EOS 5D Mark IV To Use EOS C300 Mark II Technology? [CR2]

[3kramd5]

It is funny how excited all those trashing the 'Droners' are the second their is some talk about Canon possibly getting even just a trace more DR than Exmor! ;D ;D ;D ;D :

I must have missed that

 

[Diltiazem]

… this essentially works as a hardware implementation of what the Magic Lantern guys did with software, and actually takes it a bit further.

Unbelieveable, I didn't think Canon would copy any ML innovation because af sheer principle.

They didn't copy ML if this is real, simultaneous per photosite ISO read out and if all it is is the ML hack in firmware, that won't fly! That's totally fake and hits resolution.

First, Emil Martinec first talked about such a scheme years ago on DPR.
Second, I'm pretty sure Canon had a patent out on this prior to ML's dual-ISO hack too.

They did. A 2010 Canon patent talks about combining high and low ISO signals from each pixels to increase DR. A refined version of the patent was granted in January this year.
http://patft1.uspto.gov/netacgi/nph-Parser?Sect1=PTO1&Sect2=HITOFF&d=PALL&p=1&u=%2Fnetahtml%2FPTO%2Fsrchnum.htm&r=1&f=G&l=50&s1=8,928,789.PN.&OS=PN/8,928,789&RS=PN/8,928,789/

A quote from the granted patent, "In Japanese Patent Application Laid-Open No. 2010-16416, a column amplifying circuit for each of columns generates a low-gain signal and a high-gain signal from each pixel signal from an imaging device. Signals obtained by returning high-gain signals to the same gain as the gain of low-gain signals and the low-gain signals are selectively combined to increase a dynamic range while maintaining an S/N ratio. ".

 

[Diltiazem]

The mention of magiclantern doesnt meam theyre doing the every other line different iso, which was a hardware limitation.
Im fairly sure they are just reading the sensor once, duplicating the signal and pumping it through different adc's and then combining the two resulting exposures. It would be stupid and slow to try and do ML's implementation in camera since the interpolation algorithms are extremely CPU intensive and slow, unless they developed hardware process of doing that work for them in realtime, which is idiotic since that would be more work than simply developing the hardware to duplicate the sensor output which should be fairly trivial in comparison and result in a full resolution composite and be much much faster since no interpolation is required.

May be something like this:
http://patft1.uspto.gov/netacgi/nph-Parser?Sect1=PTO1&Sect2=HITOFF&d=PALL&p=1&u=%2Fnetahtml%2FPTO%2Fsrchnum.htm&r=1&f=G&l=50&s1=8,928,789.PN.&OS=PN/8,928,789&RS=PN/8,928,789

 

[lintoni]

And the moon landings were faked; the Malaysia Airlines tragedies were a false flag operation organised by the illuminati and the US govt has incontrovertible proof for the existence of ETs, but is hiding it.

And Canon doesn't cripple its cameras or time its releases... Um, yeah it does.

Uh huh. Like people whingeing on t'interwebz that a brand new video camera aimed at professional media is too expensive for every Tom, Dick and Harry to pick up and shoot their cousin's wedding with is going to make Canon leak a rumour that a potentially forthcoming possible DSLR may use some innovative technology. Because having a new £3000 DSLR will placate some internet forum muppets who wanted a video camera for tuppence.:

 

[erjlphoto]

Sound promissing. Might turn out to be a real Woo-Hoo camera

 

[pedro]

What will this mean for high ISO sensitivity? Cleaner images above 6400 to 51200 or will a focus on DR prevent from better sensibility to high ISO. Sorry, I am no tech knowledge amateur with a 5D3 ;-)

 

[PureClassA]

Using this tech on the 5D4 seems likely, the 1DX2 a certainty, and the 5DSR damn well should have been. If this was the result of a significant re-tooling of fab process, it explains the wild pricing of this camera and it also means going forward Canon will use it in new Pro bodies. But assuming this was something that only worked well with DPAF, that nixes the use on the 5DSR

 

[LonelyBoy]

And the moon landings were faked; the Malaysia Airlines tragedies were a false flag operation organised by the illuminati and the US govt has incontrovertible proof for the existence of ETs, but is hiding it.

And Canon doesn't cripple its cameras or time its releases... Um, yeah it does.

As opposed to Nikon, Sony, and everyone else in every industry?

 

[scottkinfw]

Sorry for the uninformed question.

In reading through the announcement and this thread so far, I'm trying to figure out what this might all mean for me the one using the camera.

So my question for you sage people is, what do you think the actual improvement over a 5DIII might actually be, vs. tradeoffs? Mainly interested in stills, but opinion on video would be welcome too.

Better DR
Better resolution
Faster/better focusing (I know not part of the chip tech)
Less noise
Better IQ/color/contrast?
Auto focusing on video
Faster image capture and processing?
Some magic new feature that isn't just fluff?

Thanks for taking the time.

sek

 

[LetTheRightLensIn]

What will this mean for high ISO sensitivity? Cleaner images above 6400 to 51200 or will a focus on DR prevent from better sensibility to high ISO. Sorry, I am no tech knowledge amateur with a 5D3 ;-)

It won't mean anything in regards to mid-tone SNR at high ISO. You'll just get better DR since the read noise will be miles better. Good stuff if this really is for real!

 

[LetTheRightLensIn]

And the moon landings were faked; the Malaysia Airlines tragedies were a false flag operation organised by the illuminati and the US govt has incontrovertible proof for the existence of ETs, but is hiding it.

And Canon doesn't cripple its cameras or time its releases... Um, yeah it does.

As opposed to Nikon, Sony, and everyone else in every industry?

Canon has been worse about it over the last near decade or so.

 

[LetTheRightLensIn]

Sorry for the uninformed question.

In reading through the announcement and this thread so far, I'm trying to figure out what this might all mean for me the one using the camera.

So my question for you sage people is, what do you think the actual improvement over a 5DIII might actually be, vs. tradeoffs? Mainly interested in stills, but opinion on video would be welcome too.

Better DR
Better resolution
Faster/better focusing (I know not part of the chip tech)
Less noise
Better IQ/color/contrast?
Auto focusing on video
Faster image capture and processing?
Some magic new feature that isn't just fluff?

Thanks for taking the time.

sek

This wouldn't apply to any of that other than the DR part and better IQ (for shots that need it) part (because of the better DR).

The other stuff is all based on marketing decisions and/or other tech and has nothing to do with this at all.

But you can pretty much be assured that the MP count won't match the 5Ds just for the fact that why release the 5Ds if you just come up with the same sensor, only better, like three months later? i think it's got to hint, if this rumor is true of course, that they will hold the MP count way down on the 5D4, perhaps to about the same as is on the 5D3. If not, the 5Ds would be truly pointless then. Nothing in this tech itself says that the MP has to be lower though. It would just appear that Canon didn't want to deliver too much so they held the new tech back from the 5Ds. The 5Ds decision is bizarre IMO. That would be a lot of crippling even for the Canon of today. Of course this rumor has to be true for any of this to apply.

Anyway, things are hinting that the 5D4 might be one heck of a stills camera other than for MP count, possibly better than D810 in all ways other than for MP (and maybe color filter array).... IF this rumor is true combined with some other things. But if so, sounding potentially very impressive for the 5D4 for stills! As far video it remains to be seen whether they deliver or cripple it. If they deliver there too.... .

 

[rrcphoto]

… this essentially works as a hardware implementation of what the Magic Lantern guys did with software, and actually takes it a bit further.

Unbelieveable, I didn't think Canon would copy any ML innovation because af sheer principle.

They didn't copy ML if this is real, simultaneous per photosite ISO read out and if all it is is the ML hack in firmware, that won't fly! That's totally fake and hits resolution.

First, Emil Martinec first talked about such a scheme years ago on DPR.
Second, I'm pretty sure Canon had a patent out on this prior to ML's dual-ISO hack too.

They did. A 2010 Canon patent talks about combining high and low ISO signals from each pixels to increase DR. A refined version of the patent was granted in January this year.
http://patft1.uspto.gov/netacgi/nph-Parser?Sect1=PTO1&Sect2=HITOFF&d=PALL&p=1&u=%2Fnetahtml%2FPTO%2Fsrchnum.htm&r=1&f=G&l=50&s1=8,928,789.PN.&OS=PN/8,928,789&RS=PN/8,928,789/

A quote from the granted patent, "In Japanese Patent Application Laid-Open No. 2010-16416, a column amplifying circuit for each of columns generates a low-gain signal and a high-gain signal from each pixel signal from an imaging device. Signals obtained by returning high-gain signals to the same gain as the gain of low-gain signals and the low-gain signals are selectively combined to increase a dynamic range while maintaining an S/N ratio. ".

to be correct just before someone cries .. why has canon sat on it for FIVE years.. that patent took by the looks of it, 3 years+ to get approved and only approved in 2013.

canon also had a dual ramp ADC sensor patent, but that was not approved until January 2015.

 

[Marsu42]

They didn't copy ML if this is real, simultaneous per photosite ISO read out and if all it is is the ML hack in firmware, that won't fly! That's totally fake and hits resolution.

It would be great if Canon would manage dual pixel readout, because this would allow even more than 15ev (ML has to have some overlap between iso values for interpolation).

That being said, I encourage everyone to try the ML solution: It sounds hackish on paper and the workflow is a hassle, but you loose much less resolution in practice unless you shoot test charts and each white-black line is exactly one scanline on the sensor.

Alex has worked a lot on the interpolation algorithm since the first release, and personally I would have never that the results are as close to "vanilla" shots as they are. After a lot of experimentation and comparisons, I shoot much more confidently with dual_iso these days.

 

[emko]

They didn't copy ML if this is real, simultaneous per photosite ISO read out and if all it is is the ML hack in firmware, that won't fly! That's totally fake and hits resolution.

It would be great if Canon would manage dual pixel readout, because this would allow even more than 15ev (ML has to have some overlap between iso values for interpolation).

That being said, I encourage everyone to try the ML solution: It sounds hackish on paper and the workflow is a hassle, but you loose much less resolution in practice unless you shoot test charts and each white-black line is exactly one scanline on the sensor.

Alex has worked a lot on the interpolation algorithm since the first release, and personally I would have never that the results are as close to "vanilla" shots as they are. After a lot of experimentation and comparisons, I shoot much more confidently with dual_iso these days.

Wow really you use dual_iso? for what? everyone here says you only need 12 stops of DR or you are shooting with a lens cap on.

If this is true and has more DR then Sony sensors they will now say how weak the SONY sensor is.

 

[LetTheRightLensIn]

How is Canon going to do 15 stops of DR with a 14bit CR2 file?

Note that it is quite possible that the 15 stops of DR for the C300 Mk II is via C-Log, which is different to CR2 raw files.

Will this mean a 16bit CR2 or a 16bit CR3?

It might not do 15 stops at 100% view, maybe just at 8MP normalization. The D810 even at ISO64 doesn't do 14.8 stops at 100% view either. If this can also do 15 stops at say 20MP then super wow. They'd obviously need to go to 16 bit files then and high-quality 16bit converters. Not sure if that will happen yet or if it will be quite that good. Whatever the case, if the rumor is true, it should at least be reasonably close to as good as Exmor and maybe it could even be the same or even better.

 

[LetTheRightLensIn]

They did. A 2010 Canon patent talks about combining high and low ISO signals from each pixels to increase DR. A refined version of the patent was granted in January this year.
http://patft1.uspto.gov/netacgi/nph-Parser?Sect1=PTO1&Sect2=HITOFF&d=PALL&p=1&u=%2Fnetahtml%2FPTO%2Fsrchnum.htm&r=1&f=G&l=50&s1=8,928,789.PN.&OS=PN/8,928,789&RS=PN/8,928,789/

A quote from the granted patent, "In Japanese Patent Application Laid-Open No. 2010-16416, a column amplifying circuit for each of columns generates a low-gain signal and a high-gain signal from each pixel signal from an imaging device. Signals obtained by returning high-gain signals to the same gain as the gain of low-gain signals and the low-gain signals are selectively combined to increase a dynamic range while maintaining an S/N ratio. ".

Canon's signal to noise ration at low ISO is shocking and I'm not sure why they'd want to maintain that.

They are not referring to read noise, they mean like DxO SNR middle gray tests, nothing changes there, where Canon is perfectly fine already, but the DR becomes much better since they fix the dark tone SNR.

 

[LonelyBoy]

Canon has been worse about it over the last near decade or so.

Market leaders always are. Intel was king of it back in the day, and now is again. Every for-profit company will do it when they can afford to. If anyone doesn't, it's because they're desperate, or they like to not run the company well.

 

[BeenThere]

How is Canon going to do 15 stops of DR with a 14bit CR2 file?

Note that it is quite possible that the 15 stops of DR for the C300 Mk II is via C-Log, which is different to CR2 raw files.

Will this mean a 16bit CR2 or a 16bit CR3?

It might not do 15 stops at 100% view, maybe just at 8MP normalization. The D810 even at ISO64 doesn't do 14.8 stops at 100% view either. If this can also do 15 stops at say 20MP then super wow. They'd obviously need to go to 16 bit files then and high-quality 16bit converters. Not sure if that will happen yet or if it will be quite that good. Whatever the case, if the rumor is true, it should at least be reasonably close to as good as Exmor and maybe it could even be the same or even better.

Maybe two 14 bit converters for dual ADC output from sensor (one at base ISO and one amplified) combined off-board into a single 16 bit file. Overlapping bins could be averaged to reduce noise in mid-tones.

 

[Orangutan]

Let me summarise the situation with said people:
Any measurement or statistic where Canon has the better number or statistic is important (until Canon is behind) and vital and any measurement or statistic where Canon trails is unimportant (until Canon is in front.)

I don't think that's accurate. Any characteristic for which Canon is not "behind" far enough to overcome the many reasons to own a Canon system is somewhat important, but not vital. Any characteristic for which Canon is "ahead" is another good reason to own a Canon system.

As Neuro (and others) keep saying: it's a system not a sensor. Of course we'd all like to see a "better" sensor in the Canon system; however, until another system is sufficiently superior to justify the expense and hassle of change, we just use and enjoy what we have.