Home Made HD Cinema Cameras - Technical Discussion

Wayne Morellini · May 26th, 2006, 09:43 AM

The really too radical
Patching video cameras to make a larger image:

What do you get if you stick two 720*576p camcorders together, something better than DVCPROHD 720p.
What about if you stick three together, something much better then DVCPROHD 720p, including extra wide 720p.

This has already been done, as reported on some thread around here, I hear. But ideally you would need a prism to patch the images together (like the Olympus 8MP SHD Eng camera did with three HD chips) and to synchronise them.

An alternative, would be to use a Projection screen, like they use on the 35mm Lens adaptor projects and align the cameras to the image and record them from there in synchronisation.

Using full frame still cameras in sequence:

You may have read my Red camera price joke, where the true maximum price of a 11 Mpixel camera can be determined by aligning a number of cheap cameras together till they reached the desired resolution ;).

Another solution, I have recently come up with, but something that will probably be impractical due to the problem of angle of photography, is to put an square/rectangular array of synchronised still cameras to take pictures interleaved at the maximum rate (even circular would work). This is different as each camera is taking a full frame, but because these frame rates are like 1-7 fps, depending on model, the next cameras have to take a picture while the previous one is saving the picture. But the problem is that you will get an movement, the closer an object is to the camera the worse this shift will be. To get around this problems you can use a projection screen, but the angle of the cameras will make the screen look warped as the angle of perspective changes, as it is unlikely you can put it far enough away. So again, prism is the best solution, but the prism would have to have many output surfaces, and be complex. There is a way to make it with a series of first surface partial mirrors (these have mirror surfaces on the outside of the mirror, the partially mirrored surface transmits some of the light through instead, to be reflected by the next mirror). But each mirror has to be a different size, and different partial transmission, to capture the same amount of exposure, with different focusing distances (if you do it that way) and many of them, so it to is complex, but an interesting experiment.

Things undone: I'm not bothering ;)

Sensors, where does it end
Kodak have plenty of nice sensors, I went through just about all of them, virtually three pages to get to the basic specification of each :( (no wonder I didn't do it before). Some have at least 100K -ev well capacity. Pixim has some extended latitude stuff, and their site has development kits, and lists of box, dome and network cameras using their sensors. Micron is getting into extended latitude, on their Super Wide VGA sensor at the moment. Cypress has Ibis with extended latitude and global shutter, non cheap well capacity and descent SN, and their Smalcamera sensor company with cheap sensors with extended latitude. Extended latitude seems to be a feature on the rise across companies, which is good. Foveon has 3 colour per pixel, and I have found another, that I am trying to communicate with. Altasens has good Quantum Efficiency, SN, and I guess latitude, but what is the fill factor?. Fuji is doing a 3 color per pixel chip for next year or so, based on an more organic basis, apparently. Too, many more to go through, but this is the important ones so far.

Undone work: There are probably more than 20 other sensor companies to check all their sensors, lists in thread above :(

What resolution for shooting 720p and 1080p on the same camera:
There is a trend to producing new sensors with ever increasing resolutions, and lower resolution sensors getting left behind. So the problem is that there might be limited low cost options at 720p and eventually 1080p, so a higher resolution sensor might have to be used. What ever resolution that you want to shoot in bayer, you should have an sensor that has, or a bit more, than a multiple of the resolution you want to shoot, because proper binning will give you accurate pixel for pixel matches. For instance, 720p is 1280 pixels across, so you could use A 1280, or a 2560 or 3840 etc pixel width. For 1080p, it is 1920, 3840 etc. You see there is a match up at 1280*3, and 1920*2 and 5*768 (not a standard resolution, but one I have seen used in the past).

GPU encode enhance
I have read through the Nvidia Pure video Marketing, and it does claim GPU encode assist (not complete but only assist).

Alternative commercial cameras:
Panasonic and Sony has announced AVCHD h264 18Mb/s, for DVD, Blu-ray and SD recorders (I hope HDD also). This should be good, and hopefully fix up the problems of HDV with motion artifacting most of the time, and help with low light noise performance degradation . It does 1920*1080p. H264 is 2-3 times better than Mepg2, I don't imagine that this transfers to the problems, but the situation should be better a lot more often, and other images a bit nicer (if used on the same quality sensor etc). So prosumer version of this will hopefully be everything that HDV should have been. Going on interviews and reports (mentioned in various threads on the subject) Sony is expecting to release cameras, and Panasonic maybe releasing a SD card version late this year, or into next year.

http://www.dvinfo.net/conf/showthread.php?t=67127

Samsung announced a h264 around 18Mb/s, like AVCHD, last year, due August, uncompressed HDMI (a digital component replacement) output.

http://www.dvinfo.net/conf/showthread.php?t=58261

Ambarella has announced that a number of manufacturers will be suing it's revolutionary h264 chip for cameras starting at $799.

Still cameras will get HD video one day, Sanyo HD1 has already started with Mpeg4 9Mb/s.

Panasonic, at NAB, announced a new High Profile H264 10bit 4:2:2 50mb/s intra (it does not do inter frame compression, used for movement) codec. Cameras will be at Nab next year, maybe even a HVX200 version we hope. Even through people think h264 is purely a delivery format, the high profile was brought along to make it into a robust editing and lossless format. The profile goes to at least 12bit 4:4:4 at many hundreds of megabits. Altogether this Panasonic version will, hopefully, be a descent consistent quality professional codec, a grade step down from cineform I imagine, and two from lossless, but a step up from AVCHD. The present HVX200 with the DVCPROHD 100Mb/s goes some way to that already, but still 8 bit.

JVC showed a Hard drive Everio HD camera last year, even announcing at a dealer conference, that it was to be released around the 05/06 turnover (this is reported by member on dvinfo who was there). This camera, and a few others, have been delayed, the question was why? This new H264 stuff hopefully indicates that there might be some rethink, and hopefully we might get h264 (though JVC is not a member of AVCHD, nor is Samsung who is doing a h264 camera). If so, there are more options.

http://www.dvinfo.net/conf/showthread.php?t=62117
http://www.dvinfo.net/conf/showthread.php?t=52458

So things are looking Rosy.

There is one more possibility to come, but I am negotiating at the moment.

I am not quiet finished there will probably be at least one more post.

Thanks

Wayne

Keith Wakeham · May 27th, 2006, 08:04 PM

I think this is a first here, a real summary of sorts.

All in all some good stuff, and unfortunately makes me feel a little bad that much less has been accomplished than I would have wished for this group. Most of the people who started with the crazy ideas have moved onward and away it seems. The most successful I think being Reel-Stream. I hope that I'll be able to migrate my site off my university server, and setup some forums like they did in the comming months and firmly establish something.

Over a year ago I started work on my project which originally was to use variable timing with a KAI-2093 CCD and transport bayer data over 1/2 a HD-SDI link. Then decode it back to data, and write the data to hard drives. The first of this came as developing hard drive control and after months of pouring over info started to make some progress. But once I went to finish off the camera side I found that money was a very limiting factor. Then I went for a more direct approad and bypass the HD-SDI with some outside help, but eventually decided back on going for a pure recorder that would be general purpose. That was 4 or 5 months ago. Lots of programming latter, multiple circuit board designs, all sorts of problems and just a little further ahead. Everything I do seems to hit a huge delay and I appoligize to everyone for that, I don't think i've realistically met a single deadline in ages - I guess i'm always way to optomistic.

Honestly I think that people out their in the big name companies know about this place. Heck, they might even know my name in some circles of these companies as the crazy engineering student working on something completely doable without big backing, and honestly they probably count on this as the reason its not done yet and hope that I never finish. I recall people at panasonic commenting about Reel-Stream negatively. So I have no doubt that when I am finished and show what i've spent a year of my life developing that I too will recieve negative comments. Nothing is perfect.

However, to all those who speak negatively about this I have to say one thing. Do you think nobody from the Red development team has ever been in this forum. I'd bet that Jim Jannard himself has been in here at some point and got some little ideas. It places like this that recognize what can be done now, what we want done, and the methods to get their -- and most of us have spent a lot of money along the way.

Since I've put up my site the biggest question I've recieved is when can people get one, but what its now turned into is people stating they have a project in a month and asking if it will be ready then. Some have even gone to stating that if its not ready by the time they begin shooting then it won't be any use to them. I'm hoping this is a reaction to the "shooting season" being upon us, but I tried to put my email out their to help people and discuss ideas and try and find the people the first units will most benefit.

Anyway, for all the crazy Ideas Wayne keeps coming up with and all the reasearh and work, I got to say he's a really good fellow. He doesn't lose faith in what people are working on, and I have to say a big thanks for not losing faith in me. I have a feeling a lot of people have and although I'm not sharing my FPGA code, I feel I have let the community down a little everyday I don't have a device or product to show or some more really proof of what i'm doing.

I'll never quit trying to improve things and ideas, and hope that someday they will make a difference.

So a big thanks to Wayne for showing us what the years of forums has given us so far even if it seems small, you never know who is listening.

John Benton · June 14th, 2006, 11:54 PM

Keith,
I just shot you an email, before reading this post, asking how it was going.
Thanks for the update and Kudos...Just keep going... you will conquer it.

Steve Nordhauser · June 15th, 2006, 03:33 AM

Keith,
There wouldn't be a Silicon Imaging SI-HD-DVR (new name) if it wasn't for this forum. Two years ago we were an industrial camera manufacturer being pestered by 'some other bunch of people who wanted to use our camera in a weird way'. I started reading this group and met Jason, Davids at Cineform, Kyle, Obin and the rest of the group. When the group software didn't get off the ground, SI became convinced that the need was real and wrote our own.

Cineform completed the picture when we found that editing uncompressed wasn't easy for a budget minded Indy. So, we partnered with Cineform. It wasn't a huge leap to doing a complete camera so again, driven by the needs of the people we met here, we got a working camera to NAB, hit the press, hit the streets and should be shipping our heads in about 4 weeks and the full camera in the 3rd quarter.

So, aside from any other discussion, that is what was accomplished - a full 1080p camera with a radical workflow for $20K.

Thanks gang,
Steve

Quote:

Originally Posted by Keith Wakeham

I think this is a first here, a real summary of sorts.

All in all some good stuff, and unfortunately makes me feel a little bad that much less has been accomplished than I would have wished for this group. Most of the people who started with the crazy ideas have moved onward and away it seems. The most successful I think being Reel-Stream. I hope that I'll be able to migrate my site off my university server, and setup some forums like they did in the comming months and firmly establish something.
.........
So a big thanks to Wayne for showing us what the years of forums has given us so far even if it seems small, you never know who is listening.

Matteo Pozzi · June 15th, 2006, 06:54 AM

Hi to all
did someone can tell me what camera (gige, firewire usb2)
use the altasens cmos
I know only the Silicon imaging one!
many thanks
Matteo Pozzi

Wayne Morellini · June 16th, 2006, 04:54 AM

Yet, nobody ever started out these projects to make $20K cameras that only a few hundred would get in a year. But virtually all were aiming for sub $10K, preferably sub $5K, that they could build themselves, or would be made in 10K+ quantities. We are miles away from where we were going, led astray by some, and it looks like Pana is going to supply the minimum quality, maybe at double the price at least of the top quality.

What was not doable in 1080p editing last year is doable now or latter, and 720p could have been done sooner instead, Cinelerra has been doing for years in true 4:4:4. The PC/processing market is about to go through a major advancement in processing power over the recent advancements, and has just gone through a major advancement in drive speed. I have no problem for the numbers in 720p, even in 1080p with a bit of effort, but the advancement will make 1080p easier than before. 4K is the new target for the industry, that is where things are difficult despite all recent advancements, but something that can be done in future.

A few hundred people will benefit from direct cameras, so it is good for them, the rest of us will have to pay for the H264/XDCAM HD stuff. Thousands can benefit from the HDSDI work. Hopefully the post $10K cameras will inspire somebody to do sub $10K cameras, where you can buy double the quality compared to the regular video camera manufacturers, and force them to drop their prices.

As far as focus goes, it all comes down to who do you care about, ourselves, or everybody else. Caring for everybody else requires much more effort and less easy money, but in the long run better for everybody.

Wayne Morellini · July 22nd, 2006, 08:48 AM

Quote:

Originally Posted by Wayne Morellini

I got to have a feel of the new Playstation portable, the demo unit just came in the local store. Wow, haven't really felt like that since I bought my first Atari Lynx. Pretty awesome machine. The LCD is pretty good, but refresh a little slow.

Since my posts about PSP Camera project history and suggestion were removed, I'll repeat some of the information here.

I was planning on doing a RAW, Lossless or Visually Lossless PSP Camera. But as I couldn't get the right team together, was terribly sick, and Sony were a little reluctant, I had to give it up. Now I have other projects to concentrate on. So I decided to reveal the project as a suggestion somebody else may like to pickup.

The PSP is a mighty powerful little machine (If programmed properly, preferably most inner processing loops in proper machine code, and anything else that may cause pauses in execution, so only for good programmers). It has dual 300Mhz+ processor, I think one might be integrated with full or part DSP circuit (can't remember the details). Perfect for SD or even 720p footage. Put it this way, more powerful than the Playstation 2.

Now the problems: Interfaces are limited. One USB Slave (where their camera and extras go into), at least one card port, and one custom interface port. So connecting a camera and a Hard drive at the same time would be a challenge. I doubt that the USB can be made to connect to a hub splitter with data being read in and sent out to the drive (but I don't know). This will probably would mean making a custom interface circuit on the card or the interface port. If the camera head recorded directly to drive instead, PSP could be made to be viewer and control. They have temporarily locked the processors back closer to 200Mhz to preserve batter life. It is a surprisingly heavy thing. The only independent home-brew software works on the limited number of early Japanese units with version 1 of the firmware, all other versions don't work yet. I understand they are planning to force users to update the firmware (live update) before they can run certain games. This way they maybe able to keep independent development out.

There is psp development information at the PSP forum on:
http://ps2dev.org/

This links to the Home-brew PC development system for the PS2, I am not aware if there is a PSP version yet.

These details maybe inaccurate, as it has been a longtime since I read the data. Verify if you need to.

------------------------------------------------------------------

Some Sums:

1280*720 24p=22.1184MB/s
or 2:1 compressed=11.0592MB/s
or 4:1 Visually lossless (for instance Cineform Bayer) compressed=5.5296MB/s
(this is pushing processing a bit (for instance Cineform is very intensive) but I don't know what it is like with new DSP functionality)

Interfaces:
USB 2.0=50+MB/s
So feasible for HDD and Camera, only if USB splittable for dual use.

Custom interface:
I don't know, assume a lot.

Memory card (is there a compatible memory card to IDE interface):
I forget, I think some version of this card go upto 16MB/s.
So potentially feasible for compressed data.

Lets looks at Foveon X3, or three chip:
640*480 24p (you can probably go upto near 720*360 24p wide cinema format). If a three chip was used with 1/3rd pixel shift, you could get a theoretical 1920*1440 resolution after filtering upscaling. =22.1184MB/s
or 2:1 compressed=11.0592MB/s
or 6:1 Visually lossless Cineform compressed 3 chip=1.8432MB/s
(this is not recommended, processing, and the visually lossless may not be designed for this).

New, single, 1.8th inch drives might be able to handle 2:1 compression. The largest is supposed to have 90GB, which is plenty, but I don't know about fastest sustainable write rate.

It is possible that much greater than 2:1 lossless can be achieved, there are new codecs claiming much. I suggest nuking the sensor noise in the image, as it is not part of the original image anyway, but electrical errors.

So, interesting isn't it, of course none of us have time or money to work on it, or to go to PSP group and ask them.

Not supposed to be posting here, but I thought I would update this. Totally possible, Sony has decided to introduce a very low res camera for the PSP. People wonder why, maybe they did read the above. Back to all seriousness. It is totally possible to do the above, without anything too elaborate. A usb camera should go into the port (even if you need a plug adaptor and separate power supply) and you have SD to IDE interfaces for storage.

But have a look at this (to keep in mind for the future):

http://www.tomshardware.com/hardnews...11_162423.html
http://www.research.ibm.com/journal/.../vettiger.html
http://blogs.zdnet.com/BTL/index.php?p=1143&tag=nl.e539

The PSP is getting old and passe, many other portable devices could do the same thing.

Recently Ultra Mobile PC's (UMPC) have been turning up with inbuilt mega-pixel cameras. One with a suitable camera and programming might turn up, all that would be needed is a lens adaptor with condenser set, and wired lens (as we have seen in the thread all doable/available).

-------------------------

How has it being going on my front?

Well I am getting a few useful hours a day of limited health a lot of the time, if I take it easy, and even more if I don't do much. I even feel happy sometimes (though with all the health product I am pumping through myself I should feel like Clark Kent). Things have been slowly improving for months, at the moment very slowly. Hopes are for improvement to full days of full health long term, not just a few days or weeks, a few times a year, before I relapse.

What has been happening on the project side:

I have been trying to negotiate an OEM software developers license for camera hardware, where I can develop quality camera capture/control 3rd party applications for various platforms, that manufacturers can pack in, or end users buy. But negotiations are going slow, so I don't know how it will go. I for see a market out there that limitedly competes with the machine vision sector, but also for other non-video professionals requiring better image quality, as well as video professionals. Still a long way to go.

Obin Olson · July 22nd, 2006, 08:56 AM

out. no worries!

Wayne Morellini · July 23rd, 2006, 08:05 AM

I liked that Furp animation of yours, but can't find it anymore. But what does "out." mean?

You know, I would love to see that on TV, I could point to the screen, and say, "I know the guy that was responsible for that". Ren and STi, and Crazy Frog can stand aside.

What is the latest on that line?

Frank Hool · August 28th, 2006, 04:05 PM

Found just some settlement reading about image sensors overall and about my present investigation subject what sensors for what CMOS vs CCD.
All from Dalsa:
http://www.dalsa.com/dc/documents/Im...8-00_03-70.pdf
http://www.dalsa.com/markets/ccd_vs_cmos.asp

Wayne Morellini · August 30th, 2006, 07:07 AM

Thanks Frank.

Here is a new HD video Webcam.

http://www.dvinfo.net/conf/showthrea...337#post535337

Wayne Morellini · September 12th, 2006, 05:35 AM

For those Digital Cinema camera people still subscribed to this thread, I have some major news.

Declink has released an uncompressed HDMI capture card fro around $250US. Apparently you can select a codec to compress the footage during compression. It includes a free photojpeg codec, but something like Cineform o a free open sourced codec might be preferable. People are also talking about using a a component to HDMI converter fro older cameras without HDMI.

http://www.dvinfo.net/conf/showthread.php?t=75102

I would liked to have started a thread on this in alternative imaging for the other Digital Cinema camera people not subscribed here, but somebody beat me to a thread in news. But people are spread across the SI and Red forums now, that would have been interested.

A word of caution, I would be wanting confirmation on which cameras output pre-compressed/uncompressed HDMI. I would be wanting confirmation about the format that it records in and wherever it can precisely lock onto the exact signal format the camera actually uses.

Wayne Morellini · September 13th, 2006, 11:52 PM

Have contacted them. It looks like HDMI cameras are currently 8 bit 4:2:2 at the most. So, this would, in practicality, be a step down to what we would expect to achieve with machine vision cameras and the Elphel project (which currently is goign towards hard disk interface and bayer in the other thread) . But as a solution a step above what we are likely to achieve from a webcam without 10 bit uncompressed.

Notebook version is yet to be planned.

I'm impressed, definitely a budget uncompressed solution. But a notebook version for similar price, or budget field recorder, would make life a lot easier.

Wayne Morellini · November 2nd, 2006, 01:06 AM

This is only really for people interested in design of digital cinema cameras for volume manufacture, or manufacture based around existing micro-controller boards.

For those interested in embedded Windows operating system for a camera project, here is some news on the latest Windows CE, which now has shared source a total re-write of the core, 32 thousand thread support with 2GB of virtual memory each etc. It is noted that it is good for realtime deterministic processing, something that XP cannot achieve without expensive code add ons (something people around here have discovered). The cost per device is measured in dollars.

http://www.theinquirer.net/default.aspx?article=35478

I have been watching Windows CE for sometime, as I was not impressed with earliest versions then they adopted Tron realtime parts in the operating systems in version 5, I think. Tron was an old time favorite of mine, it was an attempt by the Japanese to take over the Operating System market. It included a more correct emphasis on realtime deterministic delivery and very fast response, and probably efficiency, etc, something some American OS's were severely lacking in. If this latest release can bring these Tron values through it should at last be a much needed OS.

Contrast to Windows XP Embedded, confusing situation all these separate embedded OS's (that are to be combined, or not, I don't know the current strategy). If it gets the same advantages great, it will also get PC support. Previously, CE has been available in PC version in times past, it will also work on Arm etc etc etc architectures. These architectures will be coming in Ghz+ multiple core configurations, with all sorts of embedded data processing units (alternatives to many of the PC processing units, including media acceleration). There are a number of reference designs, and micro-controller boards that may attract substantial discount in volume. A device based on this will undoubtedly offer lower power requirements and lower price. A handheld is possible with this level of technology (or make one based on a HD web cam).

As I stated, this is only for people interested in volume manufacture, or based on existing micro-controller boards, and their are undoubtedly other realtime OS's that shift with micro-controller boards people might want to use.

Thanks

Wayne.

Wayne Morellini · November 8th, 2006, 10:30 AM

A major side note update for all the readers of the thread.

http://www.net-gmbh.com/e/3LayersCMOS.htm
http://damien.douxchamps.net/ieee139...ras/search.php

There has already been Foveon firewire cameras, but at greatly huge prices with SD like video resolutions. I wish somebody could do full featured machine vision HD cameras (and full customisable capture software) for less than $500. If they sold it as a security camera as well they would get the volume to do the price.

I have viewed the latest 14Mp sensor spec that was sent to me months ago, and I can tell you that even at that resolution, the spec seems to exceed most machine vision alternatives. I have also read that they are planning other products. If this is a new product using recent advances (not based on older X3 designs) we could possibly expect similar performance. This performance could greatly exceed the Micron sensors we have been used to.

There is little information, it is based on an Hitachi product, but no price or further details yet (I have contacted the company). The Firewire may limit the camera somewhat. But this is a first camera, even if ti is not suitable we may see something better in future.

May 26th, 2006, 09:43 AM	#466
Wayne Morellini Inner Circle Join Date: May 2003 Location: Australia Posts: 2,762	Part 2 The really too radical Patching video cameras to make a larger image: What do you get if you stick two 720576p camcorders together, something better than DVCPROHD 720p. What about if you stick three together, something much better then DVCPROHD 720p, including extra wide 720p. This has already been done, as reported on some thread around here, I hear. But ideally you would need a prism to patch the images together (like the Olympus 8MP SHD Eng camera did with three HD chips) and to synchronise them. An alternative, would be to use a Projection screen, like they use on the 35mm Lens adaptor projects and align the cameras to the image and record them from there in synchronisation. Using full frame still cameras in sequence: You may have read my Red camera price joke, where the true maximum price of a 11 Mpixel camera can be determined by aligning a number of cheap cameras together till they reached the desired resolution ;). Another solution, I have recently come up with, but something that will probably be impractical due to the problem of angle of photography, is to put an square/rectangular array of synchronised still cameras to take pictures interleaved at the maximum rate (even circular would work). This is different as each camera is taking a full frame, but because these frame rates are like 1-7 fps, depending on model, the next cameras have to take a picture while the previous one is saving the picture. But the problem is that you will get an movement, the closer an object is to the camera the worse this shift will be. To get around this problems you can use a projection screen, but the angle of the cameras will make the screen look warped as the angle of perspective changes, as it is unlikely you can put it far enough away. So again, prism is the best solution, but the prism would have to have many output surfaces, and be complex. There is a way to make it with a series of first surface partial mirrors (these have mirror surfaces on the outside of the mirror, the partially mirrored surface transmits some of the light through instead, to be reflected by the next mirror). But each mirror has to be a different size, and different partial transmission, to capture the same amount of exposure, with different focusing distances (if you do it that way) and many of them, so it to is complex, but an interesting experiment. Things undone: I'm not bothering ;) Sensors, where does it end* Kodak have plenty of nice sensors, I went through just about all of them, virtually three pages to get to the basic specification of each :( (no wonder I didn't do it before). Some have at least 100K -ev well capacity. Pixim has some extended latitude stuff, and their site has development kits, and lists of box, dome and network cameras using their sensors. Micron is getting into extended latitude, on their Super Wide VGA sensor at the moment. Cypress has Ibis with extended latitude and global shutter, non cheap well capacity and descent SN, and their Smalcamera sensor company with cheap sensors with extended latitude. Extended latitude seems to be a feature on the rise across companies, which is good. Foveon has 3 colour per pixel, and I have found another, that I am trying to communicate with. Altasens has good Quantum Efficiency, SN, and I guess latitude, but what is the fill factor?. Fuji is doing a 3 color per pixel chip for next year or so, based on an more organic basis, apparently. Too, many more to go through, but this is the important ones so far. Undone work: There are probably more than 20 other sensor companies to check all their sensors, lists in thread above :( What resolution for shooting 720p and 1080p on the same camera: There is a trend to producing new sensors with ever increasing resolutions, and lower resolution sensors getting left behind. So the problem is that there might be limited low cost options at 720p and eventually 1080p, so a higher resolution sensor might have to be used. What ever resolution that you want to shoot in bayer, you should have an sensor that has, or a bit more, than a multiple of the resolution you want to shoot, because proper binning will give you accurate pixel for pixel matches. For instance, 720p is 1280 pixels across, so you could use A 1280, or a 2560 or 3840 etc pixel width. For 1080p, it is 1920, 3840 etc. You see there is a match up at 12803, and 19202 and 5768 (not a standard resolution, but one I have seen used in the past). GPU encode enhance* I have read through the Nvidia Pure video Marketing, and it does claim GPU encode assist (not complete but only assist). Alternative commercial cameras: Panasonic and Sony has announced AVCHD h264 18Mb/s, for DVD, Blu-ray and SD recorders (I hope HDD also). This should be good, and hopefully fix up the problems of HDV with motion artifacting most of the time, and help with low light noise performance degradation . It does 1920*1080p. H264 is 2-3 times better than Mepg2, I don't imagine that this transfers to the problems, but the situation should be better a lot more often, and other images a bit nicer (if used on the same quality sensor etc). So prosumer version of this will hopefully be everything that HDV should have been. Going on interviews and reports (mentioned in various threads on the subject) Sony is expecting to release cameras, and Panasonic maybe releasing a SD card version late this year, or into next year. http://www.dvinfo.net/conf/showthread.php?t=67127 Samsung announced a h264 around 18Mb/s, like AVCHD, last year, due August, uncompressed HDMI (a digital component replacement) output. http://www.dvinfo.net/conf/showthread.php?t=58261 Ambarella has announced that a number of manufacturers will be suing it's revolutionary h264 chip for cameras starting at $799. Still cameras will get HD video one day, Sanyo HD1 has already started with Mpeg4 9Mb/s. Panasonic, at NAB, announced a new High Profile H264 10bit 4:2:2 50mb/s intra (it does not do inter frame compression, used for movement) codec. Cameras will be at Nab next year, maybe even a HVX200 version we hope. Even through people think h264 is purely a delivery format, the high profile was brought along to make it into a robust editing and lossless format. The profile goes to at least 12bit 4:4:4 at many hundreds of megabits. Altogether this Panasonic version will, hopefully, be a descent consistent quality professional codec, a grade step down from cineform I imagine, and two from lossless, but a step up from AVCHD. The present HVX200 with the DVCPROHD 100Mb/s goes some way to that already, but still 8 bit. JVC showed a Hard drive Everio HD camera last year, even announcing at a dealer conference, that it was to be released around the 05/06 turnover (this is reported by member on dvinfo who was there). This camera, and a few others, have been delayed, the question was why? This new H264 stuff hopefully indicates that there might be some rethink, and hopefully we might get h264 (though JVC is not a member of AVCHD, nor is Samsung who is doing a h264 camera). If so, there are more options. http://www.dvinfo.net/conf/showthread.php?t=62117 http://www.dvinfo.net/conf/showthread.php?t=52458 So things are looking Rosy. There is one more possibility to come, but I am negotiating at the moment. I am not quiet finished there will probably be at least one more post. Thanks Wayne

May 27th, 2006, 08:04 PM	#467
Keith Wakeham Major Player Join Date: Jan 2005 Location: St. John's, NL, Canada Posts: 416	I think this is a first here, a real summary of sorts. All in all some good stuff, and unfortunately makes me feel a little bad that much less has been accomplished than I would have wished for this group. Most of the people who started with the crazy ideas have moved onward and away it seems. The most successful I think being Reel-Stream. I hope that I'll be able to migrate my site off my university server, and setup some forums like they did in the comming months and firmly establish something. Over a year ago I started work on my project which originally was to use variable timing with a KAI-2093 CCD and transport bayer data over 1/2 a HD-SDI link. Then decode it back to data, and write the data to hard drives. The first of this came as developing hard drive control and after months of pouring over info started to make some progress. But once I went to finish off the camera side I found that money was a very limiting factor. Then I went for a more direct approad and bypass the HD-SDI with some outside help, but eventually decided back on going for a pure recorder that would be general purpose. That was 4 or 5 months ago. Lots of programming latter, multiple circuit board designs, all sorts of problems and just a little further ahead. Everything I do seems to hit a huge delay and I appoligize to everyone for that, I don't think i've realistically met a single deadline in ages - I guess i'm always way to optomistic. Honestly I think that people out their in the big name companies know about this place. Heck, they might even know my name in some circles of these companies as the crazy engineering student working on something completely doable without big backing, and honestly they probably count on this as the reason its not done yet and hope that I never finish. I recall people at panasonic commenting about Reel-Stream negatively. So I have no doubt that when I am finished and show what i've spent a year of my life developing that I too will recieve negative comments. Nothing is perfect. However, to all those who speak negatively about this I have to say one thing. Do you think nobody from the Red development team has ever been in this forum. I'd bet that Jim Jannard himself has been in here at some point and got some little ideas. It places like this that recognize what can be done now, what we want done, and the methods to get their -- and most of us have spent a lot of money along the way. Since I've put up my site the biggest question I've recieved is when can people get one, but what its now turned into is people stating they have a project in a month and asking if it will be ready then. Some have even gone to stating that if its not ready by the time they begin shooting then it won't be any use to them. I'm hoping this is a reaction to the "shooting season" being upon us, but I tried to put my email out their to help people and discuss ideas and try and find the people the first units will most benefit. Anyway, for all the crazy Ideas Wayne keeps coming up with and all the reasearh and work, I got to say he's a really good fellow. He doesn't lose faith in what people are working on, and I have to say a big thanks for not losing faith in me. I have a feeling a lot of people have and although I'm not sharing my FPGA code, I feel I have let the community down a little everyday I don't have a device or product to show or some more really proof of what i'm doing. I'll never quit trying to improve things and ideas, and hope that someday they will make a difference. So a big thanks to Wayne for showing us what the years of forums has given us so far even if it seems small, you never know who is listening. __________________ www.engr.mun.ca/~wakeham/index.htm

August 28th, 2006, 04:05 PM	#475
Frank Hool Major Player Join Date: Apr 2006 Location: Barca Spain Posts: 384	back to basics Found just some settlement reading about image sensors overall and about my present investigation subject what sensors for what CMOS vs CCD. All from Dalsa: http://www.dalsa.com/dc/documents/Im...8-00_03-70.pdf http://www.dalsa.com/markets/ccd_vs_cmos.asp

August 30th, 2006, 07:07 AM	#476
Wayne Morellini Inner Circle Join Date: May 2003 Location: Australia Posts: 2,762	Microsoft 1.3MP HD 30p video Webcam VX-6000 Thanks Frank. Here is a new HD video Webcam. http://www.dvinfo.net/conf/showthrea...337#post535337

September 12th, 2006, 05:35 AM	#477
Wayne Morellini Inner Circle Join Date: May 2003 Location: Australia Posts: 2,762	HDMI uncomrpessed capture for around $250. For those Digital Cinema camera people still subscribed to this thread, I have some major news. Declink has released an uncompressed HDMI capture card fro around $250US. Apparently you can select a codec to compress the footage during compression. It includes a free photojpeg codec, but something like Cineform o a free open sourced codec might be preferable. People are also talking about using a a component to HDMI converter fro older cameras without HDMI. http://www.dvinfo.net/conf/showthread.php?t=75102 I would liked to have started a thread on this in alternative imaging for the other Digital Cinema camera people not subscribed here, but somebody beat me to a thread in news. But people are spread across the SI and Red forums now, that would have been interested. A word of caution, I would be wanting confirmation on which cameras output pre-compressed/uncompressed HDMI. I would be wanting confirmation about the format that it records in and wherever it can precisely lock onto the exact signal format the camera actually uses.

June 14th, 2006, 11:54 PM	#468
John Benton Major Player Join Date: Oct 2005 Location: NYC Posts: 658	Keith, I just shot you an email, before reading this post, asking how it was going. Thanks for the update and Kudos...Just keep going... you will conquer it.

June 15th, 2006, 06:54 AM	#470
Matteo Pozzi Regular Crew Join Date: Jul 2005 Location: Milan, Italy Posts: 169	Hi to all did someone can tell me what camera (gige, firewire usb2) use the altasens cmos I know only the Silicon imaging one! many thanks Matteo Pozzi

June 16th, 2006, 04:54 AM	#471
Wayne Morellini Inner Circle Join Date: May 2003 Location: Australia Posts: 2,762	Yet, nobody ever started out these projects to make $20K cameras that only a few hundred would get in a year. But virtually all were aiming for sub $10K, preferably sub $5K, that they could build themselves, or would be made in 10K+ quantities. We are miles away from where we were going, led astray by some, and it looks like Pana is going to supply the minimum quality, maybe at double the price at least of the top quality. What was not doable in 1080p editing last year is doable now or latter, and 720p could have been done sooner instead, Cinelerra has been doing for years in true 4:4:4. The PC/processing market is about to go through a major advancement in processing power over the recent advancements, and has just gone through a major advancement in drive speed. I have no problem for the numbers in 720p, even in 1080p with a bit of effort, but the advancement will make 1080p easier than before. 4K is the new target for the industry, that is where things are difficult despite all recent advancements, but something that can be done in future. A few hundred people will benefit from direct cameras, so it is good for them, the rest of us will have to pay for the H264/XDCAM HD stuff. Thousands can benefit from the HDSDI work. Hopefully the post $10K cameras will inspire somebody to do sub $10K cameras, where you can buy double the quality compared to the regular video camera manufacturers, and force them to drop their prices. As far as focus goes, it all comes down to who do you care about, ourselves, or everybody else. Caring for everybody else requires much more effort and less easy money, but in the long run better for everybody.

July 22nd, 2006, 08:56 AM	#473
Obin Olson Trustee Join Date: Jan 2003 Location: Wilmington NC Posts: 1,414	out. no worries!

July 23rd, 2006, 08:05 AM	#474
Wayne Morellini Inner Circle Join Date: May 2003 Location: Australia Posts: 2,762	I liked that Furp animation of yours, but can't find it anymore. But what does "out." mean? You know, I would love to see that on TV, I could point to the screen, and say, "I know the guy that was responsible for that". Ren and STi, and Crazy Frog can stand aside. What is the latest on that line?

September 13th, 2006, 11:52 PM	#478
Wayne Morellini Inner Circle Join Date: May 2003 Location: Australia Posts: 2,762	Have contacted them. It looks like HDMI cameras are currently 8 bit 4:2:2 at the most. So, this would, in practicality, be a step down to what we would expect to achieve with machine vision cameras and the Elphel project (which currently is goign towards hard disk interface and bayer in the other thread) . But as a solution a step above what we are likely to achieve from a webcam without 10 bit uncompressed. Notebook version is yet to be planned. I'm impressed, definitely a budget uncompressed solution. But a notebook version for similar price, or budget field recorder, would make life a lot easier.

November 2nd, 2006, 01:06 AM	#479
Wayne Morellini Inner Circle Join Date: May 2003 Location: Australia Posts: 2,762	Windows CE 6.0 embedded OS This is only really for people interested in design of digital cinema cameras for volume manufacture, or manufacture based around existing micro-controller boards. For those interested in embedded Windows operating system for a camera project, here is some news on the latest Windows CE, which now has shared source a total re-write of the core, 32 thousand thread support with 2GB of virtual memory each etc. It is noted that it is good for realtime deterministic processing, something that XP cannot achieve without expensive code add ons (something people around here have discovered). The cost per device is measured in dollars. http://www.theinquirer.net/default.aspx?article=35478 I have been watching Windows CE for sometime, as I was not impressed with earliest versions then they adopted Tron realtime parts in the operating systems in version 5, I think. Tron was an old time favorite of mine, it was an attempt by the Japanese to take over the Operating System market. It included a more correct emphasis on realtime deterministic delivery and very fast response, and probably efficiency, etc, something some American OS's were severely lacking in. If this latest release can bring these Tron values through it should at last be a much needed OS. Contrast to Windows XP Embedded, confusing situation all these separate embedded OS's (that are to be combined, or not, I don't know the current strategy). If it gets the same advantages great, it will also get PC support. Previously, CE has been available in PC version in times past, it will also work on Arm etc etc etc architectures. These architectures will be coming in Ghz+ multiple core configurations, with all sorts of embedded data processing units (alternatives to many of the PC processing units, including media acceleration). There are a number of reference designs, and micro-controller boards that may attract substantial discount in volume. A device based on this will undoubtedly offer lower power requirements and lower price. A handheld is possible with this level of technology (or make one based on a HD web cam). As I stated, this is only for people interested in volume manufacture, or based on existing micro-controller boards, and their are undoubtedly other realtime OS's that shift with micro-controller boards people might want to use. Thanks Wayne.

November 8th, 2006, 10:30 AM	#480
Wayne Morellini Inner Circle Join Date: May 2003 Location: Australia Posts: 2,762	Finally, alternative Machine Vision Foveon X3 cameras. A major side note update for all the readers of the thread. http://www.net-gmbh.com/e/3LayersCMOS.htm http://damien.douxchamps.net/ieee139...ras/search.php There has already been Foveon firewire cameras, but at greatly huge prices with SD like video resolutions. I wish somebody could do full featured machine vision HD cameras (and full customisable capture software) for less than $500. If they sold it as a security camera as well they would get the volume to do the price. I have viewed the latest 14Mp sensor spec that was sent to me months ago, and I can tell you that even at that resolution, the spec seems to exceed most machine vision alternatives. I have also read that they are planning other products. If this is a new product using recent advances (not based on older X3 designs) we could possibly expect similar performance. This performance could greatly exceed the Micron sensors we have been used to. There is little information, it is based on an Hitachi product, but no price or further details yet (I have contacted the company). The Firewire may limit the camera somewhat. But this is a first camera, even if ti is not suitable we may see something better in future.