M Vers wrote:
Obviously FF sensors haven't plateaued in the MP department yet, but I think it's safe to say that smaller sensors, like the one in the G10 and the 50D/T1i, have hit the ceiling. Unless other technological advancements occur I can't see them increasing in MP size all to much further.
It may well be that for a tiny G11 sensor with current lens technology, 15 MP v 10 MP would only offer an advantage in aliasing, not actual detail. If so, it might be nice to have the smaller file size of the new model. I just don't think noise is really the issue here.
I also don't really expect this (reduced pixel count) to become a trend that carries over to FF sensor design. Hope not.
brainiac wrote:
I only raised the Stephenson gauge because it was a good example of how popular misconceptions can favour worse engineering. If only Brunel's gauge had been adopted trains would be much faster, cheaper, safer, more comfortable, and more people would use them. Digital cameras matter less, but like the photodisc era, we are approaching an era in which some people will get less good snaps because of a popular misconception. Misconception-driven engineering is the point.
Ah, thanks for the history. We've seen plenty of examples of bad choices of standards (VHS v Betamax, eg.) for various reasons, but the railroad gauge seems like a good parallel to what people are calling for in sensor design.
Does that mean that my discontinued G10 will be worth more than the G11 once people find out the real truth?
I think that's an interesting point, although the G11 sensor will probably be better by some margin anyway since it's newer technology. We can be sure that any extent to which it is better than the G10 sensor will be put down to the wonderful larger sensels rather than the tide of technological progress. Once this great advance is heralded by DPR and co. as some kind of epiphany, will Canon really push the boundaries and realease a 7 Mpixel G12? Probably not once Fuji release a superior 15 Mpixel p+s. It all reminds me of the haggling scene in the Life of Brian:
brainiac wrote:
Sorry I didn't provide identifiers, but to my eye there is so little between the cameras at iso 12800 that who cares which is which?
The point of posting this file is to illustrate that most people's beliefs about the effects of pixel contraction are based on per pixel analysis, which is thoroughly misleading.
Well, I care (obviously) and please note, there are differences ... the illumination is slightly different between the two pictures and the shot on the right shows greater detail. Just a suggestion ... if "visual equality between the two shots" was the basis of making your point then IMO you should have simply stated that instead of letting folks flounder around thinking they should somehow find one shot to be better than the other. Also a small nit, you pulled in separate quotes from two posters and left "quote-ownership" ambiguous; obviously this isn't a big deal but you might want to edit your post for clarification.
Regarding per-pixel comparisons ... I agree with you except that they are more than just misleading ... that approach is incredibly foolish PERIOD.
AFAIC the only valid comparison is per picture ... but note, this comparison must be limited to two cameras (and only two cameras), both shots should be interpolated to the geometric mean of each camera's lineal pixel count (note, an arithmetic average is just fine if the difference in lineal pixel count is not extreme), and then sharpened equally. Another viable per-picture interpolation is to up-rez both pictures to a size that is about two times larger that the average of the two files and then sharpen both images equally. Again, regarding per-picture comparisons, if one tries to interpolate one camera up to the size of the second (or conversely one camera down to the size of the second) the camera "not interpolated" always gains a small advantage; thats why it's important to choose a final size that forces interpolation of both images.
The problem that camera review sites have is they try to show comparisons to multiple "other camera models" and are stuck with either per-pixel (e.g. DPR) or interpolating all cameras down to a ridiculously small common image size (e.g. DxO); both per-pixel approaches severely penalize the higher resolution camera (as compared to the lower resolution camera).
Nope, me either...at least not any time soon. I do, however, expect the MP race to slow or even stop in regard to APS-C. Unless Canon pulls something out of the tech closet I don't see high ISO performance becoming any better past 15mp. Unlike most people I feel Canon was successful in increasing mp count 50% from the 40-50D all the while keeping ISO performance virtually the same. Now if they can only figure out a way to continue increasing mp's in conjunction with high ISO performance I think a lot more people would be a little more accepting of higher mp count.
Don't forget that if you put a larger version of the average p+s sensor into an APS-C camera you end up with a sensor in the hundreds of Mpixels. People talk as though the limit is the sensel size. If fact what really limits the number of pixels is the camera throughput. Imagine how long you would have to wait between frames if you put a 128 Mpixel sensor in a 500D. That's why I think Canon is pandering to misconceptions. They have realised that they will never persuade sites like DPR to do fair comparisons, so being in the lead in pixel count counts against you. Better to sit back and underperform to glowing reviews.
This is awesome to see from Canon. Have they ever previously gone down on pixel count? It shows they identify the need for low noise and greater DR is greater than the need for more pixels.
brainiac wrote:
Don't forget that if you put a larger version of the average p+s sensor into an APS-C camera you end up with a sensor in the hundreds of Mpixels. People talk as though the limit is the sensel size. If fact what really limits the number of pixels is the camera throughput. Imagine how long you would have to wait between frames if you put a 128 Mpixel sensor in a 500D. That's why I think Canon is pandering to misconceptions. They have realised that they will never persuade sites like DPR to do fair comparisons, so being in the lead in pixel count counts against you. Better to sit back and underperform to glowing reviews....Show more →
What about the misconception that every camera ever mad needs to produce billboard sized images? Why do we need so damned many pixels in every body from point and shoot to 50d?
Let the tiny fraction of folks that need that kind of resolution use the 1ds and 5d and leave the rest if the world alone.
Well good move and common sense by Canon to keep the Megapixel count to a sensible amount.
However feature wise this is hardly any advance over the G9 model which had the same size sensor although at 12 Megapixels.
I doubt we will see much improvement in noise, the sensor is still too small, why can't Canon produce a good compact camera with something around the APS-C sized sensor. I used to own a Ricoh GR1 film camera which was a brilliant compact camera with a full-frame sized sensor and 28mm f2.8 lens!
I guess Canon fear producing a really good compact would compromise DSLR sales.
It looks like Panasonic will be leading the way again as with the LX3 and GH-1 type cameras.
Also why no HD video 720 in the G11? They can put it in the SX20 so this is clearly a deliberate crippling in firmware. Even the old G9 had higher resolution video and also the nifty time-lapse mode.
Also lens is only 28mm and f2.8. The S90 has an f2.0 lens so why not the G11? A 24mm lens would also be prefered, even at the sake of loosing some telephoto range.
FWIW, I recently sold a 40D *after* buying a 50D and checking the noisiness *in prints*. There's virtually nothing to choose between them at high ISOs (high for the 40D). The 50D is significantly better for big prints at low ISOs and any reasonable size above 3200, where the 40D doesn't go IIRC. With Noise Ninja and down-rezzing you can make decent small prints (up to about 5x7") from a G10 at ISO 1600. Don't expect flawless IQ, but perfectly acceptable for the sort of shots that generally call for such high ISOs. The images look pretty horrid at 100% on a screen, of course. BTW, max ISO on the g11 is 3200.
brainiac wrote:
Don't forget that if you put a larger version of the average p+s sensor into an APS-C camera you end up with a sensor in the hundreds of Mpixels. People talk as though the limit is the sensel size. If fact what really limits the number of pixels is the camera throughput. Imagine how long you would have to wait between frames if you put a 128 Mpixel sensor in a 500D. That's why I think Canon is pandering to misconceptions. They have realised that they will never persuade sites like DPR to do fair comparisons, so being in the lead in pixel count counts against you. Better to sit back and underperform to glowing reviews....Show more →
brainiac,
I fear you are arguing against the tide here, where logic and sanity of position will not prevail While I'm not privy in anyway to Canon's engineering nor their thought processes, there is another possible reason for a reduced pixel count in the sensors on the G11/S9 beyond just pandering to the Internet masses. It is entirely possible that Canon has developed an entire new class of sensor and is starting production with a fairly low sensel/pixel count. This is a fairly common thing to do in chip production where density is ramped up in later iterations as the fabrication process is perfected. It seems clear that the sensor in these camera is different than their predecessors. Just how different remains to be seen. While there is much reason to fear that marketing rules Canon as it seems to run so many companies these days, there are still some excellent engineers in the company so it is possible that they created something entirely new and better. Guess we'll have to wait and see.
nads wrote:
What about the misconception that every camera ever mad needs to produce billboard sized images? Why do we need so damned many pixels in every body from point and shoot to 50d?
Let the tiny fraction of folks that need that kind of resolution use the 1ds and 5d and leave the rest if the world alone.
It's not ONLY about print size, but breathing room in cropability AND increase in overall detail. For macro, wildlife, product, landscape and architecture higher resolution definitely lends a helping hand.
nads wrote:
What about the misconception that every camera ever mad needs to produce billboard sized images? Why do we need so damned many pixels in every body from point and shoot to 50d?
You can see the difference in resolution between 5D and 1Ds3 in prints as small as A3, and sometimes even A4. But I agree - if you don't want sharp images, ever, then don't spend the coin.
I fear you are arguing against the tide here, where logic and sanity of position will not prevail While I'm not privy in anyway to Canon's engineering nor their thought processes, there is another possible reason for a reduced pixel count in the sensors on the G11/S9 beyond just pandering to the Internet masses. It is entirely possible that Canon has developed an entire new class of sensor and is starting production with a fairly low sensel/pixel count. This is a fairly common thing to do in chip production where density is ramped up in later iterations as the fabrication process is perfected. It seems clear that the sensor in these camera is different than their predecessors. Just how different remains to be seen. While there is much reason to fear that marketing rules Canon as it seems to run so many companies these days, there are still some excellent engineers in the company so it is possible that they created something entirely new and better. Guess we'll have to wait and see.
While I'm not privy in anyway to Canon's engineering nor their thought processes, there is another possible reason for a reduced pixel count in the sensors on the G11/S9 beyond just pandering to the Internet masses. It is entirely possible that Canon has developed an entire new class of sensor and is starting production with a fairly low sensel/pixel count. This is a fairly common thing to do in chip production where density is ramped up in later iterations as the fabrication process is perfected. It seems clear that the sensor in these camera is different than their predecessors. Just how different remains to be seen. While there is much reason to fear that marketing rules Canon as it seems to run so many companies these days, there are still some excellent engineers in the company so it is possible that they created something entirely new and better. Guess we'll have to wait and see.