Another mention of a 70+ megapixel EOS R camera

[amorse]

It definitely didn't look bad among 5DIII and 6D. But follow your own link and add Nikon D810 that was released a year before 5DS..

That's true, but my point was the 5DS dynamic range didn't appear to be hindered by it's high resolution, it was hindered by Canon's sensor design and that despite being Canon's highest resolution option available, it still outperformed the bulk of other Canon sensors released in a similar time frame. You could compare literally any Canon dslr or milc to the D810 and find the same result, the DR limitations don't appear to be truly linked to resolution but rather sensor design.

 

[Quarkcharmed]

The 5DS(r) uses an older-style sensor (than, say, the 5D4), I understand, so it's a difficult comparison to make - in any case this is not unambiguous evidence that the resolution is the cause of any differences.

Yeah I agree 5DS was a bad example as we need another lower-res sensor of the same generation to compare how pixel size affects the DR, but I'm not sure if there's any from Canon. They started pushing their dual pixel tech.

Incidentally, while the pixels are smaller, and pixel-level noise may be higher, I understand that image noise will be the same (as another FF sensor of the same generation), and since noise reduction can be carried out more finely, the finished image may actually be cleaner from a higher-resolution sensor (people seem to confuse pixel- and image- level noise in these discussions, ditto motion blur etc).

Yes it may even be better after downsampling.

Quite. My question was merely a matter of terminology. As for in-camera versus PP downsampling, I imagine those asking for the former expect a faster fps, but since it must take extra processing power to do it, I'm not sure if that would be the case (though it would save storage space on the memory card).

I think writing to the card takes more time that in-camera processing, that's why in-camera downsampling may help increase the burst rate. Roughly speaking, say you have internal in-memory RAW data of the size of S and it takes T time to write. Now we want to downsample and shrink it to the size of S/2, it will now only take T/2 time to write. But in-camera downsampling will take much less than T/2 time to process, so in total writing of downsampled data will be between T/2 and T. I guess it'll be closer to T/2.

 

[dtaylor]

It definitely didn't look bad among 5DIII and 6D. But follow your own link and add Nikon D810 that was released a year before 5DS..

+3ev on the 5Ds is virtually identical to +4ev on the D810. +4ev on the 5Ds falls between +5ev and +6ev on the D810. With a little NR you can reliably do a 4 stop push on the 5Ds for a large print. +5 gets really noisy/gritty so I wouldn't count on it except perhaps for small prints.

I've repeatedly posted a 7D landscape shot where I did a 2.5ev push to open shadow detail that was just above pure black (zone IX). Excluding scenes that would have required 2-3 exposures on any camera, I found it was extremely rare to not be able to capture a scene's dynamic range on the 7D sensor using ETTR techniques and a little NR in post.

5Ds DR is not "bad" by any real world definition of DR, and there's very little practical difference in terms of DR between shooting a 5Ds and a D8x0. There's even less difference between a 5D4 and D8x0. If I were to sum the differences up, on the 5Ds you should be careful to ETTR wide DR scenes, and you might need to apply some additional NR in post.

DR is discussed ad nauseam on forums. That made some sense when it was 5D3 (banding issue) vs. D800. It really doesn't make sense today.

If I were to guess, Canon's 70mp sensor will likely use the same architecture as the 5D4 for similar DR. And there will be weeping and gnashing of teeth that its' graph on DxO or PtoP doesn't match Nikon's or Sony's best even though the real world practical impact is nil.

 

[dtaylor]

Yeah I agree 5DS was a bad example as we need another lower-res sensor of the same generation to compare how pixel size affects the DR, but I'm not sure if there's any from Canon. They started pushing their dual pixel tech.

For both off-chip ADC sensors (Canon prior to 5D4) and on-chip ADC sensors (Nikon/Sony) the highest DR sensors are also the highest resolution ones. One can point out that Canon's 5Ds sensor isn't exactly the same generation as a 6D II, but with Nikon/Sony we see variable resolutions within the same generation.

For the record, larger pixels should result in higher DR not because of read noise but because of well capacity. It's odd that this is not the case right now.

As to downsampling in camera: I think everyone would welcome a mode whereby you could shoot 35mp at a higher fps (say 5/8 or 5/10). But I can't think of an example of Canon doing this before. I certainly wouldn't mind them trying it now.

 

[Quarkcharmed]

I'm not an expert on this but downsampling 4:1 would mean you get four doses of "read noise" and while that often offsets, statistically speaking 4 of them on average are going to be twice as bad as 1. So while 4 small pixels could theoretically capture the exact same photons as 1 big one, and the downsampling (let us say) perfectly averages those out to the right photon count, you'll still have more noise.

Smaller pixels have worse signal-to-noise ratio but they're actually less noisy in terms of ADC processing. Bigger pixels have bigger SNR but the noise part of that ratio is bigger. I may be totally wrong on that but I recall a paper with explanation why multiple small pixels + downsampling may be better than one big pixel.

Anyway when I downsample in Lightroom, if I shrink each side by 2, it'll remove a lot of noise from the high-ISO shots. At a cost of resolution, of course.

 

[rrcphoto]

I think writing to the card takes more time that in-camera processing, that's why in-camera downsampling may help increase the burst rate. Roughly speaking, say you have internal in-memory RAW data of the size of S and it takes T time to write. Now we want to downsample and shrink it to the size of S/2, it will now only take T/2 time to write. But in-camera downsampling will take much less than T/2 time to process, so in total writing of downsampled data will be between T/2 and T. I guess it'll be closer to T/2.

MRAW options have never given us a speed increase in the past, why all of a sudden would it do so now?

 

[QuisUtDeus]

I was surprised by that as well since it has the same pixel density as an APS-C Canon camera. But I have to admit I rarely had good results with the shutterspeed = 1/focal length rule, or (f1.6) for crop. I've always needed (f2) or faster on non-IS lenses.

That makes sense, if the old 1/FL was from the days of film or low-density FF sensors. A rule that worked for an 18mpx crop sensor won't necessarily work for 24mpx...

 

[stevelee]

Downsampling is an art and a science. An online acquaintance years ago adopted an iterative method after his tests. He made dye transfer prints of his photos, and needed to preserve as much quality as possible in showcasing them on his web site.

 

[privatebydesign]

Downsampling is an art and a science. An online acquaintance years ago adopted an iterative method after his tests. He made dye transfer prints of his photos, and needed to preserve as much quality as possible in showcasing them on his web site.

The key to that is ‘years ago’ software has moved on in leaps and bounds and the iterative process of upsampling and downsampling has been proven to be unnecessary and ineffective. Of course there are different programs with different algorithms that preform differently, especially with specific subject detail levels, but the iterative process has been set aside for years now.

 

[Quarkcharmed]

MRAW options have never given us a speed increase in the past, why all of a sudden would it do so now?

It was discussed above in the context of very large 75mp files, perhaps around 70-80mb each. The burst speed is also limited by physical factors such as mirror, shutter etc. But most importantly the camera doesn't write to the card right away, it writes to the buffer and raw size affects how many frames fit into the buffer. Then the buffer gets flushed to the card and the size of the contents affects the waiting time.

 

[Deleted member 7828]

Exactly. High resolution has three potential drawbacks: (1) worse DR and high ISO, (2) lower writing speed that affects continuous shooting and (3) much bigger raw files. Otherwise a high-res camera is better than low-res in all regards.

I would add a 4th and that is needing a higher shutter speed to freeze action/get a sharp image, which requires more light, a faster lens or higher ISO. Coupled with #1 above, I’d consider that to be a big negative.

An honorable mention might be that it is much more demanding of your lenses and this is noticeable particularly if you view at 100%, print large or crop heavily. (I know not everyone here does those things though)

 

[rrcphoto]

It was discussed above in the context of very large 75mp files, perhaps around 70-80mb each. The burst speed is also limited by physical factors such as mirror, shutter etc. But most importantly the camera doesn't write to the card right away, it writes to the buffer and raw size affects how many frames fit into the buffer. Then the buffer gets flushed to the card and the size of the contents affects the waiting time.

I realize that. but you didn't answer the question at all.

MRAW has never provided those options in the past, ie: on the 5Ds / 5DsR,etc so why would they magically occur now?

it's still the reading of the sensor, and the initial processing of 75mp that is the problem.

there's zero in the way of evidence that canon can or will do a higher speed MRAW version of a 70+ MP sensor size.

 

[rrcphoto]

I would add a 4th and that is needing a higher shutter speed to freeze action/get a sharp image, which requires more light, a faster lens or higher ISO. Coupled with #1 above, I’d consider that to be a big negative.

An honorable mention might be that it is much more demanding of your lenses and this is noticeable particularly if you view at 100%, print large or crop heavily. (I know not everyone here does those things though)

so many myths.

1. image dr will be approximately the same regardless of pixel density.
2. shutter speeds will be approximately the same at comparable image sizes
3. your lenses will perform better again, with comparable image sizes.

the point is that 70MP allows you to oversample your image. 30MP does not.

it does not necessarily mean that a) you can print massively bigger b) crop massively more. While you may be able to, you are in diminishing returns.

 

[Quarkcharmed]

+3ev on the 5Ds is virtually identical to +4ev on the D810. +4ev on the 5Ds falls between +5ev and +6ev on the D810. With a little NR you can reliably do a 4 stop push on the 5Ds for a large print. +5 gets really noisy/gritty so I wouldn't count on it except perhaps for small prints.

In your sample, +3ev on the 5Ds is actually slightly worse than +4ev on the D810. And +4 vs +6 look the same, just different patterns of noise. 2stop difference is huge.

I've repeatedly posted a 7D landscape shot where I did a 2.5ev push to open shadow detail that was just above pure black (zone IX). Excluding scenes that would have required 2-3 exposures on any camera, I found it was extremely rare to not be able to capture a scene's dynamic range on the 7D sensor using ETTR techniques and a little NR in post.

Again the sample from 7D doesn't prove anything about 5Ds vs D810 comparison. And for that matter, again I'd like to point out it wasn't a push from almost complete black, all the details can be seen in the shadows on the 'before' image, if it's almost black to you, check your monitor calibration.

5Ds DR is not "bad" by any real world definition of DR, and there's very little practical difference in terms of DR between shooting a 5Ds and a D8x0. There's even less difference between a 5D4 and D8x0. If I were to sum the differences up, on the 5Ds you should be careful to ETTR wide DR scenes, and you might need to apply some additional NR in post.

DR is discussed ad nauseam on forums. That made some sense when it was 5D3 (banding issue) vs. D800. It really doesn't make sense today.

If I were to guess, Canon's 70mp sensor will likely use the same architecture as the 5D4 for similar DR. And there will be weeping and gnashing of teeth that its' graph on DxO or PtoP doesn't match Nikon's or Sony's best even though the real world practical impact is nil.

I actually played with 5Ds sample raw files some time ago as I considered buying it, the resolution is fantastic as well as overall image quality, but only within its limits. Overall quality isn't comparable to 5DIV. That is for how I use it, I liked 5DIV more.

 

[Quarkcharmed]

I realize that. but you didn't answer the question at all.

MRAW has never provided those options in the past, ie: on the 5Ds / 5DsR,etc so why would they magically occur now?

it's still the reading of the sensor, and the initial processing of 75mp that is the problem.

there's zero in the way of evidence that canon can or will do a higher speed MRAW version of a 70+ MP sensor size.

I'm not so interested in the burst speed actually. 5fps is enough for me. Even less. I was guessing about downsampled mode with higher DR and/or better noise at high ISO. Burst speed was just a wishful thinking. And I elaborated that internal downsampling may be faster than writing to the card so maybe Canon will use it. Maybe not.

Overall I believe Canon can deliver a 75mp camera, but I doubt it'll fully satisfy the picky audience of this forum, including DR, burst speed etc.

 

[Loswr]

Overall I believe Canon can deliver a 75mp camera anything and/or everything, but I doubt it'll fully satisfy the picky audience of this forum...

Fixed that for ya.

 

[Quarkcharmed]

Fixed that for ya.

I hope you don't charge for the repair I didn't ask for!

 

[Loswr]

I hope you don't charge for the repair I didn't ask for!

 

[dtaylor]

In your sample, +3ev on the 5Ds is actually slightly worse than +4ev on the D810.

If you scroll around the image you can find spots where the 5Ds appears better, but you also quickly realize that at any spot it's splitting hairs with differences that could be due to very slight differences in other factors, even just the default color profiles.

And +4 vs +6 look the same, just different patterns of noise.

The D810 noise is worse in the screenshot I posted, and that's made undeniably clear from scrolling around the image. Though again, the 5Ds @ +4 is not as good as the D810 @ +5, which is why I said between 5/6ev.

2stop difference is huge.

Show us. Show us a real world image you've shot where you pushed shadows +4.5ev, +2ev over my 7D shot, and the difference was huge. You've got a 5D4 which should be able to handle that so if DR is so important, so often, and so huge you should have landscape shots that illustrate it.

I'm tired of words from people on this topic. The lack of images very strongly suggest to me that the people who bring up DR on forums are not the people shooting wide DR scenes and producing wide DR prints. And that they therefore have no real idea of the relative impact, what it actually means to shoot two cameras 1ev apart, or 2ev apart, or 3ev part.

Again the sample from 7D doesn't prove anything about 5Ds vs D810 comparison.

Again, you've posted nothing.

 

[dtaylor]

What's this?

Expose to the Right. It means having your histogram shifted as far to the right as possible without unrecoverable highlights, even if that means parts of the image appear overexposed prior to post processing.

ETTR allows for maximum shadow recovery (DR) and, because of the way ADCs work, maximum tonality.