Development Platform for DIY cameras

[Obin Olson]

I can host if you need it...
this all sounds great..your biggest challenge is going to be the power and viewfinder...as you may know our system will have atouch screen to run the camera...this will also have the image display..the problem is that in bright sunlight you can't see an LCD ..this is ONE of the reason we need a viewfinder system...I would guess that we can split the VGA port going to the touch screen AND the viewfinder..keep in mind this camera NEEDS to work with CINE style equipment..matte box follow focus filters lenses etc...so you really need to get all the specs of ARRI and panavision stuff so we don't re-invent the wheel...I am thinking maybe even use PL mount lenses..this would need a converter from C-mount to PL mount..maybe you can CNC that?

I did notice that the Drake camera is using some c-mount lenses that are of a very high quality..maybe we don't need PL..not sure yet...

[Keith Wakeham]

From what i remember when i used p&s mount for the canon xl1s that primes aren't light weight and they need a hefty mount, so what you will likely need is a a full blown mounting device because i personally wouldn't want to hang a PL mount lense off of a c-mount adapter. It will actually have to be mounted to the camera body so that the body takes the weight and not the camera.

I just realized I'm thinking out loud.

I'm familar with some of the arri stuff and mounts, but I don't know anything about panavision, i'd assume similar, i could be assuming wrong. Time for research.

[Keith Wakeham]

I just put up a little website for the time being on cjb.net to keep track of everything, i'll eventually move it to my university account this coming week

http://indiedp.cjb.net/

Not much on it yet, but its a start.

[Keith Wakeham]

I don't know if its because of a lack of intrest or maybe because of me not being on this forum for long enough or not being clear.

I'm doing exams now, and i'm halfway through (phew, only 2 more and i'm homefree).

If you have checked out the website essentially i'm pretty much tailoring the indie dp to be a housing for obin's camera setup, and right now i'm working on getting info on arri mouting stuff and designing a c-mount to pl mount.

Obin, if you check this, do you want the adapter to just connect, because the sensor size is way off from 35mm, so you won't get the DOF and the lenses or do you want to use a groundglass setup, even though it would lower the amount of light. Up to you. I'm also trying to figure out how to connect all this stuff together and my orginal intent of somthing that kind of looked like a shoulder mount camera seems to be less and less of a good idea. Do you want to go similar to the Drake camera, or have any ideas for the over all body. You know what you have, and so do i (to a degree anyway) so let me know your ideas.

As for the indie dp mini, that is going to be my little project. Right now i'm looking for an imaging sensor. I've been thinking of getting a sony one from framos.de and overclocking it with some cooling so i can get 24fps, but it might be a bad idea. (blowing up a 300 dollar ccd isn't something i'd like to do) I've checked the thomas register and globalspec and contacted kodak and it's almost impossible to get a sensor, let alone a good one at a half decent price, any ideas? (gotta buy at least a 100 of the KAI-9023 at 1000 each, if i wanted that one - ouch)

If i could get a reasonably priced FF or FT ccd then i'd do a mechcanical shutter, but doesn't look promising. Time to call dalsa and see if they low volume engineering samples.

[Wayne Morellini]

Hello Keith

We have been planning to do very small shoulder mount platforms for sometime (after the software is finished). I have been looking at the VIA nano-itx boards (12cm*12cm (or less)). The processors have 1Ghz for 1 W. I have been waiting for software, camera and newer higher speed nano-itx boards with Giga Ethernet links to come into the market. With Obin's latest progress it looks like more than capture and preview might fit on a nano-itx board (compression). I was thinking of even chasing up some non working ENG cameras and rip the insides out and put a system in it, anybody with defunct GY-DV500, or 5000 can also contact me. Being a trendy sort of person, all that matters is pro filming features, trendy looks, and size (smaller the better).

If you were thinking of just capture and preview, then there is a smaller/lower powered Via platform for the handheld game system, and Pocket PC's/Palm that could even fit inside the case of a PD150/DVX etc. Don't get too phased by the viewfinder/monitor problem, many RGB type displays (LCD's) have RGB subpixels horizontally (though on a PDA this might be oriented in a less useful way). So 640(*3 sub pixels) across will give you an idea on the focus of 1920 pixels (focus is radial, so only the horizontal is needed most of the time). JVC's new camera has a useful focus assist feature and that is to edge enhance the bits that are in focus in the preview.

I happen to be looking up a mid price Kodak camera at the moment, it uses a Foveon X3 chip and I know previous cameras had their own SDK. I believe I have a technique to get it to output HD live (also has snapshot mode). I will probably publish details (except for he HD stuff which comes off another project) to the list, as I lack the strength and time to do it myself for the moment.

Don't forget that with the smalcamera sensors you have a whole primitive HD camera system for $69 (don't like that cheaper model (more expensive higher res version available) I have it here, even though it works well with dual slope like feature sometimes it blooms and moirs etc) that has enough sensor bandwidth and USB2 bandwidth for the job (though the compression engine might be limiting factor). They also have a ready made development system with SDK.

Don't forget the original Rob Scott's camera system. He has been doing it alone, and doing it tough, for $100 for capture software. Obin gets most of the volunteers to do cases etc, Rob gets nothing but the occasional correspondence with Rob lobtham. I don't know where ever Obin plans a big fully built camera system or where ever he will sell the software independently (like Rob) for roll your own.

Thanks

[Wayne Morellini]

About my suggestion of a camera on the Foveon X3. I had a further research on it and cannot find global shutter again, and the frame rate was a misprint on a site, and this Polaroid camera does not use the SDK kit I thought it did. So it is not worth it. But here are the links anyway:

http://www.foveon.com/press_x530.html

http://www.foveon.com/press_SmallFormat.html

http://www.alt-vision.com/hvduo5m.htm

http://www.alt-vision.com/fo18-50-f19.htm

[Keith Wakeham]

To bad to hear about th Foveon, I remember when everyone got supper hyped about it then the low frame rate keep it back.

Personally, Even though naming things is likely a bad idea to start with I did anyway, for those who went to the website.

Unfortunately, My university account is botched and i can't serve from it, my luck, resulting in no updates on the page cuz cjb really sucks.

I have the website updated on my hard drive, but that is useless.

I'm going to completely develop the Indie dp mini, as it is currently called, into a full camera pretty much.

Right now i'm doing paper work design and its actually getting done somewhat. I've selected the body stylingly and how it should work (This might seem backwards to most but this is how product development works) considering what needs to be inside it (with plenty of room). I'm really looking hard at the KAI-2093 Kodak sensor and timing circuit. $250 for an engineering sample, 1150 for tested one without defects.

This sensor has dual output, which, when you crunch some numbers and stuff it does something really strange. Each output has the exact same bandwidth as full rgb or yuv standard def. (960 x 1080 = 1036800 = 720 x 480 x 3 ). I've been thinking that maybe i could botch a capture card, like a usb one, to taking the two signals.

The only reason i'm doing the pushing for the Kodak CCD sensor rather than a cmos is because it is perfect resolution and specifically design for hd imaging like the altasens and is available in any quantity i want, plus one other think. The diagonal is .75 of an inch, that is even bigger than the altasens area, and makes pl mounts way more of an option for the DOF issue.

I'm also glad that someone else took some intrest in my ideas and posted to the thread. (It seems i talk to myself in it a lot - i'm so lame)

I never though on the subpixel thing in lcd's, that is a really good point, and the focus assist on the jvc is a good idea, but i think it needs improvement from what i heard. I think it only shows the center for focusing, but what if your subject is off center. So if the software could move the digitally zoomed area around for focusing than that would be awesome.

I'm gonna try and look up that via plateform, because if it could use ide or sata drives and i could make it fit inside the pov body i'm designing, that would be awesome. I don't plan on compression, so i don't think i'd need much processing power. Most of my mockups have been under impression of a mac mini or itx stuck on the back of the camera.

Anyway, first day with no exams, horray, i'm done (for this semester), i'm gonna work my ass of with this design stuff today.

[Wayne Morellini]

Good to see you here. But the problem is that there has been so much written for so long, that most people are talked out/left. So you might very few people interested in anything outside the current projects. So it might be better to be with somebodies project (unfortunately) and post on those threads.

Advice on programming, there is a very steep learning curve for the sort of programming needed for capture (real time embedded) compared to apps programming. Very few programmers would no it, and not many programmers would be up to learning it (PC's/Windows XP are complex fro real-time). So if you want to do capture i might be best to team up with somebody that knows their stuff. It is also a long process, maybe better to avoid doing camera capture.

Machines: New Apple Notebooks and Mac Mini's coming out. Present one is limited. New Xbox2, more powerful than Mac's, power PC etc, cheap. Just pointing a few extra options than PC (that people are concentrating on). I don't know how much simpler Mac is than PC to real-time but i imagine it is, and even the Xbox2.

Do you have links to information on that Kodak stuff?

There are also various free web page providers around the world. I have seen them with 50 and 100MB of account space. Google search may help.

That Foveon camera I mentioned before, misprint, was a Polaroid.

[Keith Wakeham]

Yes i do links on the kodak and I have pricing

http://www.kodak.com/global/en/digital/ccd/
-here for genereal ccd

http://www.kodak.com/global/plugins/...3MLongSpec.pdf
-here for the specs on the one I plan on purchasing

The first link gives way to all kodak imaging sensors including cmos, but the KAI-2093 is a really nice on paper sensor.

First 100 pieces purchased $1130 ea
Engineering Grade $250 ea
Evaluation Board (Includes EG Imager) $2500 ea

The engineering grade will be what i plan to work with first, the have more sensor defects but are still 100% functional and good for proof of concept, just not so representative of a working device, their might be rows of pixels dead.

Right know I'm stuck with getting the digital output from out camera block design onto a hard drive. But really, this and software were the real problems to begin with.

From my experience I am pretty sure that i could get an fpga to take the dual outputs from the kodak, link them together, debayer and output a hd-sdi signal and an rbg (analog or digital)with a virtex II or virtex II pro - But then its the giant computer problem. A computer with PCI-X or If i wait a year then PCIe with a giant breakout box.

I don't have enough knowledge or experience with PHY devices so I can't implement ata or sata. The Virtex II pro has rocket io which can be clocked for sata, except there are a couple of necessary things the rocket io would need to do that it just can't do. One of them is something called OOB. If it did then i'd be plugging away bypassing the computer and using hard drives more like tapes.

Your probably right about the threads, maybe i should just go work silently like a shark and come back up when i have something or just join someone elses project.

Sumix altasens camera is gigabit so if they got it out soon enough then its pretty much straping it to a computer, and would make all my work useless.

I know that right now i can avoid a lot of the problems people are having. I'm that i can get the framerate, global shutter, dynamic range, and solve a lot of the problems people are having with sensors, but i just can't get the data stored. What i really need is something like the reelstream andromeda modified for my camera design.

[Obin Olson]

Keep posting in here..I do all the time..we all read but don't have alot to say as it's all been said and you can read it in the threads...just keep at it like myself...

[Obin Olson]

I would like to FIRST have a NON GG no 35mm lenses...I think that 2/3rd inch cmos will be enough..then later on we can do GG when someone has a REALL good one on the market..I like the Drake style box..but better can be done in CAD...

we are now having trouble with the timing in our software..working that out now...

[Keith Wakeham]

Where I am with your design for the unit is kind of mixed right now obin. I've done some preliminary designs for the body of the unit along with doing some research for pl mount and c mount lense and for the cine style gear. Amazingly it is harder than i though to get really accurate information on the lens mounts and stuff.

I know your mainboard and i have it mocked up in solidworks and i'm working to build an enclosure around, but right now the positioning of the parts seem to be the most difficult part. Its likely perferable to keep as small as possible.

Thanx for the encouragement obin.

As for my own design I am part way through designing the camera head to output digital bayer from the ccd and have the head board partially mocked up in solidworks. I also developed a flow chart and "word" programming if i'm going to do demosaic in camera and have it output component and hd-sdi. The parallel smpte 274 to serial smpte 274 (smpte 292 defines the serial transfer mode, not the format) can be done in a chip made by national semiconductor (clc030). Amazingly the chip isn't as expensive as i though, 65 CAD and most fpga's can handle parallel smpte 274 like spartan 3 or virtex II, and right now i have the development enviroment for designing fpga's.

I also have to say, from my research, the component to hd-sdi convertors can be made with little work and could cost about $200 to make, yet companies charge 2k -3k, talk about mark up. (adc + fpga or bufferes + clc030 ic = component to hd-sdi convertor). The Cb and Cr needes to be buffered and switched between to go into the clc030 and that is all that is too it, plus maybe a microcontoller and switch to tell the clc030 what format to expect.

I've tried to setup some of those cheap online hosting sites, but they are really poor quality and make it hard to upload sites. I think I'm going to piggy back my site on one of the sites of a team i'm on at my university until my account is fixed.

[Wayne Morellini]

Originally Posted by Obin Olson

I would like to FIRST have a NON GG no 35mm lenses...I think that 2/3rd inch cmos will be enough

Drake does that, and Rai has real good GG solution somewhere in Pipe line.

Originally Posted by Keith Wakeham

I also have to say, from my research, the component to hd-sdi convertors can be made with little work and could cost about $200 to make, yet companies charge 2k -3k, talk about mark up. (adc + fpga or bufferes + clc030 ic = component to hd-sdi convertor). The Cb and Cr needes to be buffered and switched between to go into the clc030 and that is all that is too it, plus maybe a microcontoller and switch to tell the clc030 what format to expect.

Look at my home camera thread (link is on the first page of the technical thread). You will find a link to a Russian camera that uses open source FPGA designs that are available on open cores (I think).

The cost would be $200, sounds right with markups you are looking around $1-2K. I think the Australian one might be one of he cheapest (around $1K) I forget the name, but "blackmagic" is the name of one of their products.

There are a number of cameras outputting better grade signal from their component output (the JVC HD100, and Sony HDV cameras being some of them). A really cool product would be some thing that could take cameralink, HDSDI (dual link), or component, output and translate it into an industry standard format signals (4:4:4, 4:2:2 but any resolution and bit depth) on one or two Gigabit (or usb2, Firewire, or Gigabit version of Universal Wide Band) to input directly to the computer/hard dive, and have it translate industry standard commands to the camera the other way. Then you could even use freeware products like cinerella (how do you spell that) to control any camera. It might not be much more complex then a HD-SDI on it's own.

[Keith Wakeham]

Well, hd-sdi (single and dual link) are really the industry standard for high end work so going to that format and using proper smpte standards are likely the best way to go in my opinion, even if those standards have some really stupid issues (smpte 274 which defines 1920 x 1080 60i/30p/24p is 2200 pixels per line with 1920 active, making an fpga necessary for my design for all the useless black before and after the usable signal).

I have looked at the russian fpga camera, but it wasn't as helpful as i thought it would be because most of what they are doing is realitivly more involved and images are moving at a slower speed and being compressed in the chip.

I'm working on finishing up the schematic for the camera block, but the timing chip is giving me a little headache. So many outputs and documentation on the thing is really poor. I migh look for another timing device that won't require me to have a micro controller just to control the thing. I want a timing chip that i set to 24p with 1 electronic shutter per frame and never have to change it. Maybe i'm just not understanding the KSC-1000 timing chip docs properly.

Fpga programing, not as hard as i thought but the speed concepts and how it works are what are hard to master, I'm a little confused on the debayer right now, so if i can get the code to simulate outputting the bayer to the luma channel then i think i will be happy. I know it sounds simple but because of the dual output of the KAI-2093 I have to write code to output black then the active line from one output, then reverse and output the active line from the second output then output some more black, along with sync.

Building the fpga simulation model is going to be a pain.

It doesn't seem easy to model gigabit ethernet and unless you use a virtex-II or virtex-II pro with rocketIO then you will need a phy layer, meaning a gigabit controller, and then you will need to know how to control that controller. So right now that is beyond me, same with anything else involving a phy layer (usb, 1394, ata[i'm still waiting to hear back on ata code from a company, and it might be embedded or it might just be a phy layer interface], sata).

I'm not 100% sure but I thought the sony FX1's component output was after compression so it was pointless because it is decompressing the mpeg2 stream to analog so your actually gettting lose. I'm pretty confiden that the JVC 100U is not after compression.

[Wayne Morellini]

Hi Keith

That's not quiet what I meant. I was suggesting to keep the video industry file/data uncompressed formats (with some extensions for higher bit depth resolution and frame-rate and 4:4:4/Bayer) but being able to capture from any camera interface by standard GigaE converter box. For the industry it might not mean much, for us it means we can capture from any camera interface cheaply to a computer/laptop and use standard software to edit the file. There are numerous HD SDI's and they don't allow access to cheaper machine-vision cameras (and the laptop card-bus HDSDI cards won't do dual link data rate, but dual GigaE will). So rather than do what everybody else is doing you could do something different.

Anyway back to the HDSDI. I think Elektor (or it's rival in the UK) had a SDI kit last year or two (don't know if it was a HD one or not). I'm sure somebody is working somewhere on a kit for dual link HDSDI (Open core and he open source repository might have something).

Originally Posted by Keith Wakeham

I have looked at the russian fpga camera, but it wasn't as helpful as i thought it would be because most of what they are doing is realitivly more involved and images are moving at a slower speed and being compressed in the chip.

That's only his present public design, I would recommend emailing him privately and asking about his next system he is working on.

I'm working on finishing up the schematic for the camera block, but the timing chip is giving me a little headache. So many outputs and documentation on the thing is really poor. I migh look for another timing device that won't

I can't see the doc, but sometimes these people assume a lot of experience to "connect the dots" so to speak. So it might require a lot of figuring out to get to grips with the information.

I want a timing chip that i set to 24p with 1 electronic shutter per frame and never have to change it. Maybe i'm just not understanding the KSC-1000 timing chip docs properly.

I would recommend 25 and 30fps as well.

It doesn't seem easy to model gigabit ethernet and unless you use a virtex-II or virtex-II pro with rocketIO then you will need a phy layer, meaning a gigabit controller, and then you will need to know how to control that controller. So right now that is beyond me, same with anything else involving a phy layer (usb, 1394, ata[i'm still waiting to hear back on ata code from a company, and it might be embedded or it might just be a phy layer interface], sata).

I see. I think using a third party circuit and programming manual is way easier than designing half these interfaces fresh.

I'm not 100% sure but I thought the sony FX1's component output was after

One of the Sony HDV cameras, at least, is supposed to have uncompressed output (but seems to be a bit poor). There was much talk, and testing (links in technical thread and it's own thread a few months back). The JVC HD100 is before compression (4:2:2 upto 60fps). I have heard of higher end cameras before compression too.