*UPDATE* Canon Rebel T4i/650D on June 8 [CR3]

[jrista]

jrista:

I enjoy your posts and you obviously have a wealth of tech understanding.
As far as the 7d softness debate. I too, borrowed a 7d for a few days, and compared it with my t2i. The out of camera images appeared softer / fuzzier. This mystified me... as the 7d is supposed to be far superior. Could you explain again in terms for laymen, why this is - and how I can make the 7d's jpgs appear sharp ?

The 7D ships with older processing algorithms than the T2i, and its default sharpness setting is either 1 or 0. Crank it up to 3, and that should fix the problem. Also, Canon lenses do usually need to be micro focus adjusted for each body. When it comes to rebels, its often luck of the draw...and you might have received a body and lens that were relatively closely matched. The same lens may not be as ideally matched to the 7D, however it does offer micro focus adjustment, so I would do some tuning.

 

[jrista]

Anyone curious if this is still just CD-AF, or possibly some kind of FPPD-AF?? Focal-plane phase detection is a critical thing for high performance mirrorless AF. I'm rather curious if Canon will demonstrate a solid willingness to compete with the likes of Fuji and Nikon in the arena of high speed, highly capable, multi-point AF in their first mirrorless entrant...or whether they will cripple it with some kind of vastly inferior CD-AF.

How Canon play's their mirrorless card(s) will be, IMO, a clear indication of their willingness or capability to compete in the new arena of digital camera equipment. FPPD-AF would indicate they ARE willing to compete, at least to keep up with the competition. A mirrorless entrant with FPPD and a high DR 14-bit sensor would indicate they are truly serious about competing and innovating into new markets, and serving their customers quality gear. Some form of Contrast-based AF would indicate they have lost the ability or interest in solidly competing and innovating into new markets against fierce competition. A mirrorless entrant with CD-AF and the same kind of "Canonized" sensor with high low-ISO read noise and poor DR would be seriously worrying...

Maybe the 650D's continuous live AF will be an indication of things to come? A hint at Canon's competitive prowess?

rest assured, it will be only plain-vanilla CD-AF in the 650D and most likely still slower than CD-AF in recent mFT mirrorless cameras.

Canon's "technical prowess" means, they are finally offering "continuous/Servo-AF" in video mode, probably also face-detection for tracking moving head-shots across the frame ... if and when conditions [lighting, contrast, speed, direction, predictability of movement] are so good and simple, that anyone could keep it manually in focus as well. And even that feature coms years after Canon's competitors have implemented it.

I myself have been waiting for hybrid, 2-stage AF systems for a long time. FPPD for "fast and rough" stage one focussing, followed by ultrafast CD-AF for stage 2 ... "precision focusing". Closest to that concept is the Nikon 1, which currently is the only camera on the market with FPPD. Presumably Nikon is sitting on a bunch of pretty valuable patents by now. But up to now, even Nikon has not managed to implement FPPD on APS-C or FF sensors.

Overall, Canon is getting increasingly punished for not being a true innovator. During the last 4 years they have increasingly become a pure marketing (-differentiation!) company. Technically all of their cameras are at least 1 step behind state of the art in virtually every aspect - from sensor to FPPD to hybrid viewfinders ... only their in-cam jpg engine(s) do very well compared to competitive products.

I also seriously doubt, the 650D will have a really kick-ass new Canon sensor with significantly higher DR [= with better dark noise specs]. I expect only a tiny, incremental update on the current 18 MP APS-C sensor (as used in 7D, 550D, 60D, 600D).

What's worse, by now I am fairly convinced that Canon is UNABLE - as opposed to just UNWILLING - to come up with new CMOS imaging sensors that are fully competitive with Nikon's (/Sony) latest sensors. The Canon G1X was the latest disappointing proof to that. Definitely NOT terrible but rather "pretty good" ... meaning: about 1 step below current "gold standard" on the market.

You've generally voiced my concerns. Over the last four years, its been clearly demonstrated that significant gains in sensor IQ can and indeed have been made. Read noise does not need to be extremely high (like the 33+ electrons in the 5D III, one of the highest I've ever seen!), dynamic range can be pushed to the limits, pattern noise can be pretty much eliminated. Even if Canon had to pay some royalties for use of a patent or two, or needed to buy up a few companies with patents to get competitive again, they really need to. They are obviously incapable of innovating in the current market, and its really their customers who are going to start suffering as time moves on. I'm just waiting for SoNikon to move to full 16-bit CMOS sensor design, which, since their ADC's are on-die, would mean they could quadruple their dynamic range AGAIN, putting them around four stops better than Canon. That...well, it would be kind of a game ender, and people with large investments in Canon gear might indeed be stuck with them (who would buy used Canon gear when the vast bulk of the competition...anyone who uses a Sony Exmor sensor in their DSLRs, MILCs, etc. have vastly superior IQ, focal-plane phase-detection AF, etc. etc.).

BTW, regarding FPPD...I believe it is Fuji who has the patents on that. They were the first to produce a sensor with that style of AF system. It may not even be Fuji...I believe the concept was originally introduced in a research paper several years ago, in which case who knows who owns the patents. Whoever they are, they are brilliant, and deserve to rake in the royalties as it becomes a de-facto standard way to handle autofocus off in the future.

 

[dr croubie]

Anyone else notice that this is different to the Rebel T4i on the CR front page?
Notably, the absence of the 4 microphone-holes above the nameplate...

 

[wickidwombat]

in camera sharpening only applies to jpg ooc, compare raw vs raw to get a real idea if the 7D is softer or not

 

[AvTvM]

well, I have to give it to Canon: they really prove me wrong this time! :-[

Now they are first up with a 2-stage Hybrid-AF system in a DSLR, with FPPD-AF for fast focusing and CD-AF for final precision in the 650D ... I am very positively surprised! Plus the first 2 lenses with matching AF-drive. Of course this also means the sensor is new, although they kept it at 18 MP - which I consider a great decision. Will be interesting to see, how much progress they made on read noise!

Looks like they might be getting their act together ... and maybe we are going to get a decent or even great Canon mirrorless soon!

 

[nentraC]

Anyone curious if this is still just CD-AF, or possibly some kind of FPPD-AF?? Focal-plane phase detection is a critical thing for high performance mirrorless AF. I'm rather curious if Canon will demonstrate a solid willingness to compete with the likes of Fuji and Nikon in the arena of high speed, highly capable, multi-point AF in their first mirrorless entrant...or whether they will cripple it with some kind of vastly inferior CD-AF.

How Canon play's their mirrorless card(s) will be, IMO, a clear indication of their willingness or capability to compete in the new arena of digital camera equipment. FPPD-AF would indicate they ARE willing to compete, at least to keep up with the competition. A mirrorless entrant with FPPD and a high DR 14-bit sensor would indicate they are truly serious about competing and innovating into new markets, and serving their customers quality gear. Some form of Contrast-based AF would indicate they have lost the ability or interest in solidly competing and innovating into new markets against fierce competition. A mirrorless entrant with CD-AF and the same kind of "Canonized" sensor with high low-ISO read noise and poor DR would be seriously worrying...

Maybe the 650D's continuous live AF will be an indication of things to come? A hint at Canon's competitive prowess?

rest assured, it will be only plain-vanilla CD-AF in the 650D and most likely still slower than CD-AF in recent mFT mirrorless cameras.

Canon's "technical prowess" means, they are finally offering "continuous/Servo-AF" in video mode, probably also face-detection for tracking moving head-shots across the frame ... if and when conditions [lighting, contrast, speed, direction, predictability of movement] are so good and simple, that anyone could keep it manually in focus as well. And even that feature coms years after Canon's competitors have implemented it.

I myself have been waiting for hybrid, 2-stage AF systems for a long time. FPPD for "fast and rough" stage one focussing, followed by ultrafast CD-AF for stage 2 ... "precision focusing". Closest to that concept is the Nikon 1, which currently is the only camera on the market with FPPD. Presumably Nikon is sitting on a bunch of pretty valuable patents by now. But up to now, even Nikon has not managed to implement FPPD on APS-C or FF sensors.

Overall, Canon is getting increasingly punished for not being a true innovator. During the last 4 years they have increasingly become a pure marketing (-differentiation!) company. Technically all of their cameras are at least 1 step behind state of the art in virtually every aspect - from sensor to FPPD to hybrid viewfinders ... only their in-cam jpg engine(s) do very well compared to competitive products.

I also seriously doubt, the 650D will have a really kick-ass new Canon sensor with significantly higher DR [= with better dark noise specs]. I expect only a tiny, incremental update on the current 18 MP APS-C sensor (as used in 7D, 550D, 60D, 600D).

What's worse, by now I am fairly convinced that Canon is UNABLE - as opposed to just UNWILLING - to come up with new CMOS imaging sensors that are fully competitive with Nikon's (/Sony) latest sensors. The Canon G1X was the latest disappointing proof to that. Definitely NOT terrible but rather "pretty good" ... meaning: about 1 step below current "gold standard" on the market.

You've generally voiced my concerns. Over the last four years, its been clearly demonstrated that significant gains in sensor IQ can and indeed have been made. Read noise does not need to be extremely high (like the 33+ electrons in the 5D III, one of the highest I've ever seen!), dynamic range can be pushed to the limits, pattern noise can be pretty much eliminated. Even if Canon had to pay some royalties for use of a patent or two, or needed to buy up a few companies with patents to get competitive again, they really need to. They are obviously incapable of innovating in the current market, and its really their customers who are going to start suffering as time moves on. I'm just waiting for SoNikon to move to full 16-bit CMOS sensor design, which, since their ADC's are on-die, would mean they could quadruple their dynamic range AGAIN, putting them around four stops better than Canon. That...well, it would be kind of a game ender, and people with large investments in Canon gear might indeed be stuck with them (who would buy used Canon gear when the vast bulk of the competition...anyone who uses a Sony Exmor sensor in their DSLRs, MILCs, etc. have vastly superior IQ, focal-plane phase-detection AF, etc. etc.).

BTW, regarding FPPD...I believe it is Fuji who has the patents on that. They were the first to produce a sensor with that style of AF system. It may not even be Fuji...I believe the concept was originally introduced in a research paper several years ago, in which case who knows who owns the patents. Whoever they are, they are brilliant, and deserve to rake in the royalties as it becomes a de-facto standard way to handle autofocus off in the future.

i apreciate your sensor tech knowladge and would like to read your opinion if it is a new sensor or a recikled on from 600d?
Can they add a sensor layer for the focusing sistem to a eksisting design or do thej need to design a ne on?

sorry for my bad English.

 

[jrista]

Anyone curious if this is still just CD-AF, or possibly some kind of FPPD-AF?? Focal-plane phase detection is a critical thing for high performance mirrorless AF. I'm rather curious if Canon will demonstrate a solid willingness to compete with the likes of Fuji and Nikon in the arena of high speed, highly capable, multi-point AF in their first mirrorless entrant...or whether they will cripple it with some kind of vastly inferior CD-AF.

How Canon play's their mirrorless card(s) will be, IMO, a clear indication of their willingness or capability to compete in the new arena of digital camera equipment. FPPD-AF would indicate they ARE willing to compete, at least to keep up with the competition. A mirrorless entrant with FPPD and a high DR 14-bit sensor would indicate they are truly serious about competing and innovating into new markets, and serving their customers quality gear. Some form of Contrast-based AF would indicate they have lost the ability or interest in solidly competing and innovating into new markets against fierce competition. A mirrorless entrant with CD-AF and the same kind of "Canonized" sensor with high low-ISO read noise and poor DR would be seriously worrying...

Maybe the 650D's continuous live AF will be an indication of things to come? A hint at Canon's competitive prowess?

rest assured, it will be only plain-vanilla CD-AF in the 650D and most likely still slower than CD-AF in recent mFT mirrorless cameras.

Canon's "technical prowess" means, they are finally offering "continuous/Servo-AF" in video mode, probably also face-detection for tracking moving head-shots across the frame ... if and when conditions [lighting, contrast, speed, direction, predictability of movement] are so good and simple, that anyone could keep it manually in focus as well. And even that feature coms years after Canon's competitors have implemented it.

I myself have been waiting for hybrid, 2-stage AF systems for a long time. FPPD for "fast and rough" stage one focusing, followed by ultrafast CD-AF for stage 2 ... "precision focusing". Closest to that concept is the Nikon 1, which currently is the only camera on the market with FPPD. Presumably Nikon is sitting on a bunch of pretty valuable patents by now. But up to now, even Nikon has not managed to implement FPPD on APS-C or FF sensors.

Overall, Canon is getting increasingly punished for not being a true innovator. During the last 4 years they have increasingly become a pure marketing (-differentiation!) company. Technically all of their cameras are at least 1 step behind state of the art in virtually every aspect - from sensor to FPPD to hybrid viewfinders ... only their in-cam jpg engine(s) do very well compared to competitive products.

I also seriously doubt, the 650D will have a really kick-ass new Canon sensor with significantly higher DR [= with better dark noise specs]. I expect only a tiny, incremental update on the current 18 MP APS-C sensor (as used in 7D, 550D, 60D, 600D).

What's worse, by now I am fairly convinced that Canon is UNABLE - as opposed to just UNWILLING - to come up with new CMOS imaging sensors that are fully competitive with Nikon's (/Sony) latest sensors. The Canon G1X was the latest disappointing proof to that. Definitely NOT terrible but rather "pretty good" ... meaning: about 1 step below current "gold standard" on the market.

You've generally voiced my concerns. Over the last four years, its been clearly demonstrated that significant gains in sensor IQ can and indeed have been made. Read noise does not need to be extremely high (like the 33+ electrons in the 5D III, one of the highest I've ever seen!), dynamic range can be pushed to the limits, pattern noise can be pretty much eliminated. Even if Canon had to pay some royalties for use of a patent or two, or needed to buy up a few companies with patents to get competitive again, they really need to. They are obviously incapable of innovating in the current market, and its really their customers who are going to start suffering as time moves on. I'm just waiting for SoNikon to move to full 16-bit CMOS sensor design, which, since their ADC's are on-die, would mean they could quadruple their dynamic range AGAIN, putting them around four stops better than Canon. That...well, it would be kind of a game ender, and people with large investments in Canon gear might indeed be stuck with them (who would buy used Canon gear when the vast bulk of the competition...anyone who uses a Sony Exmor sensor in their DSLRs, MILCs, etc. have vastly superior IQ, focal-plane phase-detection AF, etc. etc.).

BTW, regarding FPPD...I believe it is Fuji who has the patents on that. They were the first to produce a sensor with that style of AF system. It may not even be Fuji...I believe the concept was originally introduced in a research paper several years ago, in which case who knows who owns the patents. Whoever they are, they are brilliant, and deserve to rake in the royalties as it becomes a de-facto standard way to handle autofocus off in the future.

i apreciate your sensor tech knowladge and would like to read your opinion if it is a new sensor or a recikled on from 600d?
Can they add a sensor layer for the focusing sistem to a eksisting design or do thej need to design a ne on?

sorry for my bad English.

Um...seems like purposely botched English to me (I don't recognize any form of actual language barrier in there...possibly just REALLY BAD typing skills)...but whatever.

If your asking is the sensor the same old 18mp APS-C Canon has been spinning around the merry-go-round for about four years now, possibly. Increasing NATIVE ISO by a stop would probably require more than just an improved image processor. Moving from Digic 4 to Digic 5+ certainly improves processing horsepower, but for RAW images it wouldn't do squat from an ISO standpoint. The amplifier is in the CMOS die...per pixel. It would be kind of hard to change that without designing a new sensor.

Now, Canon does do some funky things when it comes to ISO settings...their highest ISO settings tend to be a hybrid of various forms of amplification...so I could be wrong, and the additional stop of ISO could be purely handled by the Digic 5+. If that is the case...I would expect ISO 12800 to be pretty shitty...

However that would kind of be in-line with Canon's approach to "competition" these days: cheat, cut corners, weasel their way out of the tough job of ACTUALLY INNOVATING....

 

[jrista]

well, I have to give it to Canon: they really prove me wrong this time! :-[

Now they are first up with a 2-stage Hybrid-AF system in a DSLR, with FPPD-AF for fast focusing and CD-AF for final precision in the 650D ... I am very positively surprised! Plus the first 2 lenses with matching AF-drive. Of course this also means the sensor is new, although they kept it at 18 MP - which I consider a great decision. Will be interesting to see, how much progress they made on read noise!

Looks like they might be getting their act together ... and maybe we are going to get a decent or even great Canon mirrorless soon!

Where did you read it is FPPD-AD??? From what I read, it still needed a mirror-flip for PD-AF...

 

[Loswr]

Where did you read it is FPPD-AD??? From what I read, it still needed a mirror-flip for PD-AF...

That's the 'hybrid' part of the (putatively) 'new' Hybrid CMOS sensor. In Live View and movie shooting, PDAF sensors on the CMOS image sensor provide 'coarse focus' that's refined by traditional CDAF, effectively speeding up the AF process.

 

[AvTvM]

well, I have to give it to Canon: they really prove me wrong this time! :-[

Now they are first up with a 2-stage Hybrid-AF system in a DSLR, with FPPD-AF for fast focusing and CD-AF for final precision in the 650D ... I am very positively surprised! Plus the first 2 lenses with matching AF-drive. Of course this also means the sensor is new, although they kept it at 18 MP - which I consider a great decision. Will be interesting to see, how much progress they made on read noise!

Looks like they might be getting their act together ... and maybe we are going to get a decent or even great Canon mirrorless soon!

Where did you read it is FPPD-AD??? From what I read, it still needed a mirror-flip for PD-AF...

well of course any sensor-plane AF-system can only be active, when light is shining on it.

So, yes, it is FPPD and izt is ONLY available in video/liveview mode. In addition, quick-mode-AF in live.view remains functional ... flipping the miorror breifly and using the regular phase-AF sensors at the bottom of the mirror box. Those are still there for regular stills capture.

So in essence, the 650D has 3 AF systems:
* regular phase AF for stills capture and for quick-mode AF in liveview
plus
* in-sensor FPPD-AF - only available in video/liveview mode - for "phase 1": quick focusing
* contrast AF - only available in video/ liveview - for "phase 2":- precision focusing

Also, the FPPD-AF is only available in the very center of the sensor ... even more bunched together than the traditional phase-AF fields. It is NOT available outside the center of the frame. See writeup and good illustration on dpreview:
http://www.dpreview.com/previews/canon-eos-650d-rebel-t4i/3/#Hybrid

Despite all limitations I have to give kudos to Canon for being first to market with an implementation of FPPD-AF on an APS-C sensor / in a DSLR.

It certainly looks as if at least someone in their geriatric management has realized they better get things moving a bit ... or else ... ;-)

 

[Mt Spokane Photography]

The camera seems to me to be a bif advance for a consumer model.

The vari-Angle display is very useful for overhead shots or other ones where the angle does not let you view the rear of the camera, I've had a touch screen on my sony camcorder for years, and it speeds up making the settings. Smudges happen in any event, its not a big deal.

Autofocus during video (I hope its better than my D800), pentaprism viewfinder, there seem to be more than the usual tiny upgrades.

That 40mm pancake lens might be a nice feature as well, at least to help avoid undue attention.

 

[jrista]

Where did you read it is FPPD-AD??? From what I read, it still needed a mirror-flip for PD-AF...

That's the 'hybrid' part of the (putatively) 'new' Hybrid CMOS sensor. In Live View and movie shooting, PDAF sensors on the CMOS image sensor provide 'coarse focus' that's refined by traditional CDAF, effectively speeding up the AF process.

Well, if that is the case...that the sensor has some FPPD-AF pixels in the center of the frame, then the sensor is definitely new. FPPD-AF requires half-masked pixels to work, and that would require a new design. Pretty sweet though! I'm glad Canon IS finally showing some interest in actually competing, and improving their increasingly outdated technology. The use of FPPD-AF, even if its "hybrid" like this, seems to be good news for the mirrorless front.

 

[Loswr]

The vari-Angle display is very useful for overhead shots or other ones where the angle does not let you view the rear of the camera... pentaprism viewfinder, there seem to be more than the usual tiny upgrades.

The T3i/600D had a vari-angle LCD, so no improvement there...

Also, sorry...but it's a pentamirror just like all the other Rebels. There was a mistake in the initial translation of the specs from the Japanese version.

 

[AvTvM]

Well, if that is the case...that the sensor has some FPPD-AF pixels in the center of the frame, then the sensor is definitely new. FPPD-AF requires half-masked pixels to work, and that would require a new design. Pretty sweet though! I'm glad Canon IS finally showing some interest in actually competing, and improving their increasingly outdated technology. The use of FPPD-AF, even if its "hybrid" like this, seems to be good news for the mirrorless front.

exactly!

 

[x-vision]

i apreciate your sensor tech knowladge and would like to read your opinion if it is a new sensor or a recikled on from 600d?
Can they add a sensor layer for the focusing sistem to a eksisting design or do thej need to design a ne on?

sorry for my bad English.

It's a tweaked/recycled version of Canon’s current 18mp sensor.

The hybrid AF has no bearing on the sensor.
For hybrid AF, certain pixels just need to be read out by the AF processor as well (in addition to the image processor).
But they don’t need to be special pixels or anything.

As for the higher ISO setting:
This is achieved with tweaks to the read-out logic and also by the Digic 5 processor.

 

[jrista]

i apreciate your sensor tech knowladge and would like to read your opinion if it is a new sensor or a recikled on from 600d?
Can they add a sensor layer for the focusing sistem to a eksisting design or do thej need to design a ne on?

sorry for my bad English.

But they don’t need to be special pixels or anything.

Sorry, but that is completely false. For phase-shift detection to work, you have to be able to determine a phase differential. For Focal-plane pase-shift detection AF, you have to mask off part of certain pixels in a specific way to make that possible. That allows only light from one side of the lens or the other to be actually picked up by the PD-pixels in the sensor. Much the same way a split-prism focusing screen works. If they were normal pixels, there would be no way to actually know if phase shifted, since in a standard phase-detect AF sensor, there is a special lens purposely designed to split incoming light for the purpose of detecting if there is a differential between the phase of light coming through various parts of the lens.

An explanation of Fuji's specific implementation can be found here:

http://www.dpreview.com/news/2010/8/5/fujifilmpd

 

[x-vision]

But they don’t need to be special pixels or anything.

Sorry, but that is completely false.

Aah. So, we have certain pixels on the sensor wearing 'sunglasses', figuratively speaking.
This of course makes them totally different from the other pixels. Got it.

 

[jrista]

But they don’t need to be special pixels or anything.

Sorry, but that is completely false.

Aah. So, we have certain pixels on the sensor wearing 'sunglasses', figuratively speaking.
This of course makes them totally different from the other pixels. Got it.

Aye. And those pixels remain functional for standard imaging as well. Despite the masking, they do not actually remove pixels from the normal bayer image output...its just that since a few hundred thousand, maybe a million pixels are used for FPPD-AF, vs tens of millions for the total image, the IQ impact is non-existent.

 

[moreorless]

The vari-Angle display is very useful for overhead shots or other ones where the angle does not let you view the rear of the camera... pentaprism viewfinder, there seem to be more than the usual tiny upgrades.

The T3i/600D had a vari-angle LCD, so no improvement there...

Also, sorry...but it's a pentamirror just like all the other Rebels. There was a mistake in the initial translation of the specs from the Japanese version.

You've also got by far the biggest improvement we've ever seen in the Rebel lines AF though and also the largest single improvement in FPS we've seen, before now it had been creeping up in 0.5 steps at most.

 

[Marsu42]

You've also got by far the biggest improvement we've ever seen in the Rebel lines AF though and also the largest single improvement in FPS we've seen, before now it had been creeping up in 0.5 steps at most.

Sure faster fps is fine for bracketing, but with little buffer size forget action shots.

Concerning the af: This is certainly an improvement too, but I have the 60d and I have to say the af is complete crap for servo, and for still shots the af points are mostly spread too far apart so it's focus & recompose like in the good ol' days. And with large aperture you cannot recompose because the focus shifts - no, wait, we cannot really use these lenses anyway because Canon didn't include afma :-o

As a conclusion, I guess real world 650d users will be much happier about the touchscreen & video af to record their children running around than the improved af & fps.