50MP FF Canons incoming...

D.P. · 25 Nov 2014 at 14:19

maybe
http://www.canonrumors.com/2014/11/another-50mp-ff-dslr-mention-cr2/

They really are taking their time. Canon has been testing 32-48MP sensors in the wild since before the Nikon D800 was even released. Interesting that CR doesn't have much info on a 5Dmk4 or 1DX replacement yet.

I'm not sure it makes sense to release 2 bodies that only differ by the presence of the low-pass filter. Nikon did this with the D800 but it caused more issues than it helped, so with the D810 they dropped to a single model without the LP filter since it is not needed for such resolutions. You add a whole new SKU, additional manufacturing/shipping/ordering complexity, marketing issues, and some subtitles relating to AF and calibration. The D810 actual did have 2 AA filters but arranged to canal each other out. The reason for that was to ensure identical focusing and sensor depth compared to the non-E model. DSLR Photographers have now accepted the fact no AA filter is required at these resolutions so there is no need to offer models with a filter. Not adding a filter is actually cheaper (they are surprisingly quite expensive).

Will be interesting at CES and CP+ to see what gets released.

Phal · 26 Nov 2014 at 13:38

Yeah everything is still pretty vague! Won't be for me but it'll be interesting to see what they come out with judging by some of the stats they found when looking at the 7d2 sensor.

SDK^ · 26 Nov 2014 at 16:12

It's about time.
Canon had a 12mp 1Ds out in 2002 so MP progress has been very slow over the last 12 years.

Phal · 26 Nov 2014 at 17:02

SDK^ said:
It's about time.
Canon had a 12mp 1Ds out in 2002 so MP progress has been very slow over the last 12 years.

As most people will tell you though, MP isn't everything and there needs to be lenses capable of resolving that much detail too.

Can have the best sensor in the world but it'll be wasted behind an average lens.

greedy123 · 26 Nov 2014 at 17:18

Phal said:
As most people will tell you though, MP isn't everything and there needs to be lenses capable of resolving that much detail too.

Can have the best sensor in the world but it'll be wasted behind an average lens.

All too true. This video makes that point in an 'amusing'ish way.

D.P. · 26 Nov 2014 at 19:06

Phal said:
As most people will tell you though, MP isn't everything and there needs to be lenses capable of resolving that much detail too.

Can have the best sensor in the world but it'll be wasted behind an average lens.

Not necessarily. To accurately replicate an analogue signal with a digital representation you need to sample at least twice the Nyquist frequency.

SDK^ · 26 Nov 2014 at 19:37

Phal said:
As most people will tell you though, MP isn't everything and there needs to be lenses capable of resolving that much detail too.

Can have the best sensor in the world but it'll be wasted behind an average lens.

Yup, i'm fully aware of that. Was going to include a disclaimer in my post but thought it not necessary.

If Canon have poor lenses, like you implied, then they need to sort that too.
It's no good hiding behind the 'lens can't resolve the detail' tag line forever.

Phal · 26 Nov 2014 at 20:06

D.P. said:
Not necessarily. To accurately replicate an analogue signal with a digital representation you need to sample at least twice the Nyquist frequency.

Not quite sure what you mean there? All I know is that any chain is only as good as its weakest link.

SDK^ said:
Yup, i'm fully aware of that. Was going to include a disclaimer in my post but thought it not necessary.

If Canon have poor lenses, like you implied, then they need to sort that too.
It's no good hiding behind the 'lens can't resolve the detail' tag line forever.

I don't think anyone says Canon have poor lenses? What I meant was that lenses have a finite amount of resolution.

D.P. · 26 Nov 2014 at 20:38

Phal said:
Not quite sure what you mean there? All I know is that any chain is only as good as its weakest link.

But that isn't actually true in Shannon-sampling theory when recording discrete signals from continuous signals.

You need to record at twice the Nyquist limit

Phal · 26 Nov 2014 at 20:51

D.P. said:
But that isn't actually true in Shannon-sampling theory when recording discrete signals from continuous signals.

You need to record at twice the Nyquist limit

So the quality of the lens is irrelevant? Why have I wasted so much money on expensive lenses?

I still have no idea why you're talking about these people (Shannon and Nyquist

). I know I could do some googling to find out but it won't change reality which is that I'm sure that if there was a way around having good glass then someone would be making money from it.

Yes, a higher resolution is probably better but is twice the resolution twice as good? I'd imagine not... Like most things, there are diminishing returns so the object is to find the best resolution for the lenses you have and for the cameras you have.

gjc10212 · 26 Nov 2014 at 21:18

At what resolution is a canon L, e.g, 85 f/1.2 lens out-resolved?

D.P. · 27 Nov 2014 at 02:42

Phal said:
So the quality of the lens is irrelevant? Why have I wasted so much money on expensive lenses?

I still have no idea why you're talking about these people (Shannon and Nyquist ). I know I could do some googling to find out but it won't change reality which is that I'm sure that if there was a way around having good glass then someone would be making money from it.

Yes, a higher resolution is probably better but is twice the resolution twice as good? I'd imagine not... Like most things, there are diminishing returns so the object is to find the best resolution for the lenses you have and for the cameras you have.

I didn't say that glass is irrelevant, I said that even with poor glass you need a sensor at twice the linear spatial resolution of the lens in order to fully capture all the detail that the lens can resolve without experiencing aliasing. A lens with a native resolution of say 10Mp will thus require a 40MP sensor to fully capture all detail. And that ignites things like the color filter array. A 10MP sensor only has 2.5MP of red pixels. Ideally you need to sample all color frequencies at every pixel (e.g sigma foveon), this is easily achieved with a simply group of 2G1R1B pixels into a single full color pixel. Thus to fully capture that 10MP lens you probably want 120MP Bayer sensor which you down-sample in post processing.

For sure doubling the resolution of the sensor won't make photos "twice as good" for whatever definition of good you want. And there are difficulties having 100+ MO sensor currently in terms if bandwidth.

My main point is that a 50Mp is a great step up from a 20-30MP sensor, and is actually a far low spatial frequency than exist in almost all cameras. If a modern smartphone sensor was scaled to FF size then it would gigapixels in resolution and outperform the current FF sensors in terms of dynamic range and high ISO performance. APS-C sensors are 24MP, which is around 60MP if FF sized. A 50MP sensor would be lower spatial resolution than current commonly used DSLR sensors.

robj20 · 28 Nov 2014 at 15:50

What's file size going to be like though, my 36MP come out at around 32-40MB each.

Phal · 28 Nov 2014 at 18:28

robj20 said:
What's file size going to be like though, my 36MP come out at around 32-40MB each.

Well that's another factor. It's not just about having big enough CF/SD cards and enough disk space to store them all in but the camera itself needs to be able to process all that data quickly and shuffle it around inside itself to the cards. Need decent buffers etc!