Development Platform for DIY cameras

[Keith Wakeham]

Micheal
from the sounds of it the camera sounds good, now if only they could provide a few specs on the sensor used. It must be a 1394b because with overhead bayer 1080 at 24p won't make it over 1394a, well it should in theory, but because of protocol overhead it won't so 25p would definetly have to be some sort of dual link firewire or use the new 800mbit standard

As of right now I'm still working on the fpga program for the timing and the circuit for the gain and stuff. Support IC's are gonna cost a little more than expected but not a significant amount compared to the sensor. I'll likely release the fpga code for the timing in a few weeks and it will allow for 24, 25, and 30 progressive modes - so long as i can run them all off one high speed global clock, if not then a method of changing clocks. The code is vhdl written in xilinx ISE 7.1.

I'm waiting for the spartan 3e development kit to come out because it has a usb 2.0 interface and documentation so usb development might actually happen, but i'm still going for getting a a viewable signal out of the ccd that is confoming to some video standard - for now, just gotta learn how to use block ram or connect to sdram. I want to minimize the number of external chips, but chosing a ccd i should have know i was asking for it.

Anyway, good luck on contacting that manufacturer

[Michael Maier]

Keith, the sensor is a 2/3" CMOS. Don't know who makes it. Didn't ask. Is it important? I know it does 25fps with a global shutter. Check my thread "Perfect DIY camera head". I post some extra information about the camera there. It captures video in avi files. So I think it would solve most capturing problems people are having, wouldn't it? I mean, any NLE can work with avi.

[Obin Olson]

I am sure i'ts the IBIS5 sensor as that is the only sensor that I know of that's 2/3rd inch GLOBAL and color...it's shit..not a good image...looks like a spy cam used in undercover TV news..if it's firewire then it's also 8bit, not good enough for shooting you will need 10-12bit images..8bit will also look like a cheap 1ccd spy camera

Originally Posted by Michael Maier

Unfortunately no link. Just a phone number. They would actually custom make a camera to specifications. I'm currently trying to gather information as what would be the perfect camera for a DIY HD cine camera. Then I will contact him and see if it is possible and how much it would cost.

[Michael Maier]

As I posted in the other thread, it does 8 and 10bit. So maybe it's not the chip you are thinking.

I will know soon which chip it has.

By the way is global shutter bad? I thought it was the best.

[Michael Maier]

Originally Posted by Obin Olson

if it's firewire then it's also 8bit, not good enough for shooting you will need 10-12bit images..8bit will also look like a cheap 1ccd spy camera

By the way, isn't the Drake 8bit? It doesn't look like a spy camera to me.

[Obin Olson]

Keith please email me off list.

obin@dv3productions.com

thanks

[Keith Wakeham]

If the data was packed properly then at 24/25p firewire can handle the output of the ibis at 10 bit, but because of protocol overhead, 30p won't work in 10bit.

I really think the only problem with ibis is the colour filters used over the photosites, which, as obin pointed out, leads to bad colour.

I'm pretty sure that most sensors, even with onboard adc, have an analog output so you could just stick amost any adc in their you wanted. You could go as high as 30bit if you can find the adc, or likely more. Getting the bits out isn't the problem, its what they are picking up.

I'm still looking into redesigning around a 3 cmos ibis5-1300 setup because I have poured over those spec sheets all all thay look really bad are the colour filter response on the bayer filter. I would like to see what a b/w one does during some test condidtions

[Obin Olson]

great..but how are you going to get a 3cmos setup working? do you have some beam splitters?

[Keith Wakeham]

I'd have to get some setup gear for the alignment but i know i can get the desired prisms and then just need to get a proper glue and make a semi-clean-room setup. A little cost, but it could be done fairly easily with a few hours of work and research and calculations.

[Aaron Shaw]

I don't know if you guys have seent these cameras but they may be worth looking into. It's the first I've heard of them at least. I do not know pricing:

Full 35mm frame size!
http://www.fairchildimaging.com/main...Harrier447.htm

Interesting 1" Chip camera (150fps):
http://www.photonfocus.com/html/eng/...s.php?prodId=1

[Steve Nordhauser]

Keith,
Look into laminar flow tables. They look like a workbench with a fancy filter mounted above. Clean air is blown down into the work area keeping a positive pressure at the table top. Dust doesn't get in. I've used this for working with sensors without cover glasses before.

Originally Posted by Keith Wakeham

I'd have to get some setup gear for the alignment but i know i can get the desired prisms and then just need to get a proper glue and make a semi-clean-room setup. A little cost, but it could be done fairly easily with a few hours of work and research and calculations.

[Keith Wakeham]

Fairchild makes sensors that can be purchased and framos.de sells them. They are expensive, very expensive, very very very expensive. I remember on sensor being about 8000 euros without any support electronics at all, at that was just a 512 x 512 one that done 150fps. Just for the sensor used in the harrier 477 it costs 3200 euros. And the camera takes at least 1 second to read out a frame, without using multiple ports it can take 2 - 4 seconds. Very far away from the 1/24 seconds per frame we want. From my experience, a camera costs at least 2 times the cost of the sensor, usually betwen 3 - 10 times normally.

The cmos one has a nice sized sensor, but is square and when cropped it won't be much more than standard def, along with the fact that its monochrome and they don't list a bayer option it is cameralink. To me that makes it fairly useless are hard to work with, but for people interested in high speed imaging then it could be useful, but not for high def. And i would suspect that this camera will be expensive, because it falls into a high framerate category.

Its not hard to find a camera that has the resolution or framerate, its hard to find one that someone can afford. Dalsa makes an excellent cameralink camera that is native 1920 x 1080 and all that jazz, its close to 10k. Awesome specs, can't afford it, and when developing you don't want to have to risk destroying 10 000 dollars every day you try something new.

[Keith Wakeham]

Steve thanx for the tip, I'll look into a setup like that, sounds interesting to say the least.

I'm still unsure what method I want to go. The Kodak camera head design is about 1/2 way done (FPGA design and testing is has to be done before the design can be finished), but the price of a 3 cmos setup is very luring because off the sensor cost. After some initial setup the cost to build one cmos camera would be more, but even if i built 2 the cost per camera would go down below the kodak design.

[Keith Wakeham]

A little note on where my design is going and maybe some people can comment and criticize the idea.

The camera design is 2 parts plus a hard drive (The hard drive is just a hard drive pretty much, so not really a main part). I know some people will not like it after the description and some people will.

The parts are a camera head and a "deck"

The camera head will output hd-sdi, but It will output the bayer over the luma channel and the chroma will be left empty. I'm doing this because it is fairly easy to implement and provides something of an industry standard. It will also output a component signal, but the signal will just be the bayer, from playing with some bayer images i made up it should be okay for viewing, framing, focusing, but not critical viewing of the subject colours. I might be able to implement a simple debayer, but won't be their initally in my prototype.

The least worked on and more exciting thing that i've come up with is the deck. And yes, it is more of a deck than anything. No, the hard drive will not have a file system and will not be directly readable by plugging it into a computer unless someone wants to make a program for reading it. It is a small device that will take in 4:2:2 HD-SDI and record it, uncompressed, to a hard drive (raw data into sectors). Initially it will only record the luma channel for the above camera head in the prototype. It will then be expanded to another hard drive to capture the chroma, or a second bayer camera. And possibly (because its not to hard to do) take component input for use with hdv cameras. It will have HD-SDI out and component out for monitoring just like a hdcam and betacam decks does. It will have a standard serial control, rs-422 or rs-232, can't remember which is standard for video stuff, but will be usable in things like fcp and avid without much work and will act like a normal deck.

I know we all want 4:4:4 but I think that a half decently priced 4:2:2 uncompressed deck could be very useful to almost anyone from high end pro work to low end work.

I'd apperciate some feedback. Like i said the deck isn't worked on much yet, but that is mainly because of my lack for the intellectual property of a sata or ata core for the fpga. They exsist, but companies never respond to my information requests.

Don't be afriad to leave criticism, good or bad, so long as you can give me some explanation of what is wrong and why

[Rob Lohman]

I have been highly interested in FPGA and possabilities to use it with SATA,
ATA, USB or firewire (+ storage protocols). However I found it all confusing
and not much information on what is available.

One thing I am wondering. How are you going to store a bayer stream in
4:2:2? That sampling refers to YUV, not bayer. Besides, you can't afford
to loose any bytes in the bayer stream because you really need those to
construct the RGB image. You could say that bayer already is a form of
4:2:2 (since green is sampled twice as much as red and blue), but don't
degrade this signal any futher. A bayer chip cannot really output 4:4:4,
however you could call it that if you leave the bayer stream intact.

Hooking up the harddisk and specifically reading of it might prove to be a
problem on the Windows operating system (I'm not sure how easy it is to
get raw sector level access to a harddisk, but it is something I was going
to look into).

Another option (perhaps) might be to implement a network (gigabit prefered)
link if you can find a TCP/IP module for FPGA and transfer the footage in that
way. Might be a lot easier to implement on the "computer" end, for all sorts
of operating systems (Windows, Unix, Mac)