"1Dsq" & 3D [CR1]

[macfly]

Re: “1Dsq” & 3D [CR1]

I think the patents are a very pertinent point, and I really can't see Canon building a whole new system for the ever shrinking stills market. I bet that their focus will be on video all the way now.

 

[Bob Howland]

Re: “1Dsq” & 3D [CR1]

Ok, so let's see... Canon is going to bring out all new lenses for this square format sensor, and we haven't seen a single square format lens patent... right that does that one in.

As for this 3D.. Anyone who thinks Canon is going to drop their sensel density needs to give their heads a shake. ISO performance and sensel size have *very* little to do with each other, the newest Sony APS-C proves this point.The current Nikon high-iso monster which shall be unnamed is that way due to other design reasons, not the sensel size.

I *really* wish people would get over this silly internet myth that smaller pixels are bad (and it is a myth until a point we aren't close to). You actually improve DR with more smaller sensels due to highlights not being blown to kingdom come.

While I appreciate the rumors didn't originate with CR, I do wish people would apply a little critical thinking to some of them. If you guys *want* something from a sensor how about faster microlenses, something f0.6 or so to allow f1.2 lenses to actually be used to their full advantage in terms of light gathering. The only reason I don't own an f1.2 lens is I can't use it at close to f1.2 from a light gathering perspective (and I really don't need DOF that thin.. not that I couldn't handle it, but I don't need it for artistic purposes).

Give me a 40MP, 4:1 binned (optionally) FF sensor with fast microlenses and I will buy it. Immediately. Give me nice low readout noise and nicely designed wells of a good depth. Move as much of the support circuitry to the back of the sensor to improve sensel density.. these are the things people would be asking for if they stopped to think about how light and sensors *actually* work... oh and if someone (anyone, I'd switch brands for this in a heartbeat), can come up with a way to have the bayer filter be movable so I can put the camera in true monochrome mode, they would be my god (and yes I know it won't happen, but if we're dreaming.. )

1. Which new Sony sensor are you talking about, 16MP or 24MP?

2. What are the "other design reasons" for the relatively low large pixel size in the "Nikon high-iso monster"?

3. As for the "silly Internet myth" about pixel size, go to the Canon USA website and look at their bragging about the sensor for the new M40/M41/M400 camcorders. Coming from Canon, it's downright weird. If pixel size doesn't matter, is binning desirable only to reduce file size? Also, you might want to talk with the people who designed the sensor for the Red Epic movie camera. They're claiming 13.5 stops of DR with relatively large pixels.

4. I didn't know that microlenses even had f-stops. Could you elaborate?

 

[neuroanatomist]

Re: “1Dsq” & 3D [CR1]

Ok, so let's see... Canon is going to bring out all new lenses for this square format sensor, and we haven't seen a single square format lens patent... right that does that one in.

I'm with dilbert on this one...lenses are round, and they project an image circle. How you sample the light within that circle - rectangular or square - is independent of the lens. The only exception to that is where there is a rectangular baffle at the rear of the lens, which I've seen on the 24-105mm and nowhere else (some telephoto lenses have a square baffle with slightly rounded corners, but not something that would impinge on the shape of the image circle).

 

[NotABunny]

Re: “1Dsq” & 3D [CR1]

What are the "other design reasons" for the relatively low large pixel size in the "Nikon high-iso monster"?

See http://www.canonrumors.com/forum/index.php/topic,255.msg3911.html#msg3911 for the real factor responsible for low image noise: sensor size.

If pixel size doesn't matter, is binning desirable only to reduce file size?

Sensor / image resolution has nothing to do with noise per image, but it affects the noise per pixel.

 

[KyleSTL]

Re: “1Dsq” & 3D [CR1]

Not to mention I wouldn't want to see the corners on a 36mmx36mm sensor with the lens that don't have the baffle on the back of them.

The image circle for a an EF lens would make a 30.6mm square sensor. Sensor size would be constrained by the hypotenuse (43.266mm), not the width of a 3:2 ratio frame. For the record, a 36x24mm sensor is 864 mm2, a 30.6mm square sensor is 936 mm2. That is an 8% increase of surface area. That, coupled with the fact that a square sensor would have less sensor area outside of the 'sweet spot' of the lens would likely result in slightly sharper images.

 

[Orangutan]

Re: “1Dsq” & 3D [CR1]

>The image circle for a an EF lens would make a 30.6mm square sensor.

If we're talking about a camera at $8-10K+, why not make a sensor that covers the entire image circle; then let the photographer decide how to crop it. Will that create wasted silicon? Yup, but this is not a consumer camera, and it would eliminate the need for a "vertical" grip.

 

[macfly]

Re: “1Dsq” & 3D [CR1]

"They don't need to build a whole new lens system."

Dilbert, make a circle that touches the edges of a rectangle with an aspect ratio of 1.5 x 1 then make a circle around the squares made up of the 1.5 dimension, and you'll see that the area of coverage is significantly expanded.

As other her noted the square will be smaller than the 1.5 dimension if they went that way but kept the current lenses. I'd say this has a snowball's chance in hell of being true, but hey, Canon have made some really bad decisions lately, like messing up the once wonderful G series, so maybe.

 

[KyleSTL]

Re: “1Dsq” & 3D [CR1]

>The image circle for a an EF lens would make a 30.6mm square sensor.

If we're talking about a camera at $8-10K+, why not make a sensor that covers the entire image circle; then let the photographer decide how to crop it. Will that create wasted silicon? Yup, but this is not a consumer camera, and it would eliminate the need for a "vertical" grip.

So you would prefer vignetting in all images, with post-processing on every image to eliminate the black areas by cropping?

 

[Orangutan]

Re: “1Dsq” & 3D [CR1]

>So you would prefer vignetting in all images, with post-processing on every image to
>eliminate the black areas by cropping?

Cropping already occurs, it just happens before image capture. By capturing the entire available scene, the photographer gets to decide after the fact whether it's landscape, portrait, square or oval.

The reasoning is similar to JPEG capture vs. raw with JPEG conversion: if you start with JPEG capture you have limited your data up front. If you capture raw you give yourself more to work with later, even though you will need to do some kind of post-processing.

If you're thinking "I don't want to apply manual cropping to thousands of photos," consider that the software that supports this camera will have features to make this easier.

In short, no, I don't want vignetting, I just want to have maximum data, and never have to rotate my camera to change orientation.

To be honest, though, this is just hypothetical for me since I could not afford $8k for a body.

 

[neuroanatomist]

Re: “1Dsq” & 3D [CR1]

Dilbert, make a circle that touches the edges of a rectangle with an aspect ratio of 1.5 x 1 then make a circle around the squares made up of the 1.5 dimension, and you'll see that the area of coverage is significantly expanded.

As other her noted the square will be smaller than the 1.5 dimension if they went that way but kept the current lenses.

I don't think anyone mentioned that a hypothetical square sensor would be 36mm x 36mm. Obviously, geometry precludes that possibility with current lenses (except the TS-E lenses, which have a large enough image circle).

 

[EYEONE]

Re: “1Dsq” & 3D [CR1]

This is exactly what I was thinking. Why not make the sensor go all they way to the edges of the lens? The you could switch between portrait and landscape at the touch of a button. And it would be FF in both directions.

 

[bvukich]

Re: “1Dsq” & 3D [CR1]

Not to mention I wouldn't want to see the corners on a 36mmx36mm sensor with the lens that don't have the baffle on the back of them.

The image circle for a an EF lens would make a 30.6mm square sensor. Sensor size would be constrained by the hypotenuse (43.266mm), not the width of a 3:2 ratio frame. For the record, a 36x24mm sensor is 864 mm2, a 30.6mm square sensor is 936 mm2. That is an 8% increase of surface area. That, coupled with the fact that a square sensor would have less sensor area outside of the 'sweet spot' of the lens would likely result in slightly sharper images.

OK... now take your 30.6mm square sensor and crop it to a normal AR and you only have 30.6x20.4 and only 624mm^2. Significantly smaller than FF, and just a smidge over APS-H.

 

[KyleSTL]

Re: “1Dsq” & 3D [CR1]

Cropping already occurs, it just happens before image capture. By capturing the entire available scene, the photographer gets to decide after the fact whether it's landscape, portrait, square or oval.

The reasoning is similar to JPEG capture vs. raw with JPEG conversion: if you start with JPEG capture you have limited your data up front. If you capture raw you give yourself more to work with later, even though you will need to do some kind of post-processing.

If you're thinking "I don't want to apply manual cropping to thousands of photos," consider that the software that supports this camera will have features to make this easier.

In short, no, I don't want vignetting, I just want to have maximum data, and never have to rotate my camera to change orientation.

To be honest, though, this is just hypothetical for me since I could not afford $8k for a body.

Well, in the case of a 36x36mm sensor, the manufacturing cost would go up at least 50% over current FF sensors (aside from R&D and other associated costs). And a 43.3x43.3mm sensor would be at least 116% more costly.

In the case of the 36x36mm sensor 4.5% of the sensor would be outside the image circle. For the 43.3mm square it would be 27.3%.

Since CR has been reporting for quite some time that Canon would make a 'radical departure' with the 1Ds replacement, I would say this is slightly feasible. This is no more radical than the idea of a removable back similar to the MF digitals. And Canon, Kodak, Nikon and the like have experience with the selectable aspect ratio in the past with the APS system of cameras.

 

[traveller]

Re: “1Dsq” & 3D [CR1]

Well, in the case of a 36x36mm sensor, the manufacturing cost would go up at least 50% over current FF sensors (aside from R&D and other associated costs). And a 43.3x43.3mm sensor would be at least 116% more costly.

I haven't bothered checking your calculations (!), but whatever the figure I'd have to agree.

Sensor size is very strongly correlated with price; sure it may be technically possible to do multiple aspect ratios with the new 1Ds (a la GH2), but it's all irrelevant unless the market is willing to pay a premium price for this. Are pros really demanding multiple aspect ratios?

All we've heard so far is doubts that the 1Ds series can remain viable at its current price point. Will adding even more expense make it more compelling, or simply drive recession hit pros to cheaper options?

 

[Justin]

Re: “1Dsq” & 3D [CR1]

In terms of cost of production and pricing, look to the Pentax 645D for a comparison of what Canon would be shooting for for a "radical departure" body. The Pentax silicone is quite a bit bigger at 44mm x 33mm. The body is priced at $10k.

Well, in the case of a 36x36mm sensor, the manufacturing cost would go up at least 50% over current FF sensors (aside from R&D and other associated costs). And a 43.3x43.3mm sensor would be at least 116% more costly.

I haven't bothered checking your calculations (!), but whatever the figure I'd have to agree.

Sensor size is very strongly correlated with price; sure it may be technically possible to do multiple aspect ratios with the new 1Ds (a la GH2), but it's all irrelevant unless the market is willing to pay a premium price for this. Are pros really demanding multiple aspect ratios?

All we've heard so far is doubts that the 1Ds series can remain viable at its current price point. Will adding even more expense make it more compelling, or simply drive recession hit pros to cheaper options?

 

[Justin]

Re: “1Dsq” & 3D [CR1]

Thanks for reposting this. I remember reading it awhile ago, and thought it was brilliant. I would love a sensor like this in the flagship Canon or even better, the 5D successor.

"They don't need to build a whole new lens system."

Dilbert, make a circle that touches the edges of a rectangle with an aspect ratio of 1.5 x 1 then make a circle around the squares made up of the 1.5 dimension, and you'll see that the area of coverage is significantly expanded.

As other her noted the square will be smaller than the 1.5 dimension if they went that way but kept the current lenses. I'd say this has a snowball's chance in hell of being true, but hey, Canon have made some really bad decisions lately, like messing up the once wonderful G series, so maybe.

The flaw in your logic is that you're assuming the (new) square sensor will be just as wide as the current sensor.

There's no reason for it to be that way.

Click on and read the "many moons ago" story in that started this thread:
http://www.canonrumors.com/2010/05/the-cmos-sensor-squared-cr2/

Next time, read all of the article and the stories it references and links to first.

 

[gmrza]

Re: “1Dsq” & 3D [CR1]

So?
Why would you crop it to a "normal" AR?
And what is a "normal" AR?
3:2 because that's what 35mm film used?
16:9 because that's what wide screen TV uses?
4:3 because that's what a lot of computer monitors are?

If a 3:2 is always cropped in order to fill a 16:9, then what's the point of a 3:2 sensor?
If a 3:2 photo never fills a 4:3 or 16:9 screen, what's the point of a larger screen?

The point here being that aside from historical bias towards 3:2 there is no reason for digital photographs to be that way.

Aside from the legacy of 35mm film, there is the proximity to the golden ratio that favours the
3:2 aspect ratio. (Actually 16:10 would be closer.)
I get the point though of asking "what is a standard aspect ratio?"
Take a look at a wedding album - you will see all sorts of aspect ratios that can readily be printed.

What will be a technical consideration in working with a square sensor is that the sensor will be
higher (even if only 30.6mm) - quite a bit higher than 24mm. That means that the reflect mirror
and pentaprism need to be totally redesigned (assuming a SLR design).
It has even bigger implications for the shutter. Building a focal plane shutter than has to travel
30.6mm will probably be quite a technical challenge - especially when you want to maintain
a reasonable flash sync speed. I doubt that 1Ds buyers would accept anything less than 1/250sec.
My uneducated guess is that with the same curtain travel speeds, you would not be able to
reach 1/200s on a 30.6mm high sensor. (I stress - uneducated guess.) By increasing
the travel speed of the curtains, you increase the mechanical stresses, which means durability
becomes an issue. Another way around that is to go back to an electronic shutter, which however
is a compromise from an image quality perspective.

Canon would need to do a great deal of engineering to change the sensor aspect ratio. That
said, there has been a long delay since the launch of the 1Ds3.

 

[Osiris30]

Re: “1Dsq” & 3D [CR1]

Ok, so let's see... Canon is going to bring out all new lenses for this square format sensor, and we haven't seen a single square format lens patent... right that does that one in.

As for this 3D.. Anyone who thinks Canon is going to drop their sensel density needs to give their heads a shake. ISO performance and sensel size have *very* little to do with each other, the newest Sony APS-C proves this point.The current Nikon high-iso monster which shall be unnamed is that way due to other design reasons, not the sensel size.

I *really* wish people would get over this silly internet myth that smaller pixels are bad (and it is a myth until a point we aren't close to). You actually improve DR with more smaller sensels due to highlights not being blown to kingdom come.

While I appreciate the rumors didn't originate with CR, I do wish people would apply a little critical thinking to some of them. If you guys *want* something from a sensor how about faster microlenses, something f0.6 or so to allow f1.2 lenses to actually be used to their full advantage in terms of light gathering. The only reason I don't own an f1.2 lens is I can't use it at close to f1.2 from a light gathering perspective (and I really don't need DOF that thin.. not that I couldn't handle it, but I don't need it for artistic purposes).

Give me a 40MP, 4:1 binned (optionally) FF sensor with fast microlenses and I will buy it. Immediately. Give me nice low readout noise and nicely designed wells of a good depth. Move as much of the support circuitry to the back of the sensor to improve sensel density.. these are the things people would be asking for if they stopped to think about how light and sensors *actually* work... oh and if someone (anyone, I'd switch brands for this in a heartbeat), can come up with a way to have the bayer filter be movable so I can put the camera in true monochrome mode, they would be my god (and yes I know it won't happen, but if we're dreaming.. )

1. Which new Sony sensor are you talking about, 16MP or 24MP?

2. What are the "other design reasons" for the relatively low large pixel size in the "Nikon high-iso monster"?

3. As for the "silly Internet myth" about pixel size, go to the Canon USA website and look at their bragging about the sensor for the new M40/M41/M400 camcorders. Coming from Canon, it's downright weird. If pixel size doesn't matter, is binning desirable only to reduce file size? Also, you might want to talk with the people who designed the sensor for the Red Epic movie camera. They're claiming 13.5 stops of DR with relatively large pixels.

4. I didn't know that microlenses even had f-stops. Could you elaborate?

1) The new 16mp APS-C which is light years ahead of any previous Sony/Nikon APS-C despite a massive increase in sensel density pre sq mm.

2) Improvements in the core technology had more to do with the performance there than the raw size of the sensel.

3) No, binning can improve noise, but not due to increasing the sensel size, but because it can lead to a reduction in read out noise. As for Red Epic, they still haven't said exactly what they are doing either.. in fact they won't say much of anything over than quoting the DR.. there are a lot of things you can do to increase DR if you're going to be sneaky about it *and* if you have the luxury of knowing you're never going to have to support 1/8000 as a shutter speed.

4) I'll see if I can find a few relevant articles for you. They aren't f-stop in the sense of an aperture, but they will only gather light up to a certain point. It's a complex subject and well beyond the scope of posting here (hence while I'll find you links).

 

[Osiris30]

Re: “1Dsq” & 3D [CR1]

Dilbert, make a circle that touches the edges of a rectangle with an aspect ratio of 1.5 x 1 then make a circle around the squares made up of the 1.5 dimension, and you'll see that the area of coverage is significantly expanded.

As other her noted the square will be smaller than the 1.5 dimension if they went that way but kept the current lenses.

I don't think anyone mentioned that a hypothetical square sensor would be 36mm x 36mm. Obviously, geometry precludes that possibility with current lenses (except the TS-E lenses, which have a large enough image circle).

Admittedly that's the assumption I've made here. If it's smaller than convention 36mm, then I have to wonder how well it would be accepted (since if you need to crop a 2x3 image you now have a *smaller* sensor area).

 

[Joe J]

Re: “1Dsq” & 3D [CR1]



So?
Why would you crop it to a "normal" AR?
And what is a "normal" AR?
3:2 because that's what 35mm film used?
16:9 because that's what wide screen TV uses?
4:3 because that's what a lot of computer monitors are?

If a 3:2 is always cropped in order to fill a 16:9, then what's the point of a 3:2 sensor?
If a 3:2 photo never fills a 4:3 or 16:9 screen, what's the point of a larger screen?

The point here being that aside from historical bias towards 3:2 there is no reason for digital photographs to be that way.
[/quote]

Agreed, 100%