Cease & Desist order at "another website"

[-hh]

Not sure what exactly attracted the attention of the lawyers to the rumor sight in question, but it could have been another spat of rumored specifications.

For those that have been tracking the looooooong anticipated announcment of an upgrade to an existing ("and its not a Nikon") Full Frame dSLR, this list might be of slightly higher credibility...or maybe not:

* 21.1 MP 1.0x
* DIGIC IV
* ISO 100-6400 L:50 & H:12800
* 5 FPS
* 3.2" High Resolution Screen (LCD)
* 19 point AF
* HDMI Out
* Liveview
* HD Movie Mode
* Viewfinder: 100% Coverage
* Full weather sealing
* EF Lenses only

Still no clear word on announcment date, etc.

It does appear that there are some people out there with insight, but are still subject to the terms of their respective NDA's...which always seems to be the old question of if it is within the NDA to be able to mention mention when said NDA expires.



-hh

 

[Westside guy]

That seems like an odd assortment of features. Why movie mode in (what I assume is) a prosumer camera? And only 5 fps?

I call shenanigans.

 

[OreoCookie]

That seems like an odd assortment of features. Why movie mode in (what I assume is) a prosumer camera? And only 5 fps?

And ISO12800 in a 21.1 MP camera? I'm not sure that would do well with the typical crowd that loves the current 5D for its high-ISO performance.

 

[vga4life]

That seems like an odd assortment of features. Why movie mode in (what I assume is) a prosumer camera? And only 5 fps?

I call shenanigans.

Looks pretty realistic for a 5D replacement or for a "3D" mini-pro body to me. The 5D dates back to 2005 and is now ancient (still a great camera, especially at its recent $1900 price) - a new, compact full-frame body to compete with the D700 is the no-brainer obvious release for Canon at Photokina. And 5FPS will be plenty for it - that's what the current 1Ds-III does today. (No doubt the 1Ds-IV will do more.)

Movie mode is going to be in every new digital camera besides the next generation of pro bodies (1D, D3), guaranteed. (I'd even bet it will be available in pro bodies within 2 generations.)

 

[-hh]

That seems like an odd assortment of features. Why movie mode in (what I assume is) a prosumer camera? And only 5 fps?

Its only 5 fps because the sensor "size" (megapixels) increased.

The DIGIC IV chip is being used in the newly released 50D, where it cranks 6.3 fps on a 15MP sensor ... that's 95 megapixels/sec worth of processing bandwidth.

Here, 5 fps @ 21MP = 105 MP/sec ... roughly the same level of performance (okay, 10% higher).

Insofar as movie mode, it appears that Nikon, Canon and Sony all intend to release "...first dSLR with..." models at Photokina. FYI, the max clip length may be around ~5 minutes, apparently to prevent the sensors from overheating.


-hh

 

[tip]

Not sure what exactly attracted the attention of the lawyers to the rumor sight in question, but it could have been another spat of rumored specifications.

I know exactly what you're talking about, and it's not due to any leaks whatsoever.

- Somebody on "another forum" made a fictitious post posting specs on a camera successor, ending that post with "a friend who's an elf told me," or something to that effect. It was completely fake.
- The "rumor site" took those specs (in fact, nearly all their rumors are from that "other forum") and posted them as a regular article.


That "rumor site" is being approached by the camera manufacturer because the company name is in the "rumor site" URL, complete with company logos, etc. This is basically similar to when Apple approached every company and website that used "Ipod" in its name.

Those specs would be great for the next successor of that particular camera; however they're completely false.

 

[Abstract]

....however they're completely false.

I didn't know any of the specific info you mentioned, but I already figured the above statement was true.

Unless Canon is updating their 1ds III with some extra features, it's made up. I doubt this will be the 5D replacement. Canon isn't that stupid, are they? They're going to release a 12, 14, or 16 MP full frame sensor, and include weather sealing, as the lack of weather sealing was a huge criticism of the 5D when it was first released.

 

[-hh]

And ISO12800 in a 21.1 MP camera? I'm not sure that would do well with the typical crowd that loves the current 5D for its high-ISO performance.

Nikon raised the bar to 25,600 in this area, so some response is necessary if for no other reasons than political / marketing.

In terms of the actual technical challenges, one of the things that has been accomplished in the new 50D is the micro lens gap. The new 50D sensor now is "gapless", which means that effectively, the sensor has 100% coverage: no light hitting it gets lost. I don't know what the micro lens coverage efficiency was of the original 5D, but if we say that it was 50% for sake of simplicity, then in going to 100%, we could actually double the number of pixels and retain the same quantum efficiencies, S/N ratios, etc.

There's also at least three other factors that could be contributing in a positive fashion:

1) reduced noise level per pixel.
2) improved digital noise reduction processing

There's reportedly been improvements in both of these areas; IIRC, I think I've read that the 50D is considered to be 1-1.5 stops better than the 40D, so this is simply another example of "...if they've already shown it working in Product X, then they can put that technology into Product Y too".

3) "pixel binning" techniques

I think Nikon shipped this recently? In any case, its worth a quick word to mention what it is for those who aren't familiar with it...basically, it takes the outputs of multiple pixels and adds them together to create a (figuratively) virtual pixel that has better sensitivity performance...but the trade-off is that the combining means you end up with fewer pixels. For a super-simple example, if you were to take a 4MP sensor and "bin" at 4:1 (4 creates 1), you would end up with higher ISO sensitiviy, but with only a 1MP image.

At least this is how I understand the concept.


-hh

 

[Abstract]

3) "pixel binning" techniques

(snip).....


At least this is how I understand the concept.

It's a lot more complicated than that.

You get 4x the noise as well when you bin 4 pixels, which is why it has never worked well on any other camera. I think they've stopped doing it completely, but you never know.

And of course the biggest problem is that "binning" isn't as simple as taking 4 pixels beside each other and adding up the light they picked up. Each pixel is just the interpolation of R, G, B pixel values. When you try to bin, it's really difficult to bin 4 pixels because you'd need the pixel values from an outside pixel (i.e. from a pixel that's not one of the 4 pixels in your bin. You'd need to borrow a pixel from a neighbouring 4-pixel bin) in order for the interpolation to work. Of course, you could just make an awesome camera for black and white photography and forget about the interpolation, but then you may as well forget about the colour filters as well and increase the amount of light hitting each sensor by 3x (and 2x).


To be honest, I don't think I know much about this sort of thing. I may be wrong, but I'm certainly correct that it's not as simple as just binning 4 or more pixels together.

 

[Westside guy]

I have a hard time believing Canon would put a 21MP sensor in a 5D replacement for the reasons others have mentioned; but also because Canon has consistently been very careful with respect to including features in their lower-end cameras that might cannibalize sales of their higher-end cameras. Nikon has been more aggressive in this regard; but IMO that's because they've felt it necessary in order to grab market share in all segments back from Canon. If Nikon ever manages to get a commanding sales lead over Canon - which I don't think is even remotely likely - they'll probably start getting more conservative again.

Fortunately we don't have too long to wait for Photokina, when we'll likely find out the truth. It's going to be fun to see all the announcements over this next few weeks...

 

[ksz]

Random noise can be suppressed simply by subtraction, in any of the following ways:

1. If we have 4 pixels, each of them composed of R, G, and B elements, then we can subtract all R elements separately, all G elements separately, etc.
   
2. But if this leads to smearing of actual detail, we could at least subtract the two closest R elements, the two closest G elements, etc.
   
3. We could run the Bayer filter to combine 4 sets of R, G, and B (or R, G, G, B) into 4 pixels and either use subtraction or interpolation to arrive at a single pixel.

Scanning Electron Microscopes (SEM) produce high-quality, relatively noise-free images by sampling the same area multiple times and subtracting all of the frames.

Just to be clear, the word "subtract" in this context means looking for the difference between frames and subtracting the difference.

 

[Cave Man]

Looks like I'll be getting a new camera before the end of the year, but which one?

 

[OreoCookie]

Nikon raised the bar to 25,600 in this area, so some response is necessary if for no other reasons than political / marketing.

Yes, exactly.
If Canon increases the number of MP considerably (in these proposed specs, the number of pixels would be almost double that of the current 5D), then they sacrifice high-ISO performance.

No amount of engineering changes that: if two sensors share the same technology, then the one with the lower MP count will have better noise characteristics.

Nikon has made the choice to keep the pixel count lower and have very, very good high-ISO performance. Once the rumored D3X is released, they will also have a camera whose high-ISO performance is `only' very good, but with a higher resolution. If the price points of the 1D and 1Ds models of Canon are any indication, then the 1Ds' sensors are substantially more expensive. Which means the hypothetical 5D Mark II with 20+ MP would be very expensive. Plus, it would cannibalize sales of the 1Ds which is used for `slower' subjects (it can do `only' 5 fps). Somehow this doesn't fit in my opinion.

Canon has been very successful to keep the 5D in its current niche and while I don't exclude a moderate increase in pixel count (say, 15 MP or so), I doubt it will be as dramatic as these `leaked specs' suggest. I don't see that they want to change this niche (very good high-ISO performance, not super fast in terms of fps, moderate, but sufficient sensor resolution).

 

[Westside guy]

Just playing devil's advocate here, so don't take me TOO seriously - I don't necessarily believe what I'll be arguing...

It's quite possible that Canon's cost for a 21MP sensor has come down dramatically, yet they've kept the 1DsMkIII's price elevated. Most of us here are passingly familiar with a company that does pretty much the same thing.

It's also possible that Canon might feel they HAVE to go with a 21MP sensor in the 5D rev. 2 if they've had trouble matching Nikon in the area high-ISO performance. And as we've seen on our own forum, there are a not insubstantial number of dSLR users that equate pixel density to "largest printable image" without thinking about all the other contravailing issues that come with a higher-density sensor.

I am curious to see what Canon comes out with, in any case - hopefully it won't be too long! This is all a lot of fun from a gearhead-wannabe perspective.

 

[OreoCookie]

It's also possible that Canon might feel they HAVE to go with a 21MP sensor in the 5D rev. 2 if they've had trouble matching Nikon in the area high-ISO performance. And as we've seen on our own forum, there are a not insubstantial number of dSLR users that equate pixel density to "largest printable image" without thinking about all the other contravailing issues that come with a higher-density sensor.

Interesting take. In other words, Canon changes the niche of the 5D Mark II, because it has to …

 

[vga4life]

Interesting take. In other words, Canon changes the niche of the 5D Mark II, because it has to …

Or they're releasing a '3D' between the 5D and 1D to test their new technologies ahead of a pro-line refresh, as they've been known to do before. (The EOS 3 introduced the autofocus system that later made its way into the EOS 1 and 1D...)

 

[-hh]

Yes, exactly.
If Canon increases the number of MP considerably (in these proposed specs, the number of pixels would be almost double that of the current 5D), then they sacrifice high-ISO performance.

No amount of engineering changes that: if two sensors share the same technology, then the one with the lower MP count will have better noise characteristics.

Agreed; my point is that the "same technology" assumption doesn't apply.

Simplistically, minimizing the microlens gap runs against the normal rule of thumb of 'more pixels = smaller pixels = poorer high ISO'.

In a simplistic ideal world model, pixels can be added without the microlens size needing to be reduced until such time that there's no more gaps between microlenses.

Since the 50D is the first camera that is advertised with a "gapless" system, this infers that all predicessors had gaps, which represent opportunities for increased pixels without sacrificing microlens size, etc.

What's unknown is how much "room" was remaining, to determine how large of a bump up in Megapixels without any need to compromise the microlens size (and thus, ISO sensitivity).

In personally working out some hypothetical numbers based on 13MP vs 21MP sensors, my guess is that the old system's linear gaps of would need to be less than ~20% in order for a new gapless system to have inferior ISO sensitivity performance.

Canon has been very successful to keep the 5D in its current niche and while I don't exclude a moderate increase in pixel count (say, 15 MP or so), I doubt it will be as dramatic as these `leaked specs' suggest. I don't see that they want to change this niche (very good high-ISO performance, not super fast in terms of fps, moderate, but sufficient sensor resolution).

I've been surprised too, but not necessarily at this number per se, but rather the sum of the rumored numbers. It appears to me to be straightforward to get to 21MP in a FF with "same as current" ISO sensitivity; the question is being to move both of them up.

Time will tell. FWIW, I also find your other comment of "Canon changes the niche of the 5D Mark II, because it has to …" intriguing as well.

Plus there's also the other rumor floating around of there potentially being 2 cameras to replace the 1.



-hh

 

[-hh]

Agreed; my point is that the "same technology" assumption doesn't apply.

Simplistically, minimizing the microlens gap runs against the normal rule of thumb of 'more pixels = smaller pixels = poorer high ISO'.

In a simplistic ideal world model, pixels can be added without the microlens size needing to be reduced until such time that there's no more gaps between microlenses.

Just thought of an analogy that might help illustrate what I'm talking about.

Envison a billiard (pool) table.
This is the area that represents the sensor.

Now put some standard billiard balls on the table.
Each one of these represents a microlens assembly with a 'pixel' at its bottom.

Now arrange these balls on the table in neat rows & columns so that they provide even coverage across the entire table. Thus, we've built our camera.

If we started with just 16 balls on the table, the gap between each ball is huge. So let's do technology research to increase our resolution (megapixels paradigm) - we just need to figure out how to adding more balls. And after each addition, we re-arrange them for even coverage.

Add 16 more balls. .. higher megapixels paradigm.
Add 16 more balls. .. higher megapixels paradigm.
Add 16 more balls. .. higher megapixels paradigm.

Note that because we've not changed the size of the balls we're using, the light-gathering performance of each individual assembly has remained the same.

Add 16 more balls. .. higher megapixels paradigm.
Add 16 more balls. .. higher megapixels paradigm.

We can keep doing this until we run out of room on the pool table for balls. This point occurs when the gaps between the balls becomes zero...the system became "gapless".

At this point in time, the only way to cram in more balls on the table (for higher megapixels) is to change to use smaller balls. But at this point, because each ball is smaller, it gathers less light individually, so the light-gathering performance of it as an individual assembly (microlens & pixel sensor) declines.

With the announcement on the 50D being "gapless", all we know are two things: first, that its billiard table is 'full' (with whatever sized balls it is using), and second, that it is now possible to actually 'fill' a billiard table.


-hh

 

[SLC Flyfishing]

Not sure what exactly attracted the attention of the lawyers to the rumor sight in question, but it could have been another spat of rumored specifications.

For those that have been tracking the looooooong anticipated announcment of an upgrade to an existing ("and its not a Nikon") Full Frame dSLR, this list might be of slightly higher credibility...or maybe not:

* 21.1 MP 1.0x
* DIGIC IV
* ISO 100-6400 L:50 & H:12800
* 5 FPS
* 3.2" High Resolution Screen (LCD)
* 19 point AF
* HDMI Out
* Liveview
* HD Movie Mode
* Viewfinder: 100% Coverage
* Full weather sealing
* EF Lenses only

Still no clear word on announcment date, etc.

It does appear that there are some people out there with insight, but are still subject to the terms of their respective NDA's...which always seems to be the old question of if it is within the NDA to be able to mention mention when said NDA expires.



-hh

Yawn!

 

[theBB]

It's a lot more complicated than that.

You get 4x the noise as well when you bin 4 pixels, which is why it has never worked well on any other camera.

No. The total amount of light goes up 4 times, but noise in each pixel is random so the total noise on average only goes up by 2. It is counter intuitive, but "probability" dictates so.

 

[ksz]

At this point in time, the only way to cram in more balls on the table (for higher megapixels) is to change to use smaller balls. But at this point, because each ball is smaller, it gathers less light individually, so the light-gathering performance of it as an individual assembly (microlens & pixel sensor) declines.

With the announcement on the 50D being "gapless", all we know are two things: first, that its billiard table is 'full' (with whatever sized balls it is using), and second, that it is now possible to actually 'fill' a billiard table.


-hh

There is a distinction to be made:

1. There are still physical gaps between the photosites.
2. "Gapless" means that the microlenses sitting on top of each photosite are completely filling the space between photosites.
3. Light sensitivity is improved because nearly all of the light hitting the sensor is focused onto the photosites.
4. Canon can still shrink or enlarge each photosite.

Hence, the billiard table is not full, but the microlens coating applied on top of the table now fills all the space between the balls. (I've redefined "balls" to be just the photosites.)

Diagram from DPReview: