Pages: [1]   Go Down

Author Topic: The 10 MP Foveon sensor rumor  (Read 10294 times)

  • Guest
The 10 MP Foveon sensor rumor
« on: March 17, 2003, 10:15:39 am »

My sources tell me that the chip is in sampling at National Semi (who is the fabricator) but that no camera maker has yet committed to it.

Apparently Nikon has been considering it, and has shown prototypes to some people, but is considering other options as well for its next, and much delayed, camera (D2?).
Logged

Quentin

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 1222
    • Quentin on Facebook
The 10 MP Foveon sensor rumor
« Reply #1 on: March 17, 2003, 04:19:35 pm »

I always hoped Contax might jump for a larger Foveon sensor.  Contax users are generally not sports photogs with high ISO requirements.  A 10mp Foveon would possibly suit Contax very well.

On the other hand, I do not really see the point in cutting the sensor down to a 1.5 crop factor for Nikon.  Why not use the whole chip?

Quentin
Logged
Quentin Bargate, ARPS, Author, Arbitrato

Marshal

  • Guest
The 10 MP Foveon sensor rumor
« Reply #2 on: March 18, 2003, 01:36:41 am »

Nikon needs something that can compete squarely with the 1Ds in both image quality and sensor size.

A 7.9Mp Foveon equipped camera would help trump the 10D for that price range, but 10-12Mp, full-frame sensor would be best for the top of the line.
Logged

Quentin

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 1222
    • Quentin on Facebook
The 10 MP Foveon sensor rumor
« Reply #3 on: March 18, 2003, 06:35:42 pm »

This months Pro Photographer mag in the UK carries a short report on the Nikon 12-24 DX lens.  

What is interesting about this are the reported comments of the UK Marketing manager of Nikon about their future plans for digital SLRs.  

He is quoted as saying that Nikon has no plans for a full frame DSLR sensor.  He says that the next high-end Nikon DSLR will have a 1.5 crop factor sensor, to be known from now on as a "DX" sensor.  He says that the smaller sensor allows better use of digital technology, and that Nikon will not be "constrained" by the requirements of the 35mm format.

Any idea that Nikon will announce a 35mm size sensor camera seem dead in the water for the forseeable future, in view of these comments.  

So we have an interesting "battle" shaping up between three alternative views of the digital SLR world:

-  The full 35mm frame camp, populated by Canon, Kodak and Contax;
-  The 1.5 crop factor camp, with Nikon and possibly Fuji, (but who knows?)
-  the four-thirds camp, with Olympus as its standard bearer - but with Kodak and Fuji having a toe-hold as supporters of the format.

Quentin
Logged
Quentin Bargate, ARPS, Author, Arbitrato

BJL

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 6600
The 10 MP Foveon sensor rumor
« Reply #4 on: March 19, 2003, 10:12:14 am »

Ray reminds me of one factor I ignored in the "only one extra lens" argument: how does the combination of greater magnification needed in printing, using a smaller focal length, and using only a smaller central part of the image produced by the lens play out in image quality for smaller format sensors? Will some photographers need a whole array of lenses with better performance at higher lp/mm? All this assuming that the pixels are small enough for lens resolution to be significant to the final image quality.

The direct comparison might be something like this example: in 35mm format, measuring resolution with MTF at the standard 40lp/mm, I get at least 60% MTF all the way to the corners (21mm from center) for focal length of 60mm and a certain f-stop.  For the newly named DX (1/1.5) format, the linear size scaling by factor 1/1.5 means that the MTF should be measured at 60lp/mm instead of 40 [tougher on the smaller sensor], going out only to a distance of 14mm from the center instead of 21 [easier on the smaller sensor], and for 40mm focal length instead of 60. With all these competing factors, real experimental data are needed to decide what overall affect this change will have on resolution. (I looked at just a single set of MTF graphs, for the Canon 28-70L, and it was close enough to a tie.)
Logged

BJL

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 6600
The 10 MP Foveon sensor rumor
« Reply #5 on: March 23, 2003, 10:58:15 pm »

I suspect that the potential advantages of a 6MP 35mm format sensor will never be realized for economic reasons: the price of such a sensor plus the big anti-aliasing filter it needs will be almost as high as a 10MP+ sensor of the same size, and significantly more than a 6MP+ "half-frame" sensor, since cost is influenced far more by chip size (and AA filter size) than by pixel count. Given the rather good noise and dynamic range performance of the Canon 1Ds, not many people would settle for 6MP FF at almost the same price.

Pardon me for bringing up Moore's law again, but that pricing phenomenon is driven mostly by getting the same performace out of ever smaller chips; I expect that commercially viable chips of a given pixel size will tend to shrink somehat as S/N performance improves at a given pixel size, with only ever higher pixel counts being marketable in the more expensvive "big chip" formats.
Logged

rokkitan

  • Newbie
  • *
  • Offline Offline
  • Posts: 14
The 10 MP Foveon sensor rumor
« Reply #6 on: March 24, 2003, 03:34:15 am »

Moore's law actually just states that the number of transistors per chip will double every 18 months.  This comes because of two factors: decreasing size of transistors and increasing size of chips.
I belive that the chips for 35mm slr's will be fullframe and have enough resolution that the lens will be the antialiasing filter in not to many years.

Edited to add:
I believe the price will be equivalent with todays film/analog-SLRs, but if they'll be 35mm or smaller I don't know.  The resolution of 35mm chips will increase quite a bit I think, and settle at perhaps 50MPixels (or perhaps 25Mpixels i Foveon or similar becomes the norm) if the dynamic range and noise is respectivly large and low enough.  Storage, processing power and chips area are becoming cheaper all the time and will not be a factor in 5-8 years.
I want a fullframe (Eos10Ds?), but I also want a small camera.  Perhaps the Olydak or similar will be the best compromise.  The Eos 10D will probably be my first dSLR since I have an Eos30(Elan7e).  Eos3D, with 8-9Mpixels and 1.3x would be a nice compromice, but I fear even that will be too expensive this and next year.
Logged

Ray

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 10365
The 10 MP Foveon sensor rumor
« Reply #7 on: March 27, 2003, 12:01:00 am »

Quote
Kodak thinks that having 13.5 million 8 micron pixels is enough to need no AA filter in 35mm format (with Nikon lenses at least!) and the reviews I have seen so far have not criticized the 14n for aliasing problems at least, so perhaps 35mm format sensors will settle around there or a bit above, moving slowly up in pixel count if pro lenses improve to keep up.
BJL,
Well, you've probably read further reports by now that indicate the 14n does in deed have aliasing problems, and I think this was expected all along. The 14n's pixel pitch is just slightly larger than the D60's and 10D's, and therefore slightly more prone to aliasing than would be the case if the D60 had no AA filter. Canon thought it necessary to provide the D60 and 10D with an AA filter. 13.5MP full frame is clearly not sufficient to eliminate this problem, but fortunately it's not a problem in every shot and I don't criticise Kodak for leaving the AA filter out.

But I must confess I'm beginning to change my mind about the advantages of having such a high pixel density that the lens itself becomes the only AA filter that's needed. This noise problem the 14n exhibits at anything other than fast shutter speeds at ISO 80 has shaken my confidence.

I thought Canon did an amazing job when they doubled the pixel density of the successor to the D30, the D60, without increasing noise or reducing dynamic range. Kodak is finding it more difficult. The 14n seems to be suffering from (under certain circumstances) what I would call the 'small pixel syndrome'.

I'm beginning to wonder if it'll ever be possible to make a 24MP full frame sensor that does not compromise S/N and DR. And if it does prove to be possible, would I perhaps not prefer a 12MP or at most 16MP camera with an extra couple of stops of DR, rather than a very marginal increase in resolution. One should bear in mind that every doubling of the pixel count does not increase resolution by a factor of 1.4. There's a law of diminishing returns, or more to the point, a law of diminishing contrasts. ???
Logged

BJL

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 6600
The 10 MP Foveon sensor rumor
« Reply #8 on: March 27, 2003, 10:53:17 am »

Quote
I want a fullframe (Eos10Ds?), but ... The Eos 10D will probably be my first dSLR ... Eos3D, with 8-9Mpixels and 1.3x would be a nice compromise, but I fear even that will be too expensive this and next year.
In that balancing act, I think you are far from alone. To oversimplify, I see a split: lots of people saying that they want a 35mm format DSLR, but most of them not near to buying yet; lots of people saying that the 10D (with maybe the *ist D and the "OlyDakUji") has finally brought DLSR into their price range and so they are buying NOW. I think I already see the start of a surge in sub-35mm format DSLR sales that will rapidly drive technology, choices, economies of scale and price advantage in that format; I honestly do not see 35mm format surviving that assault below the EOS-1/F5/MF sector.
Logged

BJL

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 6600
The 10 MP Foveon sensor rumor
« Reply #9 on: March 29, 2003, 08:54:20 pm »

Quote
Some time ago, a rumor, supposedly with a good degree of credibility, was published  on this site about a FF 10 MP Foveon sensor to be used by a major company ...
Sorry for the three posts in a row, but I just read a report at DPReview http://forums.dpreview.com/forums....4584111
that Foveon's Chief Scientist had said at a public presentation that they have no current plans for a 35mm format sensor (henceforth I will call this format "double frame" harking back to when "single frame" 35mm was 18x24mm, as a protest against  taking 24x36mm frame size format as the ideal by which all other formats are measured.) Apparently the yields on such large chips are too low to be economical for them. Could the failure to achieve good yields have killed the once rumored Nikon/Foveon "double frame" format DSLR?
Logged

H.E.

  • Guest
The 10 MP Foveon sensor rumor
« Reply #10 on: March 17, 2003, 05:59:38 am »

Some time ago, a rumor, supposedly with a good degree of credibility, was published  on this site about a FF 10 MP Foveon sensor to be used by a major company. Since there was nothing like that at PMA, what is the status on that rumor? Is it still credible (has it been modified?) or has it been discredited?

H.E.
Logged

H.E.

  • Guest
The 10 MP Foveon sensor rumor
« Reply #11 on: March 17, 2003, 03:36:29 pm »

Thanks Michael for the info,

Actually one reason I asked was that there were some rumors recently about Nikon trying a Foveon chip. However, the most specific rumor was for a 1.5x format sensor with 7.9MP. This number of pixels, while not consistent with the pixel size in the SD9, is consistent with a the pixel size mentioned in a relatively recent Foveon patent with a new photosite structure. Foveon mention in the patent that  their design alows 7 micron pixels with 50% utilization using 0.18 micron process. It could be good news that Foveon managed to give Nikon something good enough to make it that far in their testing process.

thanks again,
H.E.
Logged

BJL

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 6600
The 10 MP Foveon sensor rumor
« Reply #12 on: March 17, 2003, 07:36:49 pm »

Quote
On the other hand, I do not really see the point in cutting the sensor down to a 1.5 crop factor for Nikon.  Why not use the whole chip?
I suppose this debate is going to be around for quite a while, but I would say that, for now, the price premium for cameras with "35mm format" sensors over the 1.5 smaller formats is almost US$3000, so the point could be that the smaller format X3 sensor could probably give many photographers, even pro's, all the image quality that they need (comfortably more than 35mm film) for a lot less than current "35mm format" DLSR's.  Even if you have to buy one extra reduced format wide angle lens that might be useless after a future shift to a larger format, the package could save almost US$2000. Isn't that a significant saving, even for some pro's?
Logged

BJL

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 6600
The 10 MP Foveon sensor rumor
« Reply #13 on: March 18, 2003, 10:25:57 am »

On image quality: pixel counts cannot be compared directly between Foveon's "X3" approach and the traditional Bayer pattern style; experience so far suggests that about a 2:1 ratio, so a 7.9MP X3 sensor MIGHT match up to 16MP Bayer pattern sensor in image quality. Also, X3 sensors apparently have the cost advantage of having little or no need for anti-aliasing filters, which are VERY expensive for 35mm format ($2000?), and which will not benefit from the rapid price improvements of electronic components.

On sensor size: it seems to me that the smaller sensor only requires BUYING at most one additional special lens, to cover the factor of 1.5 in field of view lost at the wide end; a lens like the Nikon 12-24mm DX. It would not seem to require CARRYING any more lenses unless you mix formats in your kit, and would often save on weight and bulk by allowing one to carry a somewhat smaller longest lens; for example 400mm instead of a 600mm.
Thus it seems economically sound to weight the extra cost of one additional lens ($1200?) against cost savings elsewhere ($2000+?), rather than insisting absolutely on being able to use exactly one's current lens selection.

So it seems somewhat likely that it will be feasable to offer a top of the line pro level DLSR body using such a "1.5X smaller" sensor that matches or significantly exceeds the resolution of the 1Ds, at a total cost of significantly less even when one adds the price of that one special wide angle lens. I would guess a "Nikon D2" could start at the same price as the D1x did, about $5500, so about $6700 with one DX lens. That could be a completely competitive package in both price and performance. (I will also speculate that other technologies like Fuji's latest SuperCCD could soon match 1Ds quality in the smaller "Nikon" format, with the reduced pixel size eliminating the need and cost of an AA filter.)


Having, I hope, dealt with the issue of wide-angle coverage, the only other persuasive criticisms of sub-35mm formats that I have heard are reduced dynamic range and reduced view-finder image size and/or brightness (in the D60). Dynamic range is a "wait and see" for me, as electronic improments keep helping there. On viewfinder performance, a fair comparison would be to the Nikon D1 series viewfinders, not those of the lower level Canon D60 body: can anyone comment on how those Nikon D1 series viewfinders perform compared to top 35mm format viewfinders?


P. S. Arguments about smaller sensors having more severe diffraction limits on the smallest usable f-stop are completely invalid, since that effect scales with sensor size exactly the same way the DOF does: f/11 replaces f/16 in both the degree of diffraction limitation and the DOF.
Logged

Ray

  • Guest
The 10 MP Foveon sensor rumor
« Reply #14 on: March 19, 2003, 12:33:10 am »

Quote
On sensor size: it seems to me that the smaller sensor only requires BUYING at most one additional special lens, to cover the factor of 1.5 in field of view lost at the wide end; a lens like the Nikon 12-24mm DX. It would not seem to require CARRYING any more lenses unless you mix formats in your kit, and would often save on weight and bulk by allowing one to carry a somewhat smaller longest lens; for example 400mm instead of a 600mm.
BJL,
Now here's something that I've never seen quantified. A 6MP full frame sensor with, say, 160mm lens, should produce an 'apparently' sharper print than a D60 with a 100mm lens because the full frame image, whatever the final size of the print, needs to be enlarged less.

By this reasoning, the additional ultra wide angle zoom that owners of a D60 often feel compelled to buy, might not fully compensate for the 1.6x factor.

But I'm not sure about this. A pixel is a pixel. The D60 pixel seems to be on a par, quality wise, with the larger 1Ds pixel, except for really long exposures. Does the additional enlargement required of the D60 pixel make it inferior to the 1Ds pixel? The only full frame 35mm sensors I know of are the Contax and MF backs. The dynamic range and S/N of 35mm sized MF sensors is greater than the 1Ds. Don't know about the Contax. Perhaps Canon could have produced a full frame 6MP D60 with significantly better DR and S/N than the 1Ds.
Logged

Ray

  • Guest
The 10 MP Foveon sensor rumor
« Reply #15 on: March 19, 2003, 12:38:02 am »

I meant the only 6MP full frame sensors I know of are the Contax and MF backs.
Logged

Ray

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 10365
The 10 MP Foveon sensor rumor
« Reply #16 on: March 20, 2003, 08:32:02 pm »

Now here's a new angle which I had previously not given much thought to. Many people expected Canon's follow-up to the D30 would be a full frame 6MP (or more) and were disappointed the D60 retained the the 1.6 multiplier.

It has always seemed to me the advantages of the smaller sensor (within reason) outweighs the disadvantages. The cropping factor not only provides for a greater telescopic effect, but effectively upgrades virtually any lens by cropping out the edges and corners where performance invariably falls off to some degree. And as BJL has suggested, the lack of wide angle capability can be remedied by the acquisition of just one ultra wide angle lens such as the Canon 16-35 or Sigma 15-30. Furthermore, if the performance of such wide angle lenses tends to be even worse than that of most lenses at the corners and edges, then that is also taken care of by the cropping factor. One is on a winner whichever way you look at it.

However, there's another angle to this. Let's compare a D60 size pixel (about 7.5 microns) with the size of pixel you would get if the sensor were full frame 6MP - about 12 micron pitch. Let's assume that both pixels, because of different technologies, fabrication, whatever, are of equal value in terms of DR, S/N, colour accuracy and so on. Does the 6MP full frame sensor then have any qualitative advantage, apart from it's wider angle capability?

Well, it seems to me it does. As BJL has pointed out, the performance of the smaller sensor is dependent on the performance of the lens at aprrox. 60 lp/mm as opposed to 40 lp/mm. Not only that, but its performance at 30lp/mm as opposed to 20 lp/mm, its performance at 15 lp/mm as opposed to 10 lp/mm and so on.

Look at any MTF graph of any lens and you'll see two broad shifts in performance. One is towards lower contrast at the corners (or edge of the image circle), indicated by the curve dipping towards the bottom right of the graph,  and the other is towards lower contrast at the specified higher resolutions (indicated by the different curves being lower down the vertical axis). The loss of contrast at 40 lp/mm is invariably greater than the loss of contrast at 30 lp/mm, and so on.

What is the impact of these lens effects on the larger sensor? Seems to me, a more contrasty image and, in so far as contrast relates to perceived sharpness, a sharper image - with the exception of the corners and edges.

Can the more expensive lenses achieve a performance at, say, 20 lp/mm in the area between 15mm and 20mm from the centre of the image, that is equal to or better than the performance at 30 lp/mm but closer to the centre?

Difficult to judge from MTF curves alone, mainly because there's often such a wide variance between diagonal (sagital) resolution and tangential resolution. Generally, I'd have to say that, if evenness of performance from edge to edge is the criterion, the smaller format sensor with the bigger lens is the winner. On the other hand, the centre of interest in most photos is probably away from the edges and therefore loss of contrast and perceived sharpness in these areas may be of little consequence, in which case the bigger pixel of equal quality is the winner (all else being equal).

One final point; we expect larger pixels to have a higher performance. I've just compared two different sized pixels of equal performance. The D60 pixels appear to me to be on a par with the 1Ds pixels, the advantage of the 1Ds being that it has more of them. But these pixels don't differ greatly in size. The pixel pitch of a full frame 6MP sensor should be significantly larger and should therefore have significantly higher DR and S/N. Combine this with the advantages I've outlined above and - all else being equal, except the price!! - the full frame 6MP sensor should win hands down.
Logged

Ray

  • Guest
The 10 MP Foveon sensor rumor
« Reply #17 on: March 23, 2003, 11:44:30 pm »

Quote
I suspect that the potential advantages of a 6MP 35mm format sensor will never be realized for economic reasons:
Agreed! However, the 6MP figure was just an illustrative example. With the advent of the 1Ds and 14n, 6MP is now too little for full frame, but do not the same relativities still apply, whatever the pixel count?

Seems to me there's a balancing act going on. Samir referred to this in another thread. One has to weigh the advantages of increased resolution against increased noise. Whatever state-of-the-art technology can 'eke out' from the smaller pixel, should (logically) be able to be applied to the larger pixel for an equally significant improvement, but from a 'higher' vantage point.

Cost is another matter. Does anyone doubt that a 24MP 4x5 format sensor would produce superior results to a 24MP 35mm format? I'm talking theoretically. I imagine a 4x5" sensor is currently either an impossibility or unrealistically expensive.
Logged

BJL

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 6600
The 10 MP Foveon sensor rumor
« Reply #18 on: March 24, 2003, 10:29:47 am »

Merging comments from Ray and rokkitan, perhaps one part of the balancing act will be that increases in processing speed will eventally lead to an optimal pixel size for a given sensor size that is dictated mainly by the goal of eliminating the need for that very expensive, Moore's Law immune, anti-aliasing filter. Pixel sizes can then push a bit beyond that level where lens limits are significant, with gradual but slower resolution improvements. Kodak thinks that having 13.5 million 8 micron pixels is enough to need no AA filter in 35mm format (with Nikon lenses at least!) and the reviews I have seen so far have not criticized the 14n for aliasing problems at least, so perhaps 35mm format sensors will settle around there or a bit above, moving slowly up in pixel count if pro lenses improve to keep up.

This "lens resolution limit" seems to set in at somewhat smaller pixel sizes for smaller sensors. I can only speculate on the causes; maybe optical effects like smaller focal lengths and only needing the image closer to the center, and the greater economic feasability of maintaining somewhat finer manufacturing tolerances in smaller lens components. Whatever the reason, so-called "APS" or "Nikon DX" format DSLR's with pixels as small as 7.4 microns have AA filters, and some digicam lenses get most of the theoretically available resolution from pixels as small as 3 microns.

This goes with numerous other effects that favor gradually increasing pixel size with larger sensors (another being the maximum DOF possible before diffraction limitation sets in), so I would predict a continuation of the already well established pattern that the preferred pixel count and pixel size both increase with sensor size.

So perhaps in the next few years, preferred pixel counts/sizes will settle at very roughly 5-7MP and 3-3.4 microns for the 2/3" sensors at the top of the compact digicam market, 6-10MP and 5-6.8 microns for 4/3" format, 8-11MP and 6-7 microns for what Nikon now calls "DX" format, 13-16MP and 7-8 microns for 35mm format, 20MP or more for the new wave of true medium format size sensors. I am ruling out any current pixel sizes that need AA filters.

I am fairly sure that, despite a healthy trend of price reductions, the larger format digital cameras and chips, producing far better than 35mm film quality needed only by a small fraction of the total market, will never have the economies of scale and convenience to dominate below the high-end pro quality market in competition with somewhat smaller formats that also match or exceed 35mm quality. So I expect that even moderately serious photographers will have to decide where to place themselves along this scale of sensor size, resolution, dynamic range, price and bulk, much as we have since some pros (including even Ansel Adams on occasion) started using early Leica and Contax 35mm cameras for some types of serious photography.


P. S. there indeed exist sensors far bigger than 35mm format with up to 50 million or more large pixels; their prices do seem to be "stratospheric", in that they are probably used mostly in satellites.
Logged

BJL

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 6600
The 10 MP Foveon sensor rumor
« Reply #19 on: March 27, 2003, 09:41:01 am »

It surely is risky "making predictions about the future"; yes maybe getting excellent S/N ratio and good dynamic range will always require the expense on an AA filter at least in very high end cameras. But then again, the sensor in the 14n is almost a worst case design for noise: CMOS, but without the special extra circuitry that Canon uses in its CMOS sensors to overcome the noise problems traditionally suffered by that technology.

So there are still at least three chances for getting "small, quiet, dynamic pixels" that need no AA filters: more sophisticated on-chip NR, CCD improvements like Fuji's SuperCCD HD, and X3 sensors (an official generic name, not just a Foveon product), which apparently have less or no need for AA filters.

Perhaps sensor choice will develop camps like transparency versus negative film based partly on different priorities for wide exposure latitude.

Could there be market for pro digital cameras with swappable sensors to emulate film choices if signal processing flexibility is never enough? Or at least an array of models that vary mostly in which "digital film" comes pre-loaded? Nikon made a small step with the D1h and D1x options and now seems to be pushing a "DX format" size standardization, perhaps at least to allow a variety sensor options in essentially the same body: "let the sensor technologies fight it out in the market place, so long as they are all available in Nikon bodies."

[Misplaced comment deleted.]
Logged
Pages: [1]   Go Up