I don't know if there is similar data available for say 70D. However one would think if Canon had moved to a newer tech for 70D, we would have seen a bigger difference in performance compared to its older APS-C sensors.
Thanks for the link, I had been trying to that info. No new steppers yet. So now another 6-7 months wait for 5D IV. Nikon should have the issue free D820 to go by then.
Hmm this is tragic. Reading the articel you come across the following:
Canon does have a 0.18 µm generation CIS wafer fab process, featuring a specialized Cu back end of line (BEOL) including light pipes (shown below). It is possible to speculate that Canon may be preparing to refresh its FF CIS line to supply devices for a new FF camera system. Samsung and Panasonic currently use Cu fabs to produce APS-C and micro 4/3 CIS devices. It seems that Canon is destined to do so for APS-C and perhaps ultimately FF. Part III of this series will discuss CMOSIS/STMicroelectronics’ combined effort to produce FF CIS using sub 0.18 µm design rules for the first time.
This article is 2 years old and Canon has still not moved to 0.18micron CIS for DSLR despite having the fab process, which may be used on smaller sensor cameras. You will not ever see more than the current 22MP if they stay at 0.5micron. It's what's happening on the backside of the chip that's important and 0.5micron will be too large for a 40MP camera.
Maybe reason for this is cost savings while the economic times are tough? Canon has milked the 0.5micron process for all it can and given the lower cost fo the 7D II where margins are much smaller still decided to stay with this process as 0.18 (or maybe even smaller) will cost more. However yields are better on APS-C than FF so that would help amortise costs. If we do see a high MP 5D IV it'd be safe to assume they have implemented a smaller fab process.
You guys overstate the importance of the 0.XX micron process or the sensor to begin with. Most people buying cameras don't even know what that means. Techies love to seize on numbers but often times miss the big picture in doing so.
Most of the photographers out there stopped caring about sensors around 5-6 years ago (roughly 2008) if not before then. For the typical photographer, the sensors have surpassed 'good enough' and when one of these sees the 7d2, they will take note of an autofocus that improves upon the already good autofocus of its predecessor, the improved fps, etc. The sensor might be the same, it might be slightly improved, or it might be dramatically improved, but this is likely to be little more than an afterthought to the typical user.
You guys tend to get tied up on a specific piece of hardware but that is not the whole picture even when it comes to image quality. Firmware and software and everything else that goes around that sensor is just as important, look for instance at the demonstrate-able improvements that ML has made on some of the latest sensors using dual ISO mode and so on. This should demonstrate the folly citing a number such as process geometry and saying it will never be as good, because of this -- when in reality it's just a number.
Pixel Perfect wrote:
I think we know what to expect AF wise, as it's using a 1D X class AF, but IQ was the most important improvement required IMO and there is nothing to suggest it will be much better than 70D at this stage. There is yet again now new sensor tech, just minor evolution. But we should wait for the tests to start. I'll most likely get one and what other options do we have anyway as Nikon looks to have given up on a D300s successor despite rumours of a D400. Other than sensor, specs are terrific
Basically other than shooting stills at ISO100-400 (and a little at ISO800) the sensor should be very good and other than for video image quality which have some doubts about even beyond just lacking 4k, the specs are awesome. So if you don't shoot video and can ignore the high DR scenes and tend to shoot mostly in other scenarios, it's pretty ace. So it's a mix of behind the times and totally ace, grand slam.
The 5D4 is a tricky game though. It simply 100% has to either have 4k video and also either 1080pRAW or at least Cxx level 1080p in every last way OR an exmor-like sensor. It really needs both. Without either it just seems DOA. Without the video improvements, you may as well stick to a 5D3 for video then or start buying other stuff from other makers. And for stills all you'd get it going from maybe 23 to 39MP, which is nice, but this late in the game, without better per pixel quality, more DR and better colors and all....
The 7D2 can get away with the older low ISO performance and even the 4k, although it;s a shame, the rest is totally amazing (barring some 1D3 AF type fiasco).
The 5D4 no way no how can get away without matching exmor and having a really high quality 4k 10-12bits 4:2:2 no wax-works (at the very least it has to have at least one of those two things at the super minimum).
mmurph wrote:
I have been told that there is no way that Canon could do the 40+ MP Dual Pixel Auto Focus with the 0.5 micron fab, that they have to be using the new fab already.
I don't know if that is true or not - I don;t have enough knowledge to say (although I could pretend, this being the internet ...)
The thing though is then why have they still not implemented dual iso reads per photosite or on chip ADC and so on?
Access wrote:
You guys overstate the importance of the 0.XX micron process or the sensor to begin with. Most people buying cameras don't even know what that means. Techies love to seize on numbers but often times miss the big picture in doing so.
Most of the photographers out there stopped caring about sensors around 5-6 years ago (roughly 2008) if not before then. For the typical photographer, the sensors have surpassed 'good enough' and when one of these sees the 7d2, they will take note of an autofocus that improves upon the already good autofocus of its predecessor, the improved fps, etc. The sensor might be the same, it might be slightly improved, or it might be dramatically improved, but this is likely to be little more than an afterthought to the typical user.
You guys tend to get tied up on a specific piece of hardware but that is not the whole picture even when it comes to image quality. Firmware and software and everything else that goes around that sensor is just as important, look for instance at the demonstrate-able improvements that ML has made on some of the latest sensors using dual ISO mode and so on. This should demonstrate the folly citing a number such as process geometry and saying it will never be as good, because of this -- when in reality it's just a number....Show more →
It';s not just a number, without a smaller process you can't implement dual iso per photosite in hardware (which being per photosite it far better than the ML hack) or on chip ADC, etc. and without that it's incredibly hard to match exmor.
skibum5 wrote:
Basically other than shooting stills at ISO100-400 (and a little at ISO800) the sensor should be very good and other than for video image quality which have some doubts about even beyond just lacking 4k, the specs are awesome. So if you don't shoot video and can ignore the high DR scenes and tend to shoot mostly in other scenarios, it's pretty ace. So it's a mix of behind the times and totally ace, grand slam.
The 5D4 is a tricky game though. It simply 100% has to either have 4k video and also either 1080pRAW or at least Cxx level 1080p in every last way OR an exmor-like sensor. It really needs both. Without either it just seems DOA. With the video improvements, you may as well stick to a 5D3 for video then. And for stills all you'd get it going from maybe 23 to 39MP, which is nice, but this late in the game, without better per pixel quality, more DR and better colors and all....
The 7D2 can get away with the older low ISO performance and even the 4k, although it;s a shame, the rest is totally amazing (barring some 1D3 AF type fiasco).
The 5D4 no way no how can get away without matching exmor and having a really high quality 4k 10-12bits 4:2:2 no wax-works (at the very least it has to have at least one of those two things at the super minimum).
Yes sensor is brand new, it's not the same as 70D, but the underlying tech is exactly the same. Same ADC pipeline, 0.5um fab. There might well be changes to photodiodes, microlenses, improvements in circuitry etc that improve noise and along with digic 6 we hopefully will see better IQ, but don't expect to see much improvement in say DxO mark scores for DR and SNR.
Access wrote:
You guys overstate the importance of the 0.XX micron process or the sensor to begin with. Most people buying cameras don't even know what that means. Techies love to seize on numbers but often times miss the big picture in doing so.
Most of the photographers out there stopped caring about sensors around 5-6 years ago (roughly 2008) if not before then. For the typical photographer, the sensors have surpassed 'good enough' and when one of these sees the 7d2, they will take note of an autofocus that improves upon the already good autofocus of its predecessor, the improved fps, etc. The sensor might be the same, it might be slightly improved, or it might be dramatically improved, but this is likely to be little more than an afterthought to the typical user.
You guys tend to get tied up on a specific piece of hardware but that is not the whole picture even when it comes to image quality. Firmware and software and everything else that goes around that sensor is just as important, look for instance at the demonstrate-able improvements that ML has made on some of the latest sensors using dual ISO mode and so on. This should demonstrate the folly citing a number such as process geometry and saying it will never be as good, because of this -- when in reality it's just a number....Show more →
Yes hence the fact we actually own Canon gear and don't just come here to troll and many of us will buy the 7D II.
Just because most people don't care what's under the hood, and I disagree as I believe a good % of advanced DSLR users are technically inclined and do care about this stuff, doesn't mean we shouldn't discuss the issues. Canon has done a good job with its old fab and milked it for what it's worth, but it is time to move on and unless they do you will never see a high MP camera from them. There's a good reason Canon has been stuck in the 20-22MP range for so long now.
aladyforty wrote:
seems to be a lot of bitching going on across the net on various sites about this camera, wouldn't it be better to wait until its been thoroughly tested?
Yes and no. Out of discourse come insights - read some of the posts.
I'm curious - how will reducing the fab process size improve DR?
I think we can take it as read that a smaller fab process will enable greater pixel density, but DR is a function of the noise floor and the efficiency of the sensels to collect photons. How does a smaller process size improve this?
15Bit wrote:
I'm curious - how will reducing the fab process size improve DR?
I think we can take it as read that a smaller fab process will enable greater pixel density, but DR is a function of the noise floor and the efficiency of the sensels to collect photons. How does a smaller process size improve this?
Note the smaller Fab doesn't affect pixel size so much, as they are relatively large compared to the line traces - 5-7um vs 0.18-0.5micron. You want your photodiodes to be as as large as possible to collect as many photons as possible.
Well Sony uses a 0.18micron fab process. This means the line traces for the back-side circuitry is much smaller. Sony has managed to keep the entire readout, ADC-pipeline, NR etc on chip. Canon has to do a lot of this off chip and Sony's entire pipeline after the initial ADC is digital. The very short pipeline means less chance for noise to be introduced. The Sony has low ISO DR advantage due to the fact it's read noise is about 10x less than Canons' comparing D800 and 5D III. DR is the maximum signal divided by the noise floor in a pixel at each ISO and at low ISO the Canon noise floor is much higher. At higher ISO the read noise improves quickly for Canon and this is why at about ISO 800 or so the DR of the 5D III is similar to D800 and pulls ahead at higher ISOs.
I woundn't expect more than a very small incremental improvement compared to the 70D ... since higher readout speed usually comes with a noise penaly it may even be somewhat worse ...
phibes wrote:
I woundn't expect more than a very small incremental improvement compared to the 70D ... since higher readout speed usually comes with a noise penaly it may even be somewhat worse ...
Unless they move from an 8 to 16 channel readout like on the 1D X.
Pixel Perfect wrote:
Note the smaller Fab doesn't affect pixel size so much, as they are relatively large compared to the line traces - 5-7um vs 0.18-0.5micron. You want your photodiodes to be as as large as possible to collect as many photons as possible.
Well Sony uses a 0.18micron fab process. This means the line traces for the back-side circuitry is much smaller. Sony has managed to keep the entire readout, ADC-pipeline, NR etc on chip. Canon has to do a lot of this off chip and Sony's entire pipeline after the initial ADC is digital. The very short pipeline means less chance for noise to be introduced. The Sony has low ISO DR advantage due to the fact it's read noise is about 10x less than Canons' comparing D800 and 5D III. DR is the maximum signal divided by the noise floor in a pixel at each ISO and at low ISO the Canon noise floor is much higher. At higher ISO the read noise improves quickly for Canon and this is why at about ISO 800 or so the DR of the 5D III is similar to D800 and pulls ahead at higher ISOs....Show more →
Thank you for the excellent explanation. What resources are you using for this information? I know more than a few people besides myself who would like to read more.
Ziffi3: 4K resolution TVs are already being offered for sale. So the sooner 4K resolution cameras are cheap enough to buy the sooner it will be worth buying these new TVs.
There area already prosumer cameras that have 4K resolution video like the Sony A7s, Panasonic GH4 and FZ1000 to name a few. So why shouldnt a $1800 Canon SLR not have it as well?
