3 channel 36 bit 1280 X 720 low $ camera - Viper?

[Obin Olson]

also what I think you need to do is have some type of chip/computer take the 3 RGB images and layer them into one before the images go to the hard disk, this way you will have much less strain on the system for capturing

[Steve Ipp]

Thanks, Obin, I agree, the bigger CMOSs will be better.
The software will do the math, - only RGB output frames will be written. There even might be some sort of lossless compression.

[Obin Olson]

Steve FYI the single IBIS5 chip looks really bad almost no color at all...it's very washed out looking

[Wayne Morellini]

Sounds like just what I need. Sumix could make themselves a good deal of sales when done right. I think up to 100 thousand sales could be generated (10 thousand could be likely). I want to get something like this, but I'll wait until everything is sorted out.

If a complete system can be worked out then these sales can be easily sold to inde and video clubs, and videocamera forums, and by mail order. As most, more average of us, don't want to DIY the system themselves, before they know it will work. Still that would leave hundreds of people left for the first batch.

Whatever we do we should not forget, TV production, so 25p and 30p, with variable shutter speeds, gains, would be very important. A SD picture can be extracted from that.

That thing with the recrptor pads of the CMOS chip beign surrounded by transistors, what about the increased flyscreen effect?

We should look at 1080p or more, if it's not much more expensive, otherwise, if the 720p camera is very successfull, leave it till next time. I would pay a few hundred more for it, a bit more for SHD even, as long as the picture, latitude, niose, range, min lux was still good (JVC HD10 is not). If those characteristics are bad then I wouldn't be.

What video editing packages do we use, will linux give better capture performance, are their in Open Source editors that could add support for this camera project?

As I understand a normal high speed HDD will do 66-100Mb/s without raid, what about if we have a Gigbyte of memory to buffer data into write burst (most normal drives overheat during continouse writing and have to cool down before continuing) to keep the drive cool. Would that help.

I know that you can get 5/14 inch drive sized consumer computers. You can get simular sized cases to put your own board into. www.VIA.com.tw, have mini-ITX, and nano-ITX motherboards (17*17cm and 12*12cm pc mainboards), Would that be suitable for just capture? You could make a sholder mount camera with that form factor.

Looseless compression, and decompression for editing, is fine.

Soem thoughts on outputs: coomponent ad RGB, and s-video if availble, for dumping commercials to local TV stations Beta decks. USB2.0 is standard for many new computers, can 2 or three USB2.0 ports be useded to give three times more bandwidth? IEEE1394a is more standard in video circles, version 1394b goes from 800-1.6Gb's, but who's got it? Direct camera interfaces, are best for dealing with TV stations and PRO equipment, but espensive, does anybody know any cheap alternatives? Gigabyte ethernet, OK as lonhg as it's cheaper that three USB2.0's. Wireless, standardised or custom, as long as it is cheap (less than a couple of hundred), would be excellent. Need a hardwire interface and a wireless interface, buffer memory, as certain things could break the signal up (explosions, motors, electrical niose, lightnisng etc).


A guy over at camcorderinfo has some very good suggestions:

http://www.camcorderinfo.com/bbs/showthread.php?s=&postid=70544#post70544

The smal sensor is over 1 inch and has 500 times the latitude.

http://www.smalcamera.com/technology.html

[Steve Ipp]

Hey, Wayne
Thanks for support.

>What video editing packages do we use, will linux give better >capture performance, are their in Open Source editors that >could add support for this camera project?

If you get plain RGB frames, the options are endless. I like shake a lot. It's simplicity, speed and power are pretty amazing. Apple did a great job adding sound editing capabilities to the package.
Although shake isn't an NLE (ruther a compositing system), it can do pretty good if there were no other options.
I'm sure there are other HD editing progs out there but my other favorite is 5D Cyborg 2.0 which has been swallowed by Discreet a few years ago. If you are lucky, you can still get it from eMule.

As for the portable system ideas you outlined - I am planning to assemble just that after I see the first results of the camera output. PCs are getting more power more speed and shrinking fast. One deterrent here though is the final cost of such a system. All the PC components could be somewhat around 2000 US.
If we could get Linux guys signed up, the cam could shrink even more. I saw a website by a Russian electronics engineer, he's built one chipper 1024 X 1300 (I think) with bayer filter totally around linux server. The cam is tiny and it can transmit the stream wirelessly.

Three USB 2 ports is teasing, I'll bring it up, as well as s-video output, in my next reply to the company.

[Patrick Wright]

<<<-- As I understand a normal high speed HDD will do 66-100Mb/s without raid, ... -->>>

The ATA66/100/133 and SATA150 specs only allow for burst speeds that fast. There is yet to be a single drive that can reach that speed even on the outer/fastest tracks. Most drives can only send data out of their 2-8MB cache at that rate and cannot sustain it.

The most important speed rating of a drive is its minimum sustained write speed. That is the absolute slowest rate at which it will write data to the drive from a streamed source such as the proposed cameras. It has been suggested that 25MBps is the very minimum needed and unfortunately there are still a few 7200rpm drives out there that can dip below that. Alot of manufacturers only give their average and fastest speeds and this can be rather misleading. Yes, many hard drives can write data at 50MBps but this is only at the farthest outside tracks and is not to be considered consistent. Even average write speeds can be misleading if the drive spends an inordinate amount of time on the inner/slower tracks.

Western Digital produces a 10,000rpm drive that has a minimum write speed of around 50MBps and it only gets better form there. The only downsides are the price, power needs, and heat generated. The power required may be too much for any battery pack to resonably handle though. A review of this drive and a really good 7200rpm drive can be found here http://www.tomshardware.com/storage/20040123/index.html

Another option might be 7200rpm laptop drives in a Raid0 (2+ drives) or Raid5(3+ drives) array. Laptop drives by themselves are a bit slower than the same RPM desktop counterparts but raid0/5 would solve that. The power requirements for laptop drives is much lower than desktops and a hotswapable unit could be made to house and control three of the drives which would still probably have lower power requirements than one desktop hard drive. When one unit got filled up it could be swapped out for a fresh one. The filled unit could then be attached to the editing system by usb2/firewire. This would be similar to the Panasonic Flash card system except with laptop hard drives, though more expensive than regular hard drives, they are far cheaper than flash media.

---Patrick

[Wayne Morellini]

<<<-- Originally posted by Steve Ipp : Hey, Wayne
Thanks for support.

If we could get Linux guys signed up, the cam could shrink even more. I saw a website by a Russian electronics engineer, he's built one chipper 1024 X 1300 (I think) with bayer filter totally around linux server. The cam is tiny and it can transmit the stream wirelessly.

Three USB 2 ports is teasing, I'll bring it up, as well as s-video output, in my next reply to the company. -->>>

Thanks. By the way, Link, Link? ;)

One other consideration is to where ever Windows XP and USB2 can keep up with the realtime considerations of the sampling (without pausing, stuttering or breaking up). Thats the only major reason I suggest Linux, as I have had many bad experiences with Windows performance (third parties used to sell real time OS replacement cores for it, it was that bad). I understand that Firewire had better rliability in this respect to USB1, is USB2 any better?

One other aspect is that what we are looking at buying, pluss video editing package $1k-2k + computer could add upto $7000. Now, I saw a local ad for a twin security camera monitoring system for around $230US, that is not a cheap company either. So it occurs to me that just a HD head may not worth 12 times the price in high volumes. Ask about major bulk discounts past the 10 they are asking for. 1K might be possible (the prism might be the major hassel) + the rest.

The video editing software, I don't know the ones you suggested, but is there any that can support the format, timecoding, and all the pro editting features we would need for commercial/inde work, auto: colour, gamma, lighting and contrast adjustment, at a price of less than a few hundred, rather than thousands, on a packaged OEM deal, or open source?

It occured to me last night that the camera will need control software for manual/auto, iris, gain, colour correction/white balance, focus (live doco work) etc. Along with possible looseless compression and decompression for codec and editting suit (I remember a place on the web that lists the compression codec, including opensource but can't remember where). This is definetly where open source might help over OEM solutions (which might be licensed)?

I suggested Linux, because of the real time processing capabilities might be better. In capture mode, I would suggest that the system comes on instantly in a special mode without Windows. Linux offers some recompilable portability between PC and MAC, but I would like to put forward the Toas vOS system (RTOS and intent) as much more portable, efficent and faster. It is the system at the heart of the new Amiga initatives, I believe it would give the best efficencies, is very small for instant on, and is available for many hardware platforms, and Linux I believe, though a number of consumer electronics companies are shareholders. Actually I would not be surprised if some cameras use Toas for the firmware OS (just looked at the site and it is used). For the PC I believe it is probably the best system around at the moment. It is it's own Virtual OS but also can be programmed in Java (it had the most efficent Java plugin, Intent, on a PC). Through either of these (but especially through machine code, or TOAS) you might also get a greater through put of data, requiring a slower cheaper processor. Some www.via.com.tw ITX mainboards have flash instant on capacities. I believe www.via.com.tw is easy to work with for projects using their technology. I am surprised that capture required such horsepower as is quoted before, VIA will probably have 2 ghz low powered pocessors this year, I imagine. Even two of their boards put together is 12*12cm *2. We are talking about 14.5 cycles for every pixel at present 1.2ghz, but we are talking about many times that with double 2.4Ghz server cpu's. It all comes down to how you program the underlying hardware's efficencies (also getting rid of interference from background OS operations), the PC game industry used to have these problems, and any fast game, programs around them. The game industry is probably one of the few industries that takes this into concern, as some of the inefficencies mount upto 100's of times speed reduction (in real life, on average, maybe only double). Just compiling in C (Linux) can drastically reduce performance, I don't know if any C compiler (with highly optimised C code) approaches faster than 80% Machine code speed, and it might be less than that for the sort of capture we are talking about.

http://tao-group.com/

Searching the following opensource sites should reveal related Open Source projects:

www.sourceforge.com is one of the most common place for Opensource projects. There is a group (maybe more than one) working on mutimedia software for Linux, I can't remember the name, there might be one for editing software or something, and I don't know about capture software, but these are the people to approach, and their forums, as they might be interested in coding such a beast (as well as ex engineers and software programmers for camera companies). You can sell, them on the idea, that if they thought it was bad to pay Micro S too much money for something they don't like, imagine how it is paying $35K too much for a quality HD camera system, also stopping them from owning their own cheap high quality PC HD camera system :)

www.opencores.org is opensource hardware, but I believe there is another more suitable bunch with more general opensource electronics designs, but it is burried among many thousands of pages of paperwork around here. Somebody at opencores may know.

http://www.opencores.org/projects.cgi/web/video_systems/overview

If opensource software and hardware is put in place, it canbe reused by other future camera projects.

So those are the programming options I can tell you about.

Don't let any of this stop you getting a cheap, well worked out, comprehensive version out as soon as possible, as Open Source can take a while. All the custom programming can be done in the background for future upgrades, and could take years.

Pactrick, Raid issues, I thought that was the sustained rate, my appologies. I have not read anything for a while, but I think that the last article I read was on tomshardware.com about drives for multimedia or video editing, but there are a number of other sites I could have read it on. One computer can carry 4 drives, so we are talking about enough speed in parallel (the memory system can deliver many times more), adding 4 120 gb drives together will give you oddels of space. I suggested the 3.5 inch drive because of low cost compared to laptop and multimedia 3.5 inch drives, and performance (I'm interested in a shoulder mount). I had another consideration for power, but that is not relavent here. Large main memory buffer and twin drives will probably allow close to maximiun burst speed (one at a time) without stalling because of overheating. The data goes into a buffer, is written to one drive, before it overheats it then goes to the other drive, or buffer, some overlap in the drives maybe possible, I actually don't know.

Additional. I followed the link to the site of the cmos sensor, and found many new sensor designs there, there was a link to a 1000fps camera made with their sensor, probably too expensive for our needs. It's good for some ideas, but I think it is far better to stick with Sumix, as it will be far better to work with them.

How much extra would the Panavision sensor cost?

http://www.visiblesolutions.com/
http://www.visiblesolutions.com/phantomv9.html

[Wayne Morellini]

<<<-- Originally posted by Steve Ipp :
If we could get Linux guys signed up, the cam could shrink even more. I saw a website by a Russian electronics engineer, he's built one chipper 1024 X 1300 (I think) with bayer filter totally around linux server. The cam is tiny and it can transmit the stream wirelessly.
-->>>

Steve do you have a link for this camera? If you do please feel free to post it in my alt imaging thread as well.

[Steve Ipp]

http://www.linuxdevices.com/articles/AT2441343146.html

http://www.elphel.com/

[Wayne Morellini]

Good old Linux Devices ;), I keep a book mark file of around 1 MB, and a library of 8-10 thousand printed pages, of information for my variouse projects, I've been there many times. Thanks for the link.

Wayne.

[Obin Olson]

steve, how is it going?

[Steve Ipp]

I sent a detailed explanation with camera requirements to a Sumix engineer, I also sent him a link to this discussion to get a grip on people's needs :)
I reckon the guys have started the development. According to them, they need at least 3 month to come up with a prototype.
As soon as I get any news, I'll post here.

[Steve Ipp]

My pleasure :)

[Obin Olson]

steve...got a lead on a GREAT single CMOS chipcam!

http://www.siliconimaging.com/Sample...0DO%20lens.jpg

check that out! that is sinlge chip CMOS...I am going to buy that system, 60fps 30fps and 24fps and 48fps...that chip is lightyears ahead of the quality you can get from the IBIS5 chip!

it is 1/2inch..a bummer but better then 1/3inch and I think I may just build a mini35 for it and use standard Canon 35mm lenses

[Obin Olson]

ohhhhh...look at this!

3mp image

http://www.siliconimaging.com/Samples/3170%20Flash.jpg

I am not sure if it can do 24fps or not