Canon's New 50 MP Sensor

[Guillermo Luijk]

There is still this strange idea that digital cameras have "less highlight dynamic range than film" even though most system cameras now have far more dynamic range than (color) film, due simply to default exposure metering that leaves only about three stops between the metered average level and the clipping level, even when the sensor then has nine or more usable stops below the metered average level. (Yes, I am happy to  have a camera whose default metering and signal processing positions metered mid-tones about four or more stops below the maximum raw level; I can adjust my exposure up ["over-expose"!?] when it is more important to reducing noise in deep shadow regions at base ISO speed than to avoid blown

For some reason camera makers do not calibrate their exposure metering (i.e. effective ISO) according to the dynamic range of modern sensors, usually wasting a good amount of it.

Specially for Sony's ISOless sensors, with a huge DR at base ISO that quickly vanishes when pushing ISO, in my opinion exposure metering should allow much more highlight headroom for the lower ISO settings than they currently do, since shadows can be lifted with a strong tone curve still maintaining a good SNR. This would prevent more easily skies and other highlight sources from getting blown in the camera JPEGs.

For higher ISO settings highlight headroom from the exposure meter could be reduced to keep a good balance of shadows vs highlights DR.

Regards

[dwswager]

Back to the rumor!  It could very well be a correct guess, but tracking back the sources quoted, it seem no more than that: an anonymous tipster's guess that Canon will use its most recently revealed pixel design (as in the 7DII) in its next 36x24mm sensor.

That is how you might guess it would work...just use the current pixel pitch of the 7DmkII APS-C (1.6x) on the new 5DmkIV full frame.  But then if Nikon followed that pattern, the D810 (4.8 micron pixel pitch) would have the 3.9 micron pitch of the 24MP D7100 APS-C (1.5X) sensor and end up with basically 48MP instead of 36MP.  But given the same technology of two sensors, while smaller pitch increases the resolution, it also decreases the SNR which Canon is already having issues with.  They seem to have tempered the noise some on the 7DmkII at higher ISO with a better gain strategy, but a 50MP camera would be designed more for base ISO shooting.

[BernardLanguillier]

That is how you might guess it would work...just use the current pixel pitch of the 7DmkII APS-C (1.6x) on the new 5DmkIV full frame.  But then if Nikon followed that pattern, the D810 (4.8 micron pixel pitch) would have the 3.9 micron pitch of the 24MP D7100 APS-C (1.5X) sensor and end up with basically 48MP instead of 36MP.  But given the same technology of two sensors, while smaller pitch increases the resolution, it also decreases the SNR which Canon is already having issues with.  They seem to have tempered the noise some on the 7DmkII at higher ISO with a better gain strategy, but a 50MP camera would be designed more for base ISO shooting.

On the Nikon side, it would in fact be 54mp if you simply scale the D7000 sensor design, but I believe that Sony/Nikon understand that DR is an order of magnitude more important than resolution and there is no way they would release a new high res sensor/camera with less DR than their respective current FF champion (a7R or D810).

So I believe than they will stick to a less agressive AA filter less 46mp sensor and cross the 15 stops DR barrier.

Future will tell.

Cheers,
Bernard

[dwswager]

On the Nikon side, it would in fact be 54mp if you simply scale the D7000 sensor design, but I believe that Sony/Nikon understand that DR is an order of magnitude more important than resolution and there is no way they would release a new high res sensor/camera with less DR than their respective current FF champion (a7R or D810).

So I believe than they will stick to a less agressive AA filter less 46mp sensor and cross the 15 stops DR barrier.

Cheers,
Bernard

48MP/54MP same damn thing.  And I agree that no company would sacrifice DR for additional resolution, especially at the additional expense of throughput and the smaller market size.  Which means, if Canon actually does release a 50MP sensor they would have made a significant technology breakthrough that has eluded them for years or they bought it.

[BernardLanguillier]

Which means, if Canon actually does release a 50MP sensor they would have made a significant technology breakthrough that has eluded them for years or they bought it.

Yes, that's my guess as well. From what I hear, it seems certain that it will be a Canon sensor though.

Cheers,
Bernard

[hjulenissen]

48MP/54MP same damn thing.  And I agree that no company would sacrifice DR for additional resolution
...

I am guessing that practically all cameras sacrifice some DR for some resolution. I.e. that there is a trade-off to be made, and if you don't care much about one, you can do somewhat better at the other.

I do, however, disagree with some "megapixel-haters" in that I believe that there is a pragmatic sweet-spot for each technology/generation, where you get sensible DR and sensible resolution, and sacrificing one will only buy you small amounts of the other.

I am sure that some photographers would find a 50+MP 24x36mm camera useful, even if it offered only moderate DR at low ISO (was not some films in that cathegory?). For me, however, I would rather have it the other way around.

-h

[barryfitzgerald]

I would say that if the exposure is what is needed to avoid blown highlights, it is not "under-exposure", it is just correct exposure! (Correcting for the default exposure metering of most digital cameras, which assumes no more than about three stops between the average illumination and the peak illumination in the scene.)

There is still this strange idea that digital cameras have "less highlight dynamic range than film" even though most system cameras now have far more dynamic range than (color) film, due simply to default exposure metering that leaves only about three stops between the metered average level and the clipping level, even when the sensor then has nine or more usable stops below the metered average level. (Yes, I am happy to  have a camera whose default metering and signal processing positions metered mid-tones about four or more stops below the maximum raw level; I can adjust my exposure up ["over-expose"!?] when it is more important to reducing noise in deep shadow regions at base ISO speed than to avoid blown highlights -- which is rare for me!)


Back to the rumor!  It could very well be a correct guess, but tracking back the sources quoted, it seem no more than that: an anonymous tipster's guess that Canon will use its most recently revealed pixel design (as in the 7DII) in its next 36x24mm sensor.

I continue to shoot film and regarding the DR comments mostly come from those who have not shot film or don't
DR has improved hugely with digital (over the last 10 years) Normal procedure for colour (or b&w) negative film is to expose for the shadows depending on the emulsion used it's not really that easy to blow out negative film in the highlights working this way (expose for shadows) is quite easy and liked by many (I dislike slide film for it's limited DR)
Digital requires a different approach but we're at a level now where the best sensors have excellent "film like" DR albeit with a bit of shadow pulling or HL recovery in post.

It's far less a problem..unless you shoot Canon in which case you're in for a rough ride you can't aggressively pull the images at base ISO like you can on other sensor makers, the DR and latitude just isn't there. So the DR issue isn't really one anymore, but it is if you're stuck in the Canon timewarp (which includes full frame)

Regarding the "big pixel count" body it looks like something yet again that has little practical beneift for most people big numbers likely very expensive one for the boy racer camera users of the world

[shadowblade]

Regarding the "big pixel count" body it looks like something yet again that has little practical beneift for most people big numbers likely very expensive one for the boy racer camera users of the world

All the former medium format studio, landscape and advertising photographers who have moved from Hasselblad or Leica to the D800e/D810 would argue otherwise.

As would anyone who regularly makes huge prints of highly-detailed subjects.

[dwswager]

All the former medium format studio, landscape and advertising photographers who have moved from Hasselblad or Leica to the D800e/D810 would argue otherwise.

As would anyone who regularly makes huge prints of highly-detailed subjects.

So should others.  It can be about achievement or, for me, OPTIONS.

I bought the D810 as a general use camera and with 36MP in FX it also gives 24MP in 1.2X crop and 15.6MP in 1.5x DX crop.  Basically, I can choose to bring the D810 instead of the D7100 where I might want both.  Or it gives one the option to only have 1 camera. 

In addition, for single row, multi shot images, it gives the option to shoot in landscape instead of portrait, shooting less images to get to the whole, if one believes they might never need the extra size.

Finally, because I do like to print big, I have always been a frame filler.  While it has served me well, there are times that I have limited the use of a particular image because I had no crop/aspect change space around my subject.  Sometimes I can work it in Photoshop and sometimes not.

[Dan Wells]

Like it or not, one of the BIG drivers in high pixel count cameras is getting the pixels in the long dimension up high enough to allow 4k video by binning (3840x2=7680, ideally 4096x2=8192). Since video is a long, skinny frame, the vertical dimension will automatically work if there are enough horizontal pixels. Beyond that, I'd imagine that they might try and use existing designs, or at least common technology - how's the DR on the Sony 24MP APS-C sensor (which becomes 54 MP at full frame?

Dan

[dwswager]

Like it or not, one of the BIG drivers in high pixel count cameras is getting the pixels in the long dimension up high enough to allow 4k video by binning (3840x2=7680, ideally 4096x2=8192). Since video is a long, skinny frame, the vertical dimension will automatically work if there are enough horizontal pixels. Beyond that, I'd imagine that they might try and use existing designs, or at least common technology - how's the DR on the Sony 24MP APS-C sensor (which becomes 54 MP at full frame?

Dan

I own the APS-C 24MP D7100.  This sensor is about 1 year older than the D810, but it's performance is consistent.  The D810 outpaces it by 1EV at base ISO because the base ISO of the D810 is 64 while for the D7100 is 100. 

Bottom Line: Given the same technology, smaller pixel size will mean somewhat less DR and high ISO performance.  It's simple matter of less energy entering each pixel and less signal generated with a somewhat constant electronics noise for both. 

[BernardLanguillier]

I own the APS-C 24MP D7100.  This sensor is about 1 year older than the D810, but it's performance is consistent.  The D810 outpaces it by 1EV at base ISO because the base ISO of the D810 is 64 while for the D7100 is 100. 

As a side comment, I believe that the D7100 uses a Toshiba sensor as opposed to the Sony part used in the D810.

The measured DR level of the Toshiba part is similar to the Sony 24mp part (and significantly ahead of any Canon sensor till date), but there is some banding in the deep shadows of the Toshiba sensor.

Cheers,
Bernard

[shadowblade]

Bottom Line: Given the same technology, smaller pixel size will mean somewhat less DR and high ISO performance.  It's simple matter of less energy entering each pixel and less signal generated with a somewhat constant electronics noise for both. 

That's on a per-pixel basis.

Normalised to a set standard resolution (or print size), the performance will be the same, assuming gapless microlenses.

[barryfitzgerald]

I would be surprised if many required such a high pixel count I find 24mp overkill for most shooting conditions 50mp isn't something most would need though it will sell cameras
It doesn't matter what you give some folks 50mp or 100mp they will always want more, hence the term "boy racer" no matter how fast the car is 3mph extra is so much faster except it isn't not really

I doubt you will find the move from 36mp to 50mp that big the more mp you have to start with the less you see the extra pixels it's quite small really. 16 to 24mp is barely noticeable even with good primes we're at a level where resolution simply isn't an issue for 97% of shooters.

[shadowblade]

I would be surprised if many required such a high pixel count I find 24mp overkill for most shooting conditions 50mp isn't something most would need though it will sell cameras
It doesn't matter what you give some folks 50mp or 100mp they will always want more, hence the term "boy racer" no matter how fast the car is 3mph extra is so much faster except it isn't not really

I doubt you will find the move from 36mp to 50mp that big the more mp you have to start with the less you see the extra pixels it's quite small really. 16 to 24mp is barely noticeable even with good primes we're at a level where resolution simply isn't an issue for 97% of shooters.

Depends how big you print.

My large prints with a 36MP sensor (either single-frame or shift-stitched panoramas using a tilt-shift lens) are noticeably more detailed than my older ones with a 21MP sensor and the same lenses.

If you shoot for web or for small print sizes, you may not notice a difference. But it won't be worse, either (even the increased file size can be mitigated by pixel binning) and the greater resolution gives you more options as to what you can do with the image.

It's like anything else in photography - it makes no difference to you unless you're already operating at the limits.

Resolution matters to me because I print at huge sizes. I'm operating at the limit. If you're only publishing to web or making small prints, you're not at the limit and any improvement makes no difference to you.
Low-ISO dynamic range matters to me because I'm already pushing the limits at both ends when shooting high-contrast landscapes. If you shoot in studio and can control your lighting, or mainly shoot in limited-DR situations, it may not matter to you because you're not operating at the limit - if your scene only contains 5 stops of DR, it makes no difference whether your sensor can record 10 or 15 stops.
Colour gamut matters to me because I shoot very colourful scenes. If you shoot in black and white, you're not encountering the limits, so any improvement has no impact on you.
For most purposes, AF doesn't matter to me because it's not even switched on, except on the rare occasions I shoot action. It's nice to have the capability in the camera, though, so I don't have to drag around extra bodies for combined landscape/wildlife trips. If you mainly shoot action, you're at the limit and any improvement makes a difference.
High ISO capability doesn't matter to me because I operate well below the limits of performance - base ISO for landscapes, up to maybe ISO 1600 (very occasionally 3200) for wildlife in dim conditions. If you shoot live music or stage performances, then you're operating at the limit and every extra stop makes a huge difference.
Frame rate doesn't matter to me because I usually shoot at less than 1 frame a second - shooting at that rate, it makes no difference whether the camera can shoot at 4fps or 14fps. If you're shooting sports, then you're operating at the limit and every extra frame per second makes a difference.

[dwswager]

I would be surprised if many required such a high pixel count I find 24mp overkill for most shooting conditions 50mp isn't something most would need though it will sell cameras
It doesn't matter what you give some folks 50mp or 100mp they will always want more, hence the term "boy racer" no matter how fast the car is 3mph extra is so much faster except it isn't not really

I doubt you will find the move from 36mp to 50mp that big the more mp you have to start with the less you see the extra pixels it's quite small really. 16 to 24mp is barely noticeable even with good primes we're at a level where resolution simply isn't an issue for 97% of shooters.

There are advantages to more resolution, beyond just printing big.

But the customer demand curve is certainly not linear when it comes to resolution.  As MP count increases, the demand will drop will much faster.  Camera manufacturers know this.  Some of it will be due to the performance, file size penalty.  Some of it just lack of need and cost.  I suspect that Nikon outsold their D800/D800e projections though, which is why the D810 appeared so swiftly. 

As I said before, I bought the 36MP D810 over the 24MP D750 and I'm glad I did.  The extra MP, while not always neccessary for general use, has its advantages in flexibility.  But for general purpose use, I probably would have waived off of a 48-50MP camera.   Don't know where I sit on the curve, but seeing that tons of people have gone the D750 route, at least partly due to being scared of 36MP, I figure I'm further than most.

[dwswager]

As a side comment, I believe that the D7100 uses a Toshiba sensor as opposed to the Sony part used in the D810.

The measured DR level of the Toshiba part is similar to the Sony 24mp part (and significantly ahead of any Canon sensor till date), but there is some banding in the deep shadows of the Toshiba sensor.

Cheers,
Bernard

I have no idea who manufacturers the sensor in the D7100, but I find that it and the omission of the OLPF make it the sharpest APS-C camera I have run across.  And I have not experienced significant shadow banding...thank goodness.  I would recommend it to folks as the closest APS-C camera you will come to the D810 on a budget!  And right now you can get a Nikon refurb for like $798.  I see rumors of a D7200 which likely has a new sensor, but other than more buffer (same 6fps speed) and flippy screen, doesn't appear to offer much...at least for stills.

[NancyP]

Sadly, the Nikon lens universe doesn't have a lightweight inexpensive fast-AF prime supertelephoto like the Canon EF 400 f/5.6L no-IS, which is a joy to use with birds in flight or on hikes.

[dwswager]

Sadly, the Nikon lens universe doesn't have a lightweight inexpensive fast-AF prime supertelephoto like the Canon EF 400 f/5.6L no-IS, which is a joy to use with birds in flight or on hikes.

Are you talking about this thing?

While weight is relative, not sure I would call 2.75lb light, nor f/5.6 fast.

Might try the new Nikon 300mm f/4 VR at 1.6lb with the TC-14EIII 1.4x Teleconverter.



The 300mm w/o the TC:

[DeanChriss]

While weight is relative, not sure I would call 2.75lb light, nor f/5.6 fast.

Weight aside, it's always preferable to have a prime lens of the right focal length rather than a shorter lens plus a teleconverter that gives the same focal length. Lenses with teleconverters are noticeably slower to focus and have noticeably lower contrast and resolution than the lens alone. Resolution of a high quality lens and teleconverter is better than cropping the image, but not by much. The slower focus is a killer when photographing birds. There are plenty of great lenses that are faster, like the Nikon 400mm f/2.8 at 10.2 pounds and a Canon 400 f/2.8 at 8.5 pounds, but you can't easily hand hold them. Hand holding a 2.75 pound lens with camera attached is no problem.