Now that Sony has killed their superb fixed-lens R1, and no one else is producing such a large sensor fixed lens zoom, my hopes for someday upgrading to a 20+ MP R2 or R3 lie in ruins. That's too bad, as the ultra short backfocus lens design has some real advantages over SLR lens mounts. However that is not my primary reason for choosing the R1 over a dSLR.

What I want is a live LCD higher megapixel camera. The advantages of composing with the LCD instead of an optical viewfinder are numerous:

1) autofocus using a slower but more reliable contrast-detection system for landscape shooting (never any front/back focus issues).

2) You see the cameras sensor image with its limitations on DR instead of your eye with unlimited DR. Thus you get a sense of the captured file.

3) Camera does not have to be glued to your face! Makes shooting at kids eye level easier, frog eye level MUCH easier! Composing landscapes can be done with both eyes open, looking at the scene and at the image instead of peeping through a tunnel. No need to crouch down to tripod height if you are shooting low, and you can use a very tall tripod and see what you are composing.

4) Get your exposure right the FIRST TIME, EVERY TIME with a live histogram.

5) Of course, the camera should ALSO have a standard mirror and standard dSLR autofocus / metering for all the times when the TTL optical view is an advantage.

6) Video. While the R1 does not allow video recording, virtually all other fixed-lens live LCD digital cameras do. This is a great feature to have, and should be included in future digital cameras.

Olympus has done most of this this with the E330 and Fuji to a lesser extent with the S3. I believe even one of the Canon astrophotography cameras can do some of these things. Come on, let's add this to all dSLRs. Geez! I feel that Canon and Nikon are stuck in some kind of time-warp with dSLR design, totally unwilling to do anything truly innovative, perhaps out of fear that their customers are too hide-bound to appreciate it.

I've never cared for live LCD composing and that's why I choose a DSLR. I did mess with my friend's Oly and it sucks as a SLR (tiny 'n dark) and the live preview was underwhelming. What I'd really to see is Canon bring back ECF. Well, it's still current on the the Elan 7NE and EOS 3, but not on any DSLR. ECF really sped up AF point selection for me. That joystick thingie is medieval!

I had mixed results w/ ECF - for some moving subjects, I want focus in one spot, but I want to look at another in order to get good focus AND good composition...

It seemed to work well, though, when I calibrated it properly. I had both an Elan-IIe and an EOS-3, so...

I suspect many who "don't care for it" would find themselves using it at least sometimes. . . if it were offered.

It's a no-brainer for certain shots, especially in the "swivel" version that is offered on some cameras.

I'm not trying to take away anyone's optical viewfinder, just to get the manufacturers to add a missing feature to dSLRs.

I wouldn't mind having the option, for macro shots and the like, and the occasional exotic angle portrait, but I think it would be a minority use feature for me. I would like to try it though.

I love my R1 and am sad to think that it's a dead end product.

It's worth it's weight in gold for street photography where the ability to shoot from the waist really keeps the process unobtrusive.... and quality is great... and RAW (albeit massive, uncompressed files - I just run the DNG plug in from Breeze Downloader Pro which I unload the files).

I love my R1 too, and also shoot only RAW + 1mpx jpeg normal and compress to DNG. Probably should switch to 4mpx jpeg normal to make my chimping for critical focus more productive. . .

Just amongst those of us who do see advantages in some situations of live video preview on an interchangeable lens camera, I predict that Olympus will soon announce several more cameras with this feature and 10MP or more: announced if not shipping by PMA 2007, March 8-11.

The sensor in at least new such model is likely to be the new "10.68MP" Kodak KAI-10100 interline CCD, announced so far only through its use in a special astro-photography camera from SBIG: http://www.astropix.com/wp/category/news/

That sensor is possibly the one in the Olympus E-400, but has a few features not exploited by the E-400, like electronic shuttering, probably video preview support (as video is the main reason that Kodak ever bothers with interline style CCD instead of full frame CCD), and 2-1, 4-1 and 16-pixel binning, for the first time binning has been offered on a color sensor with Bayer CFA as far as I know.

Electronic shuttering offers possibilities like flash sync at any shutter speed and the ability, when combined with a video viewfinder, to eliminate both mirror and shutter motion from the "shutter release" process. The latter is enticing, as it could offer negligible shutter lag, negligible shutter release noise, and negligible shutter release vibration: the true digital heir of range-finder cameras?

2-1 binning gives the option of output at half the pixel count (5.34MP?) in exchange for roughly twice the usable ISO speed, and perhaps about double the frame rate too. The improvement in low noise at high ISO is better than got by down-sampling to half the pixel count later, because it potentially halves the effect of the "per pixel" noise accumulated during read-out from the sensor.

The options of 4-1 and 16-1 binning are less interesting for taking photographs, giving only about 2.67MP and 0.67MP, but they could be useful for video output.

Yes, very useful. Whilst overseas recently, I picked up the new Sony DSC T30 ultra compact camera. This is a 7.2mp version of the original 5mp T1. I was pleased that the sensor size was slightly larger (than the T1) to accommodate the additional pixels. Viewing sample picture comparisons on the net (at imaging-resource for example) it was quite clear that the T30 had improved resolution as well as improved noise. Other features such as zoom facility during recording of video, image stabilisation with stills and video, much improved battery life using a bigger battery, impressive 3" EVF (who needs an optical viewfinder with that?), tripod socket that was left off the T1 etc etc, made it an irresistable buy.

However, low light video capability was pretty awful. There were many situations when I wanted to record cultural shows in the evening with less than ideal lighting. The results are way too dark. Perhaps this is because Sony haven't used binning to get the 640x480 video resolution.

Whatever the reason, I thought that next time I go on such a trip, I might take with me one of the new HD camcorders such as the Canon HV-10 that has a 3mp senor and records 1920x1080i. However, checking out the reviews and specs for this camcorder, I was surprised to find that the sensor is really tiny. In fact, the same size as in my first ultra compact P&S camera, the Sony DSC T1, which 1/2.7", or 5.3mm x4mm.

Not surprisngly, the Canon HV-10 does not get good ratings for low light capability. How can it with such a tiny sensor? I suspect this is the situation with all consumer grade, compact HD camcorders.

So after this long preamble, the one feature I want in a DSLR, that currently doesn't exist as far as I know, is High Definition video capability.

Consider what might be possible with a camera such as the 5D, which as you know has rather amazing low light capability at ISO 3200. 4 to 1 binning would give you a 3mp sensor with a huge dynamic range and superb low light capability at maybe ISO 6400.

Now that would be a very, very useful feature. Why is no manufacturer providing it? I know there's the argument that you can't expect good quality stills from a dedicated video camera for the obvious reason that the sensors don't have the resolution. But I see no reason why the reverse of this argument should be true. Admittedly, faster processors will be required. If the 5D can shoot 12.8mp images at 3 frames/sec, one would not expect it to shoot 3mp images at greater than 12 frames/sec, or 2mp images at greater than 18 frames/sec, which is not fast enough. However, a 5D upgrade with the 8/sec frame rate of a 1D Mk2 would appear to be fast enough. Perhaps the write speed of the flash card would be the bottle neck, and if it is, I'd settle for slightly less than full HD video resolution. 1024x768p would be nice .

Or, if 16:9 aspect ratio is preferred, the upgrade to the 5D, which will probably be a 16mp camera at a similar or lower price, could probably do 16 to 1 binning which would provide adequate resolution for un uncompressed, RAW mode, super DR, 1280x720p video at maybe even ISO 128,000.

The potential advantages seem so great, I can't understand why there's no movement in this direction, unless of course there's some conspiracy going on to protect the existing videocam market.

Ray,
that is a good question to end with, because I tend to believe that no-one is manufacturing such a camera even in high end professional video cameras for good practical and technical reasons, not due to the higher cost of 24x36mm sensors, or an industry-wide failure to appreciate the widely touted advantages of larger sensors and larger pixels.

What formats are judged by experts with very large budgets judge to be good choices for high end digital video? 2/3" and smaller. The current state-of-the-art seems to be Sony CineAlta 24P HDCAM cameras, $80,000 items using 2/3": 6.6x8.8mm in 4:3 mode, cropped to about 5x8.8mm for 16:9 format.

As to why, I can only speculate:
1) a camera's high shutter speed/low light performance is, as you should well now by now, limited by maximum aperture size, not directly by pixel or sensor size. Using larger sensors and pixels with the same aperture size if anything makes things worse: the same amount of light gathered and delivered to the sensor and so the same signal (in photons) is available at each photosite, but there will probably be a bit more read-noise.
2) HD video camera lenses can have very low aperture ratios. Many zoom lenses for 2/3" format HD cameras offer f/1.8 over a wide zoom range like the Fujinon HA22x7.8DERD-S 2/3" 22x High Definition Lens: f/1.8 at 7.8-122mm ("31-488mm equiv.") and reaching 172mm ("690mm equiv") at f/2.5. (Standard definition zoom lenses go even faster, like one offering f/1.4 at 9.5-253mm ("40-1000mm equiv.") Strangely primes seem not to get much faster, say f/1.5: maybe sensor/micro-lens characteristics limit minimum f-stop.
3) AS to dynamic range: at HD pixel counts like 1080x1920, pixel pitch on a 2/3" sensor is about 4.8 microns, and despite some wild exaggeration of DR needs in forums, this is large enough to give quite good dynamic range, way beyond what a straight print can display. For example, it is about the same pixel spacing as in the Kodak KAI-10100 4/3" interline CCD mentioned above and used in an astrophotography camera, and surely the astrophotography market cares about dynamic range.

So maybe speed limits are already set in 2/3" format largely by lens size and weight (or even desire for adequate DOF!), so that larger formats would by limited much of the time to similar aperture sizes, and thus not much opportunity for speed gain.

In this case, you can not argue that Sony is not capable of making sensors larger than 2/3" format.


Aside: Any suggestions as to why all high end video cameras, including those from Canon, use interline CCD's rather than CMOS sensors?

Ray,
the key word here is probably "understand". Might you entertain the idea that the HD video industry understands the technological issues better that you? (Likewise for the dominant industry-wide preference for formats smaller than 35mm, which you also seem to have trouble understanding.)

Perhaps this and many other dilemmas expressed in forum posts lie largely in the posters' inadequate knowledge and understanding, rather than the industry's collective incompetence. If so, the solution involves us putting more effort into seeking rational, technical/commercial reasons for why the digital imaging industry goes in the directions that it does.

P. S. I am still open to another kind of cynicism: that profit motives can sometimes lead to decisions that are not in the best interests of customers.

Ray,

to be fair, there is at least one digital motion picture camera using a format larger than 2/3": the Panavision Genesis used for commercial motion picture making. It is in traditional 35mm format, but in the movie world, that means 23.62mm x 13.28mm. The output is "1920x1080" with each of those roughly 2 million pixels getting its light from a 12µm square region, but each pixel comes from six photo-sites, arranged
RGB
RGB
with binning of output from each pair of photosites for the same color. So each photosite is about 4µm x 6µm, with binning to effective 4µmx12µm. (Why do it this way, always binning? Maybe making one long thin 4x12 photosite for each color is more difficult.) So, no Bayer interpolation.

This camera is so expensive that it is only ever rented, not sold: Au$5,000/day for you at http://www.panavision.com.au/RentalCatalogue.htm
This is perhaps another hint as to why your idea of a video camera with 24x36mm sensor has not yet been adopted. And clearly it is not just sensor cost that dictates the extremely high overall cost of this camera, or of those $80,000 Sony video cameras with three tiny 2/3" CCD's.

By the way, the lenses for this format are generally less bright (higher minimum f-stops) than for 2/3" format professional video, but brighter overall than for 35mm still camera format: zooms as bright as 14.5-50 constant T2.2 (so minimum f-stop a bit faster than f/2.2), 17.5-75mm T2.3, and 135-420mm T2.8. This again emphasizes the trend that larger formats tend to use higher f-stops, and so do not in general achieve shutter speed gains in proportion to their ISO speed gains.

Changing topic a bit here...

The Genesis has a 2 meg capture and a 3 meg EVF?

BJL,
I can see there's a DoF disadvantage here, using a 5D with, say, the Canon 50/1.4 for video. The lens on the Canon HV-10 camcorder, for example, is f1.8, just half a stop less but with significantly greater DoF. However, if the choice is between dark and grainy footage with good DoF and clear and bright, noise-free footage with shallow DoF, I prefer the latter, especially considering that what's at a distance from the focussed subject in low light conditions (with inadequate electric lighting for example) is usually so dark that great DoF is irrelevant.



I think I mentioned once, after getting my first P&S camera, the tiny Sony T1, I did a comparison of that camera and my D60, increasing ISO on the D60 to correspond with the smaller apertures required for same DoF, keeping shutter speeds the same on both cameras.

The D60, despite being an earlier model of camera with noise levels at high ISOs hardly being better than underexposed images at base ISO, still produced better and cleaner images than the T1.

This is one type of comparison that reviewers of cameras never seem to make. Consequently the consumers is misled into thinking their small sensor P&S camera with great DoF at small f numbers, is ideal for low light shooting.

It appears to be the case, judging by my own experiments as well as the occasional report on the internet from other individuals who have made similar experiments, that there's no advantage for the small P&S in this respect. In fact, the reverse appears to be true, but of course one can't be sure to what extent the reverse is true because reviewers simply don't make the comparisons. It might well be the case that there is some P&S camera out there that does a better job at ISO 160 and f3.5 than the 5D at f11 and ISO 1600.



There's no doubt that expensive broadcast quality video cameras (costing much, much more than a 5D) seem to have adequate dynamic range. It's the cheaper consumer grade camcorders that tend to suffer from blown highlights or grainy shadows. I'm not suggesting that a high end DSLR should take the place of all dedicated video cameras, but rather that the superior 'still image' capability of DSLRs could be downgraded (through binning) to provide very acceptable HD video, better in many respects than we are currently getting from camcorders such as the high definition Canon HV-10.

I wonder if that would be technically possible. Are the internal electronics of still cameras and video cameras the same ? In video, we are talking about refresh/storage cycles continually happening every 1/25th to 1/30th of a second. That will be moving (and storing) a HUGE volume of data, while the sensor is continually running and potentially heating up. If it is technically feasible, why has that not been done yet ? I believe most high-end consumer camcorders have sensors in the 1MP range. The 3CCD versions operate similarly to a FOVEON sensor I guess, just that there are 3 dedicated CCDs catering to the 3 color channels.

I think the R1 goes through some hoops to make the live-preview work and Sony makes it work by the CMOS sensor going into a low power mode (I guess CCDs consume a lot more power) to prevent the sensor heating up, or something along those lines. The Olympus sensor, which is slightly smaller than the R1 sensor, also uses something similar. But in either of these situations, we don't have video capability.

I love my R1 too and use it extensively in situations that demand "stealth", where we can silently be shooting images, with the targets being absolutely unaware that their likeness is being captured and stored. After they went into the dSLR business, I have zero hope that Sony will come out with an R2/R3 etc. I believe the "killing" of the R1 was a deliberate business decision to remove the cannibalistic competition from within their fold, for their dSLR.

The only downsides to the R1 was a shallow buffer (thus compromising its continuous shooting ability, which is exacerbated by its large RAW files), slow (but accurate) focusing in low-light conditions and probably the lack of an optical viewfinder. If you can work around the above limitations, it is truly a gem.....the only camera I carry around on a vacation !

Good point. That sounds like a major technological obstacle and could be the reason why we don't get DSLRs with high quality video capability. Overcoming such a heating problem might add considerably to the cost of the camera.

The wonders of creative pixel counting: with LCDs and EVFs, usually each signal of each color is counted, so that the Genesis sensor would be 2MPx3=6MP. Indeed, Panavision describes the Genesis sensor as 12MP! I had to read a while to find out about the 2-1 binning.

I can only repeat my main point from above: rather than thinking in terms of what you would prefer, without having had any experience using such equipment or in professional video photography, I invite you to consider what a great number of people far more knowledgeable than you or I have chosen, and try to understand why. By the way, have you actually seen a lot of "dark and grainy footage" from professional video cameras?


You are comparing products using very different technologies (Sony interline CCD vs Canon CMOS and in very different price ranges: too many variables to be able to attribute the results to photo-site size difference alone. Also, when pixel sizes get far smaller as with the 5MP 1/2.4" sensor of the Sony T1, they start to suffer from diminishing fill factor: a smaller fraction of the photo-site available for the electron well. Those 2MP 2/3" format video sensors have photo-sites about four times the area of the T1's sensor. What is more, newer fabrication technologies increase fill-factor at a given pixel size: with under 2 micron interline CCD photo-sites now viable, the 4.7 microns of a 2MP 2/3" sensor is probably easy for Sony to get a good fill factor with. Especially with the high budget allowed by the $80,000 price of those video-cameras
Yet another point is that these are three CCD cameras, with no color filter arrays on the sensors, improving sensitivity.

Why not try comparing two sensors using similar technology from a similar era, like those of the Canon 20D, 1DsMkII and 1DMkII (or the Canon 5D, though that is a somewhat newer design.) Another interesting sensor to look at in this context is the Kodak KAI-10100, an interline CCD with about the same pixel size as those 2MP 2/3" format Sony video sensors.

I'm not a fan of EVF for many reasons, even with the R1 which I owned previously. However, I'm surprised that Nikon/Canon haven't yet found a way for a hybrid viewfinder that could overlay a live histogram alongside the optical viewfinder image. Now THAT would be a killer application of EVF technology that would help with accurate digital metering!

You are straying off the point, BJL. Of course there might be good technical reasons why this feature I'd like is not available; reasons which I don't understand.

I'm merely making the reasonable observations that:-

1. 'Still' image quality from P&S cameras, whether from 1/2.7", 1/2.5" or 2/3" size sensors, is not as good as it is from the larger DSLR sensors.

2. Video quality from P&S 'still' cameras is not as good as it is from cheap, consumer grade video camcorders with similar size sensors to the smaller variety of P&S still cameras.

3. On the face of it, it would seem possible to use the higher picture quality status of the larger DSLR sensors to produce better quality video than is currently available from consumer grade videocams. This is a feature I would like and I wonder why it's not being offered.

So far, the most convincing reason that's been mentioned in this thread is a possible overheating resulting from a large sensor running continuously.

I thought that the point was to try to understand the clear fact that no digital motion picture camera uses a format nearly as large as the "36mm wide" that you propose. (Frame width seems the safest measure, as aspect ratio is often adjusted by a vertical crop). Indeed, even the largest current format, the roughly 24mm wide "movie 35mm" of the Panavision Genesis, has been ignored by the world of HD video and makers like Sony, Panasonic and Canon in favor of options like 3 CCD 2/3" format, about 9mm wide.

One important pont is that Panavisio has shown th technical feasibility of a format larget than 2/3", and yet it has been adopted on only averyt limited basis. THis alone should be enough to dismiss the myth about larger sensors generating too much heat in video mode. Anyway, it should be clear that there is plenty of room for active cooling in a unit the size of a professionbal video camera. A final piece of evidence: the Olympus E-330 has a live video mode using a sensor four times larger than 2/3" format without problems of overheating or excessive power consumption.

So once again I come down to one main explanation:
1. the advantages of the larger format largely relate to better low light performance and lower minimum DOF.
2. These rely on using lenses with both longer focal lengths and larger maximum aperture diameters.
3. The larger maximum apertures in particular come with disadvantages in cost, size and weight.
4. These disadvantages, probably size and weight more than cost, are judged by the industry to be too great to justify the advantages in most or all situations.

The issue here is one of balancing advantages in both directions, bearing in mind the law of diminishing returns. This is the common pattern that as you go further and further towards an extreme (like ever larger or smaller sensors and photo-sites), each subsequent step brings less increase in advantages than the previous step, and more increase in disadvantages, with a given level of technology. (technological progress can diminishes the disadvantages in each direction, so there is perhaps no universal answer as to which way it shifts the balance.)

This means for example that it is invalid to argue for the superiority of one option over others by pointing to the clear disadvantages of a format that is far smaller or far larger than any the options being considered. The hopefully obvious disadvantages of making a 1920x1080 video camera in "Hasselblad 48mm format", giving 25 micron wide photosites, are irrelevant to the comparison of format and pixel size options like 4.7 microns (2/3"), 12 microns (the 35mm format Genesis) and 18 microns (36mm wide used with 1920x1080). Equally irrelevant is evidence of the disadvantages of the far smaller 2.3 micron pixels of three year old sensor in the Sony T1. More so when you consider that 2/3" format HD video uses three CCDs, with no color filter arrays or Bayer interpolation, so the advantage in sensitivity over the T1 is greater.

P. S. The video quality advantages of a video camera over video from a still camera of the same sensor size seem easy to explain: the still camera sensor has far more, far smaller photo-sites, and then probably uses sub-sampling to get down to video pixel counts. That is, reading only some of the photo-sites. To get the higher video frame rates, either this sub-sampling or on-sensor binning is needed, and binning s not yet done in digicam interline CCDs as far as I know.

That means that the still cameras' video modes have the limitations of their smaller photo-sites compared to the video cameras, with the expected lower sensitivity and dynamic range. It could be a different story in the future with on-sensor binning, which Foveon offers in its X3 CMOS sensors, and Kodak is now offering in Bayer CFA interline CCDs. (Up till now, binning has been monochrome only, possible for example with three CCD video cameras.)

BJL,
With all due respect, that's not the point. I don't think I have anywhere suggested that an add-on video feature in a DSLR would make all professional video recorders redundant. It might well be true that there are very few professional HD videocams that have sensors larger than 2/3". So what? There are also very few 1-2mp 'still' cameras that have sensors larger than 2/3rds. What's the point of designing something that is unnecessarily cumbersome and expensive and which produces a quality that far exceeds the requirements of the end user??

I believe that until recently most professional video recorders have produced anamorphic 1440x1080i images, which are smaller than 1 megapixel. TVs and video projectors capable of displaying 2mp images are now gradually becoming available and more affordable. There's no mystery here as to why videocams do not use sensors larger than 2/3rds.

My experience with video capture is as follows. Standard definition camcorders that claim to offer 'near broadcast' quality are far from broadcast quality. The image stabilisation in such cameras often works on a 'pixel shifting' principle which unavoidably reduces even further what is already poor resolution. Performance in excellent lighting conditions is hardly acceptable and in less than ideal lighting conditions, woefully inadequate.

A similar situation exists with the add-on video feature in P&S still cameras, except it's generally worse.

So, the point for me is to try to understand the technical difficulties that have prevented manufacturers of DSLRs from offering the same add-on video feature that most P&S cameras offer, but with correspondingly higher quality. You cannot fit 2m 5D pixels on a 2/3rds sensor let alone a 1/2.7" sensor. Even without binning, there should be a significant advantage. Most consumer grade videocams do not use 3 CCDs per pixel.

The reasons that seem likely are: 1. Overheating, without additional technology to combat it, which would no doubt add to the weight, bulk and cost of the camera.

2. More processing power and speed, parallel processing etc. which again adds to the cost.

3. The consumer grade camcorder market would take a big hit if this feature were implemented in the way I've suggested.



I presume this is standard definition video. If the quality is better than the video from a consumer grade camcorder, then that would be one good reason to buy an E-330, wouldn't it.

BJL,
Perhaps once again such arguments can be reduced to lens quality. We've had many discussion before about aberration limitations in lenses. Are general purpose videocam lenses superior to Canon prime lenses, at f1.8 for example, or, if good DoF is required, at f8 (ie. videocam f1.8 compared with 35mm f8)?

We know that at low resolutions, 10 lp/mm for example, that good 35mm lenses are up in the MTF 90%+ range. Is much improvement possible here?

Consider the flexibility of CMOS sensor design with on-board processing. If all we're interested in is a 2mp image from a 12.8mp 24x36mm sensor, then (potentially) we have a choice of which pixels are going to be accessed. If we start from the outer edges of the sensor, then we have the full focal length of all our 35mm lenses.

If we grab 2m pixels from the centre of the sensor, we have the advantage of better resolution and an effectively greater focal length. According to my calculations (and I know maths is your strong point, so if you correct me I shall not be insulted ), 2mp in the centre of a 5D sensor occupies an area of 9.4x14mm. My 400mm lens becomes a 1000mm lens in video mode.

But I have a choice, if the camera design offers a choice, of which pixels are going to be active when I shoot video. My 15mm Sigma can be either 15mm or 37.5mm, depending on which button I press.

These possibilities seem under-exploited, in my view, and I have a strong hunch the obstacles are not entirely technical.

I think I've misunderstood you here, BJL. I've just searched the net for comments on the video capability of the Olympus E-330... and of course it doesn't have one.

Your point was, I guess, that in order to get a live preview using the EVF that comes with the E-330, the sensor has to be continuously active, and that that doesn't appear to cause overheating... which sort of gets us back to square one (although a 5D sensor is 4x the area and heating might well be a problem).

All I can say is, if the future 16mp upgrade to the 5D contains a sensor that can access even 1mp of image information through a variety of spacing options, in order to deliver 1280x720p video footage at ISOs up to 3200, then it'll be on my shopping list. I'll consider it a very desirable item, although 2mp would be better.

A quick calculation of the zoom range with any prime lens attached to such a camera gives me a figure of around 4.5x with a 1280x720 16:9 aspect ratio.

My 50/1.4 prime effectively becomes a 50-225mm zoom, but perhaps not continuously adjustable. Anyone fortunate enough to own the Canon 600/4 gets a 600-2700/f4 zoom.

Someone point me in the direction of the patents' office .

I was of course referring to the E-330's video output from the sensor to the LCD (or to an external monitor. Surely that continuous operation generates as much heat (or as little) as if the sensor output was being recorded. (Aside: I imagine one could record the video output used for an external monitor if one wished, but that output is probably sub-sampled, so not of particularly high quality for recording purposes.)
And as far as sensor heat as an explanation for the dominance of formats 2/3" and smaller, you would need to explain first why not even formats like 4/3" and movie 35mm are not used in HD video, given evidence that those sensor formats do not have a significant heating problem.

I suggest we apply Occam's razor, in the form that common phenomena are most likely explained by common causes. The common phenomenon I refer to is an overall shift to smaller formats, with both still and moving pictures, and with both electronic sensors compared to film (or analogue vidicon tubes), and also with film over the decades.
Sensor heating cannot explain all of the downsizings, and nor can price difference dues to sensor costs (including film and processing), since that price gap went down over the years with film, not up.

The only likely common cause I can see is the opportunity to use smaller, lighter and likely less expensive lenses as resolution and sensitivity of the light sensitive media improves.



Indeed, maybe DSLR's will add video recording modes, if and when live video preview becomes more popular, and color binning becomes available.

BJL,
If I apply Occam's razor, the answer I get is, 'you don't need a large sensor to produce a 2mp image which ends up being highly compressed'. If the current Canon G7 were a 2mp camera instead of a 10mp camera, and in 16:9 format; if it could produce 1920x1080p video at 30fps instead of its actual 1024x768p at 15 fps, I think the buyer would drool over such video capability.

It doesn't seem an impossibly big step to go from 1024x768 at 15fps to 1920x1080 at 30fps; about 3x the processing power/speed, wouldn't you say? Of course, maximum print size from a still image at 240ppi, without interpolation, would then be just 4.5"x8". I don't think anyone would produce such a camera because of this limited print size and whilst the video quality would be outstanding, dedicated videocams would be more flexible and easier to use. The G7 doesn't allow optical zooming in video mode, for example.

Since you've discounted 'heat build-up' as a major technical obstacle to video from large sensors, then applying the Occam's razor principle to the question, 'Why isn't a high quality video mode made available on DSLRs, which could take advantage of low light situations?', I arrive at the answer: because it would cannibalise the profits of consumer grade videocams already being offered by the same manufacturer.

I see another explanation as far more likely, partly because the need to compete with other companies usually precludes the idea of deliberately withholding an attractive feature.

The standard SLR approach relies on having the mirror down for (phase-detection) auto-focus and light metering, and then raising it to make the exposure. This is not compatible with video operation, unless you only focus and meter before the start of a "take". To support video recording from the main sensor of an SLR, changes are needed, and the options I have heard of are
1) Change to doing AF and metering with the sensor, requiring a change to contrast detection AF, which is generally considered inferior to SLR style phase detection AF. This might be appropriate for entry level models aimed at those stepping up from fixed lens compact digicams but not for the more discerning "5D" crowd.
2) Use a partially reflecting mirror ("pellicle mirror"), reflecting enough light up to the AF and exposure meters and maybe to an optical VF, while letting a good proportion of light to go to the sensor. (Canon used a fixed pellicle mirror system in one high frame rate EOS-1 film camera, but dropped it in the more recent model.) For still photography, the camera could have the traditional option of swinging the mirror up, if the benefit of letting all the light go to the sensor is sometimes significant.

Well, I agree that design changes would have to be made and that would probably increase the retail price, but probably not by nearly as much as the additional cost of a separate, high quality videocam. The option I would prefer is mirror lock-up for video mode, rather than a partially reflecting mirror which might reduce sensitivity or increase noise at high ISOs, and a secondary AF and metering system. The review LCD on the back of the camera could act as the live preview and EVF in video mode.