Crystal Ball?

[Arne Johnson]

This is actually more about making news than reporting it, hope this is the right section...I'm writing an article about the next year in DV and HDV production, that's a sort of looking into the crystal ball kind of thing. What do y'all see coming down the pike, and what do you wish/hope for in 2006? Obviously the HVX is a big deal, but what else? Software? Hardware? Come up with something good and I'll quote ya in the piece, which will run in a northern california publication I edit called Film/Tape World...

Thanks!

Arne

[Wayne Morellini]

Arne, probably not the section, but. Hit all the major manufacturers about unreleased cameras. There was a guy that went to a NAB a last year, that posted stuff about previous canceled plans. But the thing is that Sharp had plans, probably every manufacturer (Samsung makes interesting Mpixel sensors, and I wonder when they are out with something). Every major manufacturer should have plans for at least one model, if they do pro-sumer equipment then another model, if they do pro equipment then maybe at least a third model in the consumer/pro-sumer space, at this early stage. If you see a major manufacturer with less than this number of models, then it might be worth asking (but if they don't want to talk they might get annoyed if you ask again).

http://www.geocities.com/mammacow3/

JVC, has been rumoured to be releasing a cheap camera for a while, and we have seen reports of an hard drive Everio (posted here a in the last month or so). Canon showed off their xm2 HD replacement.

On the technical side you could have a look at the following. There is h264 chips being developed and reports (here recently) of manufacturers that will use that on their cameras (but doesn't mean they will push it through to market). I don't know where ever digital still cameras and or video cameras. But a proper h264 at even Mini-dv's 25Mb/s bandwidth, would be a great improvement over HDV standard (that not everybody is signed up to). An camera based on the better open-source alternative, Ogg Theora Videoencoder (of which there are people like Elphel in Russia doing hardware FPGA designs for an previous version of it) would be even better. The issue with anything low bandwidth (25Mbit/s and less, Digital still cameras get much less) in HD, is that there are situations where there is too much change on screen (motion usually) to fit into the bandwidth available, so the video breaks up or the pixels become bigger (eventually turning into clearly visible square blocks). HDV suffers from this, if low enough bandwidth is used h264 would also suffer from this. Video on digital still cameras/Mobile phones suffer from these problems (I assume a lot more because of low bandwidth). Another problem with digital/mobile cameras (and a number of network cameras) is that the video resolution they are giving is below mini-dv (640*480/320*240/160*120 etc) and often they are using a slower frame rate (15fps etc). They often use very lousy cmos sensors (on mobiles/network) compared to Mini-dv cameras, but the latest sensors, like from Micron, are a great improvement. So a Mini-DV camera is quiet often better.

You could go to the major sensor manufacture players and see if you can get access to projects using their sensors. Rockwell Altasens, is highly lauded as one of the best HD video performers, people here are trying to make cameras based on it, but it is in short supply. But what is not commonly recognised is that they also have a smaller cheaper version (which I hope heads to consumer equipment). There are a couple of major sensor manufacturers for mobiles (but I don't remember them, though the article on the latest h264 codec chip mentions at least one of them) and the Micron I mentioned. But I suggest enquiring with the following manufacturers that have a number of advanced technologies, to see what people are doing with them in the consumer space: Altasens, as mentioned, nearly industry leading low noise and low light pickup) Fill-factory (have global shutter, dual slope, one of the highest well capacities for latitude, and full fill factor, Smalcamera (an auto gain feature for latitude, they do a sensor for Kodak and one for one of the top Cinema camera manufacturers) (Fill-factory and Small have both been bought by Cypress semiconductor). Micron (cheap quality compared to many cmos, and the new Mobile chip looks nice). Foveon, the originator and sole manufacturer of X3 sensor technology (this is a cheaper single chip alternative to quality three chip cameras) it is potentially the wave of the future. Dalsa, is supposed to be upto something, they do a lot of expensive stuff, so I don't expect it to be in the consumer electronics oriented device.

http://www.elphel.com/articles/index.html

Probably good for another side box, is the many DIY camera projects we are trying here (as good information on the issues of quality, and also as a warning of trying to DIY) and other manufactures trying to offer cheap alternative cameras (but mostly for the indie film market at the moment) good for a separate article on it's own. In network cameras, the Elphel camera is worth a look at. My technical thread follows, it links to one of the projects that had the biggest thread on the web ever:

http://www.dvinfo.net/conf/showthread.php?t=28781

Please, give a plug for the benefits of full 1080HD 50MB/s+ 10bit+ 4:2:2/4:4:4 recording for the pro-sumer (under $5K) like less motion artifacts, better color, better resolution, more to play with in the right hands.

Thanks

Wayne.

[Arne Johnson]

Wayne, thank you so much! I don't actually have time to pursue all of those leads you give, as I'm up against deadline today, but I'll definitely use some of the thoughts you've put down.

Also, you read my mind exactly about the DIY article...we've already got one in the works, and I'll be trying to get interviews with folks soon.


Arne

[Wayne Morellini]

Just read the rest of your post and now understand you are probably writing to what is probably a very savvy audience, so I have a bit more.

The threads on the cinema projects have a number of good links, and posts, for learning about the technical issues of cameras image quality. But the state of sensor technology is telling, and that they are hitting real limits that should effect future purchasing decisions:

The efficiency of a sensor is measured in QE, Quantum Efficiency. This is a general measurement, of how much electric charge is produced by the light hitting the sensor. It works by the percentage of electrons moved by the photons of light hitting the sensor. In the past 35% would be considered high, the Altasens CMOS outstripped CCD technology by raising this to 70%. Recently there has been CCD technology doing upto 90%+. The maximum is supposed to be 100%, but I have read of recent scientific research that has been able to move two electrons with open Photons, so the maximum is now likely to be at least 200% efficiency. Also it is not at maximum efficiency across the range etc. So there is still room for significant improvement in sensor efficiency. But on present day technology 70% is way high, and so you might not expect much improvement for quite a while. So a sensor with this efficiency should do you good for quiet a while. Also, once sensors based on all the technologies I mentioned in the post above (from cross licensing etc) become available, there maybe slow improvement from then on, making them good buys. This point might hit the pro-sumer space within 2 years.

Latitude rules (the number of stops) with good Signal to noise ratio (level of noise in picture) to take advantage of extreme latitude. When you get too much latitude compared to the human eye, the picture will get murky (not much a problem with generally low latitude sensors) but this higher latitude is advantageous for professional film making and for auto systems on consumer equipment, as it allows you to select a range of latitude (the upper and lower clip level) of choice during product or in post, with less dependence on the iris. But to do this requires not only extra bits, but low enough noise to keep bit accuracy (a bit technical). If you had 10 bits, but the range you want is the middle 6 bits then you are, in effect, working with 6 bit video, but 20 bits would allow you to use/record the middle 12 bits, for 12 bit video etc. I don't know how much improvement there can be in latitude, or even noise, but this is one area of improvement and filming change in the future. You see, with the recent article on the DOF camera (now moved to Alternative imaging forum, where you will find a number of the camera projects) that allows you to insert Depth of field in post production, that a major change in the way of filming is possible. Eventually you can set up the camera much quicker (less lighting problems) with modest operation to record everything, and in post to set the DOF, the colour, and the latitude range (lighting). For an indie production this is an stupendous leap in quality and operation. I have been thinking of these sorts of issues for many years, and this is only the tip of the iceberg in what can be done to streamline and lower the cost for production.

The other thing that gets people tied in knots about, is the sensor size. Sensor size is not everything, it directly effects the depth of field with rising sensor size, so over 2/3rd inch is often considered enough (with ultrafast lens on a single 2/3rd inch chip), but apart from that it is sensor pad size that matters. Different techniques have different sizes of sensor pads, the bigger the pad on the same technology (as technology also effects performance) the more charge it may store from the light hitting it and the lower the noise is compared to that charge (better for bit accuracy and low light accuracy). Three chip generally present a bigger area than one single chip of the same size. Different technologies have different ratio of pad to space around the pad used by support electronics etc (too much give an invisible fly screen problem) the fill-factor. To get over this they have tired to shrink the wasted space with different schemes and smaller chip processes, but generally the method used in to put a micro-lens array on the sensor to divert the light away to the pad away from the dead zone. But this is also a problem, the Micro-lens array may start developing aberrations around 1.7 (or was that 1.3f) stopping the 2/3rd inch chip from using super-fast primes. Three chip prisms also effect how wide the aperture can go (is that the 1.7). So a normal three chip, or a good one chip that depends on Micro-lensing has there limits. But back to sensor size. As the number of pixels on the sensor increases the pads get smaller and these problems also increase. So, this is a major problem for cheap cameras with small chips and high pixel counts.

You might like trying to look up the Drake camera thread on the alternative imaging forum, the camera team has split into two parties at the moment, but there was a German feature done with the prototype, called Dragon Feather (they hadn't used the sensors range extending capabilities and 12 bit mode either).

Please forgive my sloppy writing it is nearly 4 am.

[Wayne Morellini]

Originally Posted by Arne Johnson

Wayne, thank you so much! I don't actually have time to pursue all of those leads you give, as I'm up against deadline today, but I'll definitely use some of the thoughts you've put down.

Also, you read my mind exactly about the DIY article...we've already got one in the works, and I'll be trying to get interviews with folks soon.


Arne

You really needed to ask a few weeks ago ;) Also sorry for missing your reply when I posted.

Thanks for taking an interest in the DIY projects. How does one get your publication?