Panasonic. Sony. Canon. JVC. They do that in their video cameras. I would love to have a nice big 3CCD still camera! Here's a thought, you know who would be the ONLY DSLR maker to remotely be in a position to do it.... Olympus!
I have to ask though, how many people would put all their faith in a brand new technology like truecolor and buy the 1st generation camera? Nikon D1, Canon D30, the Sigmas. All brand new, unproven technology. For all the talk, people were a little hesitant. You read it everyday on the boards.
Panasonic. Sony. Canon. JVC. They do that in their video cameras. I would love to have a nice big 3CCD still camera! Here's a thought, you know who would be the ONLY DSLR maker to remotely be in a position to do it.... Olympus!
I haven't dug down into the mysteries of 3CCD gadgets, but why do you say Olympus? Because they use a smaller sensor?
nikt wrote:
I have to ask though, how many people would put all their faith in a brand new technology like truecolor and buy the 1st generation camera? Nikon D1, Canon D30, the Sigmas. All brand new, unproven technology. For all the talk, people were a little hesitant. You read it everyday on the boards.
All that would be new is the sensor. DIGIC III is an evolution. The AF chippery is too. We know what to expect of the 10/20/30D build, so that's not a gamble either. All your EF lenses will work on the 40D, no surprises there. So the gamble isn't as big as the first of the line, the D1, the 1D, the D30 and so on.
If the image quality is there? If the speed is there? If the high ISO performance is there? If the dynamic range is there? I'll take it!
(Well, no, I won't, I'll have to settle for something second hand, because I can't afford it financially. It's not because I'd be hesitant.)
Chris Hauser wrote:
From what do we know that Canon doesn't mean that the 10Mp of the 40D are real, and not fake like fom sigma ?
Because than it would be much bigger than anything else :-P
Well, the Sigma is not entirely fake - but we need to sort out if the nomenclature is the same as Sigma's.
The Sigma figures probalby overestimate their equivalence to Bayer rather, but a 6MP Foveon would give about the same rez for most pruposes as a 10MP bayer.
A 10MP foveon would exceed the 1DsIII - and have bigger files
Regards,
DaveMart
Well, the Sigma is not entirely fake - but we need to sort out if the nomenclature is the same as Sigma's. The Sigma figures probalby overestimate their equivalence to Bayer rather, but a 6MP Foveon would give about the same rez for most pruposes as a 10MP bayer. A 10MP foveon would exceed the 1DsIII - and have bigger files.
Just like a bayer pattern, Sigma (or rather, Foveon) gives the number of photo sites. That seems fair to me, photo sites are photo sites. Bayer conversion to pixels will interpolate each cluster of 4 photo sites to a pixel. Then shift by one unit, and interpolate again, re-using 2 values of the previous interpolation. A Foveon 3-layer sensor doesn't need to do this, obviously.
So if you talk about a 10 Mpixel Foveon sensor, then I assume you mean 10 Million photo sites, so 10 / 3 = 3.333 Mpixel per layer.
Your numbers suggest that you actually mean 6x3 or 10x3. If you take the 10:6 ratio from the SD10 and 10D comparison then you can guesstimate that a 6x3 Foveon sensor would be the equivalent of a 10.8 Mpixel Bayer sensor and 10x3 Foveon would be 30 / 1.6666 = 18.
Assuming that the lossless compression ratio is the same (which is a strech, since the spacial information isn't the same between a bayer and a foveon sensor, so it may cause it to differ) then the number of photo sites and the bit-depth will determine the RAW file size.
A 6x3 foveon will have to store 18 million photo site values at 12 bits. An equivalent Bayer sensor will be 11 or 12 million photo sites, also at 12 bit per output value. Bonus points for Bayer, one would say at first glance. But perhaps not so, because the foveon output will contain more color information, it doesn't have to interpolate, so that might be a bonus in RAW conversion.
nikt wrote:
I have to ask though, how many people would put all their faith in a brand new technology like truecolor and buy the 1st generation camera? Nikon D1, Canon D30, the Sigmas. All brand new, unproven technology. For all the talk, people were a little hesitant. You read it everyday on the boards.
I'm very interested in the Foveon style technology. I've always thought it was a more elegant design. However, I also think it needs some maturing and it needs to be in a body that I can use with my Canon lenses.
I'm seriously considering the Sigma DP1 and will try one as soon as I see it, but I'd rather it had a good zoom lens preferably with optical IS.
I'll also check out the SD14 when it becomes available but the inability to use my Canon lenses on it will most likely keep me from buying it.
Since a Bayer sensor needs 4 sites and a lot of interpolation overhead to create enough information for a pixel and the Foveon requires only 3 sites and less overhead, any megapixel comparison between the two seems to me to be pointless.
I'd think the Foveon RAW files would be very interesting. I'm not that much of in a hurry so I can wait a few more milliseconds for them to be stored in a buffer.
Edit: Something else I just thought of, since per unit area the Foveon sites could be larger, defraction limiting becomes less of an issue.
Tentacle wrote:
Just like a bayer pattern, Sigma (or rather, Foveon) gives the number of photo sites. That seems fair to me, photo sites are photo sites. Bayer conversion to pixels will interpolate each cluster of 4 photo sites to a pixel. Then shift by one unit, and interpolate again, re-using 2 values of the previous interpolation. A Foveon 3-layer sensor doesn't need to do this, obviously.
So if you talk about a 10 Mpixel Foveon sensor, then I assume you mean 10 Million photo sites, so 10 / 3 = 3.333 Mpixel per layer.
Your numbers suggest that you actually mean 6x3 or 10x3. If you take the 10:6 ratio from the SD10 and 10D comparison then you can guesstimate that a 6x3 Foveon sensor would be the equivalent of a 10.8 Mpixel Bayer sensor and 10x3 Foveon would be 30 / 1.6666 = 18.
Give or take some, of course. ...Show more →
I'm just faking it around all the different interpretations of how the different numbers equate, and being studiously vague to try to avoid hair-splitting debate on foveon - I've been in a few of those.
Really, alll this discussion of equivalents is arising from dcmiller's post, regarding the use of a truecolour sensor on the Canon.
I can't see Canon coming out with a 3MP truecolour chip at this stage, and as far as I can work out dcmiller is talking about one with 30million photosites, but that seems far too high to work - he still has not got back to us on wherte his info is from.
Something like a 6MP chip with 18million photosites would work, and is probably do-able.
What do you think?
Regards,
DaveMart
I'm just faking it around all the different interpretations of how the different numbers equate, and being studiously vague to try to avoid hair-splitting debate on foveon - I've been in a few of those.
Really, all this discussion of equivalents is arising from dcmiller's post, regarding the use of a truecolour sensor on the Canon. I can't see Canon coming out with a 3MP truecolour chip at this stage, and as far as I can work out dcmiller is talking about one with 30million photosites, but that seems far too high to work - he still has not got back to us on wherte his info is from.
Something like a 6MP chip with 18million photosites would work, and is probably do-able. What do you think?...Show more →
Currently the mainstream lithography feature size is 90 nanometer. Intel, AMD, IBM and the like are at 65 nm and have 45 nm in prototype up, or nearly up and running...
Hmmm... Okay, now, zoom out, factor 10. 900 nm. Zoom out 10 times again, 9000 nm, or rather, 9 micron. (µm)
Now, imagine 6 mpixels on a layer of 36x24 mm. That's 6,000,000 units on 864 mm², so 0.000144 mm² per unit. Take the square root and you end up with a pixel pitch of 0.012 mm. That's 12 µm. A mere 133 times bigger than the current mainstream lithography size.
Yes, if I'd hazzard a guess, a 6x3 sensor is a piece of cake. A 10x3 sensor too because that would have a unit pitch of 9.3 µm. Still a comfortable 100 times bigger than the feature size.
Compare this to a Nikon D2x: 12.2 Mpixel on 1.5x crop, that's 5.5 µm pixel pitch.
Tentacle wrote:
Currently the mainstream lithography feature size is 90 nanometer. Intel, AMD, IBM and the like are at 65 nm and have 45 nm in prototype up, or nearly up and running...
Hmmm... Okay, now, zoom out, factor 10. 900 nm. Zoom out 10 times again, 9000 nm, or rather, 9 micron. (µm)
Now, imagine 6 mpixels on a layer of 36x24 mm. That's 6,000,000 units on 864 mm², so 0.000144 mm² per unit. Take the square root and you end up with a pixel pitch of 0.012 mm. That's 12 µm. A mere 133 times bigger than the current mainstream lithography size.
Yes, if I'd hazzard a guess, a 6x3 sensor is a piece of cake. A 10x3 sensor too because that would have a unit pitch of 9.3 µm. Still a comfortable 100 times bigger than the feature size.
Compare this to a Nikon D2x: 12.2 Mpixel on 1.5x crop, that's 5.5 µm pixel pitch....Show more →
Noise issues tend to kill it first, anyway, the suggestion from dcmiller is for 10MP (Sigma pixels, ie 30m photosites) on a 1.6 sensor, not FF.
6MP, sigma style, and 18million photosites sounds a lot more in line with present technology, but what do I know?
Regards,
DaveMart
Noise issues tend to kill it first, anyway, the suggestion from dcmiller is for 10MP (Sigma pixels, ie 30m photosites) on a 1.6 sensor, not FF.
6MP, sigma style, and 18million photosites sounds a lot more in line with present technology, but what do I know?
We know nothing. But we can try to make educated guesses
Ok, so re-do the math, but then for 1.6x crop. That "robs" the sensor of 1.6² the surface: 337.5 mm² instead. That's a pitch of 7.5 µm for a 6x3 sensor and 5.8 µm for a 10x3 sensor. Still plenty of space.
But all of this is silly talk for a different reason. Yes, it can be done. Sharp managed to pry more than 8 mpixel from a 1/2.5" sensor. That's a pitch of 1.75 µm. So size is not the point.
Canon isn't stupid. Why would they allow the 40D to make such a huge jump to 10x3 ? That's very bad marketing because it will rob you from a lot of sales. Talking about canibalizing your own market.
A 6x3 gets around 11 Mpixel equivalent, and a 7x3 sits around 12 to 13. Oh, 8x3 perhaps, for a nice wide gap to nearly 15 Mpixel equivalent. That'll hit the current 10 Mpixel competition AND the next generation 12 Mpixel. It's for this reason that I don't see Canon put a 10x3 Foveon-like sensor in a 40D, it'd be overkill.
Tentacle wrote:
We know nothing. But we can try to make educated guesses
Ok, so re-do the math, but then for 1.6x crop. That "robs" the sensor of 1.6² the surface: 337.5 mm² instead. That's a pitch of 7.5 µm for a 6x3 sensor and 5.8 µm for a 10x3 sensor. Still plenty of space.
But all of this is silly talk for a different reason. Yes, it can be done. Sharp managed to pry more than 8 mpixel from a 1/2.5" sensor. That's a pitch of 1.75 µm. So size is not the point.
Canon isn't stupid. Why would they allow the 40D to make such a huge jump to 10x3 ? That's very bad marketing because it will rob you from a lot of sales. Talking about canibalizing your own market.
A 6x3 gets around 11 Mpixel equivalent, and a 7x3 sits around 12 to 13. Oh, 8x3 perhaps, for a nice wide gap to nearly 15 Mpixel equivalent. That'll hit the current 10 Mpixel competition AND the next generation 12 Mpixel. It's for this reason that I don't see Canon put a 10x3 Foveon-like sensor in a 40D, it'd be overkill....Show more →
Actually 8 by 3 fits in with one of the figures dcmiller gave - he said it 'might' be 3512 pixels across- it's the data throughput that gets me though - 8 *3* 5fps = around 120p/sec - at normal Canon compression ratios about the same number of MBs 3 times as much as the curent 30D, and twice the 1DsII!
The equivalent figures fro the 1DsIII are not so surprising, although high, but memory and processing costs in the camera are not too important in this class of camerra - although it could be tough to get perhaps 200MB/sec on to a CF or SD card.
I don't think I really believe it, but we will see...
Regards,
DaveMart
Perhaps I wasn't paying attention 18 -20 months ago, but I recall the 5D being a surprise - so perhaps the 40D will be an XTi sensor in a 30D body, perhaps with Digic3 just for fun.
And there will be a 5Dn or 1DnMk3 with a 'truecolor/foveon' chip?
Any idea on cost of manufacture? Canon has no D40 competitor, and they often relese new tech in the consumer bodies (the dust bit in the 400XTi being the most recent example).
