Possible Canon EOS 5D Mark IV Spec Talk [CR2]

[Lee Jay]

Every time there is a megapixel advance some people claim "no one needs that many pixels etc. etc. etc."; they said this at 3 mp, 5mp, 10mp etc. More pixels is always better if you don't have to give up anything (not sure yet what the 5ds will require us yield).

ISO 6,400 limit with expansion to ISO 12,800 (same as the pocketable S120 with 1/1.7" sensor, by the way).

I suspect that the quality will not be the same : : :

Right - which is why the 5DS should go much higher, not the same.

 

[bdunbar79]

I agree with Lee Jay.

I think sensor size is a heck of a lot more important than pixel size in determining ISO performance. You can always increase the QE and smaller pixels already have less read noise. Noise = SQRT #photons. If I can collect and use more photons with smaller pixels (admittedly tech improvements need made) then why not?

 

[privatebydesign]

Every time there is a megapixel advance some people claim "no one needs that many pixels etc. etc. etc."; they said this at 3 mp, 5mp, 10mp etc. More pixels is always better if you don't have to give up anything (not sure yet what the 5ds will require us yield).

ISO 6,400 limit with expansion to ISO 12,800 (same as the pocketable S120 with 1/1.7" sensor, by the way).

I suspect that the quality will not be the same : : :

Right - which is why the 5DS should go much higher, not the same.

As always, comparing completely different technologies and extrapolating, you really should know better.

The S120 has a back-illuminated sensor and is not comparable to the current DSLR sensors.

 

[privatebydesign]

I agree with Lee Jay.

I think sensor size is a heck of a lot more important than pixel size in determining ISO performance. You can always increase the QE and smaller pixels already have less read noise. Noise = SQRT #photons. If I can collect and use more photons with smaller pixels (admittedly tech improvements need made) then why not?

If you only talk about the same sized sensors then bigger pixels have the potential to be less noisy if built to the same tech, if for no other reason than less of the sensor area is taken up with the lithography. Once you start comparing different sized sensors the smaller sensor will always have to have better tech to even keep up. Which leads us to the BSI-CMOS tech in the P&S's which gives much better IQ than even a medium format digital on a per area basis, but as it uses such a small area the total image IQ is lower.

I have said for years, you can look through my posting history, sensor area is king, a smaller sensor with the same tech will never be able to give you the IQ of the bigger sensor, at what point that becomes noticeable, or the IQ is not 'good enough' is something each person has to judge for themselves.

 

[Lee Jay]

If you only talk about the same sized sensors then bigger pixels have the potential to be less noisy if built to the same tech, if for no other reason than less of the sensor area is taken up with the lithography.

Microlenses pretty much equalize that.

Smaller pixels tend to win whenever noise is shot-noise dominated. Shot noise is pretty much pixel size invariant for the the same size final image, but noise reduction software beats out the block averaging done by larger pixels. If you're read-noise dominated, bigger pixels can win, but smaller pixels tend to have less read noise too, partially or (in some cases) totally eliminating that advantage. But you tend to be read-noise dominated only at very high ISOs in the 5 to 6 digit range or more.

 

[privatebydesign]

If you only talk about the same sized sensors then bigger pixels have the potential to be less noisy if built to the same tech, if for no other reason than less of the sensor area is taken up with the lithography.

Microlenses pretty much equalize that.

Clearly they don't, but why let honest actual objective real world results get in the way of a good argument and theoretical discussion?

If you compare any same generation and tech sensors the ones with the smaller pixels have more noise per sensor area. Well that has been true in every instance of a genuine fair comparison I have seen.

 

[Tugela]

Having higher pixel densities just reduces ISO performance.

No it doesn't.

Yes it does.

Someday someone needs to explain to me why this myth persists after a decade of things going the other way despite similar basic sensor performance (QE).

Because when your small pixel drowns in a sea of noise a large pixel will still be getting a signal. It is simple physics.

Way to advance an argument: "I'm right, you're wrong". I think both could do with an added [citation needed].

How about a test?

Pixels that would give you a 208MP full-frame sensor on the left, a 13MP full-frame sensor on the right. Same ISO, same f-stop, same shutter speed, same lighting, same focal length, both processed from raw using the same settings on the same converter.

When you hit the bottom of the well you have hit the bottom of the well. No amount of averaging between wells is going to make it anything other than the bottom of the well. A larger pixel is a deeper well and will always provide a better signal than a smaller pixel.

Like I said, its physics.

 

[privatebydesign]

Having higher pixel densities just reduces ISO performance.

No it doesn't.

Yes it does.

Someday someone needs to explain to me why this myth persists after a decade of things going the other way despite similar basic sensor performance (QE).

Because when your small pixel drowns in a sea of noise a large pixel will still be getting a signal. It is simple physics.

Way to advance an argument: "I'm right, you're wrong". I think both could do with an added [citation needed].

How about a test?

Pixels that would give you a 208MP full-frame sensor on the left, a 13MP full-frame sensor on the right. Same ISO, same f-stop, same shutter speed, same lighting, same focal length, both processed from raw using the same settings on the same converter.

When you hit the bottom of the well you have hit the bottom of the well. No amount of averaging between wells is going to make it anything other than the bottom of the well. A larger pixel is a deeper well and will always provide a better signal than a smaller pixel.

Like I said, its physics.

Also, as I said, that 'comparison' is between two totally different sensor technologies, one is back-illuminated and the other isn't, so the two images mean nothing.

 

[Machaon]

If these are the specs of the 5D mkiv I hope there will be two versions: this one and one with 2X (36mp) the resolution and half the FPS (5-6fps).

Isn't that the 5Ds? Or are you after a 5Ds with better ISO performance?

I shoot both a 5d3 and a d810; I shoot a variety of subjects and I can tell you, having 36 mp and then going backwards by 50% to 18 mp would be hurtfull.

That is, unless you're going from Bayer CFA to a multi-layer sensor, in which case the spatial resolving power of the 18 MP sensor may be similar...

 

[Lee Jay]

Having higher pixel densities just reduces ISO performance.

No it doesn't.

Yes it does.

Someday someone needs to explain to me why this myth persists after a decade of things going the other way despite similar basic sensor performance (QE).

Because when your small pixel drowns in a sea of noise a large pixel will still be getting a signal. It is simple physics.

Way to advance an argument: "I'm right, you're wrong". I think both could do with an added [citation needed].

How about a test?

Pixels that would give you a 208MP full-frame sensor on the left, a 13MP full-frame sensor on the right. Same ISO, same f-stop, same shutter speed, same lighting, same focal length, both processed from raw using the same settings on the same converter.

When you hit the bottom of the well you have hit the bottom of the well. No amount of averaging between wells is going to make it anything other than the bottom of the well. A larger pixel is a deeper well and will always provide a better signal than a smaller pixel.

Like I said, its physics.

Also, as I said, that 'comparison' is between two totally different sensor technologies, one is back-illuminated and the other isn't, so the two images mean nothing.

Nope. Same basic sensor technology, same generation.

 

[Lee Jay]

Having higher pixel densities just reduces ISO performance.

No it doesn't.

Yes it does.

Someday someone needs to explain to me why this myth persists after a decade of things going the other way despite similar basic sensor performance (QE).

Because when your small pixel drowns in a sea of noise a large pixel will still be getting a signal. It is simple physics.

Way to advance an argument: "I'm right, you're wrong". I think both could do with an added [citation needed].

How about a test?

Pixels that would give you a 208MP full-frame sensor on the left, a 13MP full-frame sensor on the right. Same ISO, same f-stop, same shutter speed, same lighting, same focal length, both processed from raw using the same settings on the same converter.

When you hit the bottom of the well you have hit the bottom of the well. No amount of averaging between wells is going to make it anything other than the bottom of the well. A larger pixel is a deeper well and will always provide a better signal than a smaller pixel.

Like I said, its physics.

Yeah, but you don't know any physics.

An ideal sensor would capture the location and wavelength of every incident photon individually. Think of that as an infinite number of infinitely small pixels, most of which capture no photons at all.

 

[RGF]

I like the idea of a 24-32mp sensor as it gives the flexibility of both resolution and of ISO performance. The 1Dx2 would be the high ISO champ in the future lineup, the 5Ds studio camera, and 5D4 as a general purpose all rounder.

I'd probably keep my 5D3 with a spec list like this.

I see the need for 2 5D cameras. High ISO camera, lower MP, no need to 1D build (though it would be nice if price is right) and an intermediate MP camera, similar to the Nikon D810.

Don't expect 12 FPS, that is the territory of 1Dx.

 

[adhocphotographer]

My thoughts:

Looks great to me, a slimmed down 1DX... fantastic! As for the 18mp sensor... i think this is just the 1dX sensor, and i would guess it would change for the retail... I think it will stick around the 20-23MP range with increased ISo and DR.

This looks perfect for me! The 5Ds for high MP, the 5Dx for the general and a 5Dc for the more video orientated!

This is all very exciting.... Fingers crossed that this prototype is not too far off the mark!

Price wise, i think it will be the same as the 5Ds/sr cameras. ~3600 USD.

 

[scyrene]

Every time there is a megapixel advance some people claim "no one needs that many pixels etc. etc. etc."; they said this at 3 mp, 5mp, 10mp etc. More pixels is always better if you don't have to give up anything (not sure yet what the 5ds will require us yield).

ISO 6,400 limit with expansion to ISO 12,800 (same as the pocketable S120 with 1/1.7" sensor, by the way).

I suspect that the quality will not be the same : : :

Right - which is why the 5DS should go much higher, not the same.

At that level though, can't you effectively increase the ISO in pp? There's not much difference (if any?) between doing it in camera at the boost settings or after the fact, is there? That's the impression I get from similar discussions I've seen.

 

[scyrene]

So the original quote was:

Having higher pixel densities just reduces ISO performance.

Now pixel densities have been increasing for years. And yet high ISO hasn't got worse (has it?). Clearly there is more going on here. "It's physics" isn't good enough, sorry. I think Privatebydesign's point of comparing like with like is relevant - especially as each generation of cameras has managed to increase pixel counts *and* high ISO performance (as a general trend, I guess there will be exceptions). So when we are discussing a future camera, which presumably has a more modern sensor than past models, why should we not expect this trend to continue?

Otherwise, those seeking the best low light performance would always use the oldest, lowest-resolution sensors.

I'm just trying to understand reality, I don't have an axe to grind. Surely something as seemingly basic as this can be settled objectively one way or another?

 

[zim]

My thoughts:

Looks great to me, a slimmed down 1DX... fantastic! As for the 18mp sensor... i think this is just the 1dX sensor, and i would guess it would change for the retail... I think it will stick around the 20-23MP range with increased ISo and DR.

This looks perfect for me! The 5Ds for high MP, the 5Dx for the general and a 5Dc for the more video orientated!

This is all very exciting.... Fingers crossed that this prototype is not too far off the mark!

Price wise, i think it will be the same as the 5Ds/sr cameras. ~3600 USD.

I really really hope you are correct but that would mean 4 variants of the 5D!!

What if it turns out to be
5Ds and 5Dsr for high MP and a 5Dc for the more video orientated and general

:'(

 

[bdunbar79]

This is the way I see it. And of course Lee Jay and PBD can add/correct me if needed.

To increase high ISO performance to me is to increase DR at high ISO/light-limited situations. I need to increase FWC or max signal per pixel or QE, however you look at it. You need to lower read noise. HOW you do those two things I'm not really commenting about but if you can do that you can increase S/N at high ISO. You already have less read noise with smaller pixels so why can't you increase the size or efficiency of the photodiode in the pixel? I've been in discussions about smaller parts in and around the pixel to make way for larger photodiodes, for instance. I can also see where the FWC could be more important than read noise and overcome the higher read noise by adding more signal because signal is additive whereas noise is added SQRT. So in that case, larger pixels might still win. This of course is all at equal sensor size and equal technology.

 

[RLPhoto]

I like the idea of a 24-32mp sensor as it gives the flexibility of both resolution and of ISO performance. The 1Dx2 would be the high ISO champ in the future lineup, the 5Ds studio camera, and 5D4 as a general purpose all rounder.

I'd probably keep my 5D3 with a spec list like this.

I see the need for 2 5D cameras. High ISO camera, lower MP, no need to 1D build (though it would be nice if price is right) and an intermediate MP camera, similar to the Nikon D810.

Don't expect 12 FPS, that is the territory of 1Dx.

I wouldn't buy a lower MP 5D4 than my 5D3. Plus knowing canon being canon, they'd want you to spend more money on the high ISO champ the 1Dx2 rather than a low MP 5D4. 24mp or a 32mp sensor with improved performance would make the 5D4 more General purpose than a 18mp sensor when the extra resolution is needed. (Especially since stock sites usually require at least 20mp to sell the largest size.)

 

[clifflwms]

I can believe most of this, it looks great, but I'm skeptical about the 18MP AND the Dual CFast. Why would they add a new card to the camera while removing the card that most people already have? Traditionally (At least in recent history), cameras in this range have had one SD and one CF. Why wouldn't they just replace the CF with CFast? why remove SD?

 

[Lee Jay]

Every time there is a megapixel advance some people claim "no one needs that many pixels etc. etc. etc."; they said this at 3 mp, 5mp, 10mp etc. More pixels is always better if you don't have to give up anything (not sure yet what the 5ds will require us yield).

ISO 6,400 limit with expansion to ISO 12,800 (same as the pocketable S120 with 1/1.7" sensor, by the way).

I suspect that the quality will not be the same : : :

Right - which is why the 5DS should go much higher, not the same.

At that level though, can't you effectively increase the ISO in pp? There's not much difference (if any?) between doing it in camera at the boost settings or after the fact, is there? That's the impression I get from similar discussions I've seen.

The difference is speed, metering, and verification of correct settings.

I go though this on my 5D, which has a maximum ISO of 1600, 3200 base. It's acceptable up to ISO 12,800 but using ISO 3200 and EC-2 is a pain, it's hard to use the image preview to see what you got, it doesn't work well at all in JPEG mode, and it's a hassle even in raw.

Just being able to set the exposure index you actually want to use would be a big help in practice.