. Lets say you stop down to f/8. Well, your 3um sensor is recording some of the effects of diffraction...they are minor, and your still capturing a lot of detail, but diffraction is having an effect. The 6um sensor, however, isn't even experiencing the effects of diffraction yet. Its still capturing LESS detail, despite the fact that the 3um sensor is now a full stop beyond its initial DLA. Now lets stop down to f/22. Diffraction will visibly affect image quality now...softness will be prevalent. The images from both sensors now are roughly the same...however the 3um sensor is still going to produce a slightly better result, since its recording the image at a finer level than the 6um sensor. Think about pixellation of an image...if you pixellate at 6x6 pixel and 3x3 pixel squares, even if you blur the image, the 3x3 pixellated image is always going to exhibit more detail.

You can't get worse quality from a higher resolution sensor as you stop down to progressively smaller apertures. You only approach a normal baseline, wherein all images from any camera of any resolution are pretty much the same.

[Circles]

Precisely. However, when it gets dark, I find the AF on the 5D MkII more reliable...as long as I use the centre point

[RedEye]

If that was the only reason, then why is DLA quoted in terms of pixel size? I agree, that a large part of the increased softness is due to looking at a smaller area and more detail, but there are also relationships between the pixel:Airy disk size ratio. When using the 17-40 and 24-105 though, I certainly noticed soft corners (although I also see it on the 5D because of the lack of sweet spot effect from crop sensors).

The misconception is in the (incorrect) idea that once the airy disk has grown larger than about a 2x2 pixel area, IQ DROPS. That is ABSOLUTELY NOT the case. If you take a camera with a pixel pitch of 6um, and another camera with a pixel pitch of 3um. You can fit twice as many 3um pixels into the total area of a single 6um pixel. Diffraction will set in at around f/11 for the 6um camera, where as it will probably set in around f/6 for the 3um camera. Note the pixel sizes there...at f/6, the airy disk has grown large enough to become apparent. The 3um sensor is now diffraction limited, in that as you continue to stop down the aperture, diffraction will interfere more and more with detail. Lets say you stop down to f/8. Well, your 3um sensor is recording some of the effects of diffraction...they are minor, and your still capturing a lot of detail, but diffraction is having an effect. The 6um sensor, however, isn't even experiencing the effects of diffraction yet. Its still capturing LESS detail, despite the fact that the 3um sensor is now a full stop beyond its initial DLA. Now lets stop down to f/22. Diffraction will visibly affect image quality now...softness will be prevalent. The images from both sensors now are roughly the same...however the 3um sensor is still going to produce a slightly better result, since its recording the image at a finer level than the 6um sensor. Think about pixellation of an image...if you pixellate at 6x6 pixel and 3x3 pixel squares, even if you blur the image, the 3x3 pixellated image is always going to exhibit more detail.

You can't get worse quality from a higher resolution sensor as you stop down to progressively smaller apertures. You only approach a normal baseline, wherein all images from any camera of any resolution are pretty much the same.

Thank you for the excellent explanation!
+1