I've been trying to find an affordable but highly capable video camera for some time now, but it's always either one or the other. What I'd like is simply high-resolution, reasonable resolving power, and capture in a common format, not one that may be very compressed or obsolete in a few years. I've done some chromakeying with my 350D, and the results have been outstanding, much better than I could ever achieve with my current video equipment. So, I've been wondering... I've already been building a lot of my camera hardware myself, but this isn't exactly making your own tripod. My first thoughts: Using a single lens, but hacking together several camera bodies to somehow utilize that one lense. Each of the donor dslr cams shoot at a rate equal or close to burst mode, all together in a sequential manner. I believe Canon EOS cameras run on DOS. I don't know how buffers work; can they be increased? If writing to a single card can't be done at 24/30 fps, then multiple cards could be used, even if it would be a pain when it comes time to do post. All the cameras are triggerd by a master processor that sets frame rate and exposure. As for how to link up all the cams to use one lense- I don't know yet. A spinning central mirror to reflect the light, or another setup? I haven't looked up how the internals on the EOS cameras work yet. Are the components capable of running at such a rapid pace, or for more than several seconds? I know it must be possible somehow, but whether it's practical... what do you think?

I see that TerraMatrix raised a highly technical issue by saying,

been trying to find an affordable but highly capable video camera

I have a Sony CamCorder.  It seems to take reasonable videos -- but
when I try to use it to capture detail (such as videos of cars in traffic),
I find that the resolution is sub-par -- in fact, unusable for my purpose. 

I have an inexpensive hp point and shoot with a video mode.  While this
camera is capable of reasonable resolution for stills (~4MP), in the video
mode, the pixel resolution is 1/64 of the still resolution = ~60,000 pixels. 
This camera does have a burst mode that will take about 5 images in
sequence -- up to the available RAM in the camera.  OTOH, that must be
manually triggered -- unsuitable for my purpose. 

I have not yet tried an HD CamCorder, but from the published specs it
seems that the pixels per image would still be too low for my purpose. 

I have to guess that the problem facing the camera manufacturers is
that the file size of high resolution still images (4Mp image requires about
a 1.5MBy image file) at 30 fps exceeds the data recording rate of flash
memory.  That would require a file system capable of recording 45MBy
per second -- and to sustain that rate for the entire scene. 

Perhaps there is a camera with a video mode that is full resolution but
at a reduced frame rate.  Of course that might still not meet your needs. 

--
Martin

I see that TerraMatrix raised a highly technical issue by saying,

been trying to find an affordable but highly capable video camera

And then went on to ask about,

Using a single lens, but hacking together several camera bodies to somehow
utilize that one lense. Each of the donor dslr cams shoot at a rate equal or
close to burst mode
A spinning central mirror to reflect the light, or

My thoughts on these proposals ...

(A) Spinning mirror

Could be problems of distortion; dust; synchronization; ...

(B) Multiple camera bodies runing in tandem/sequence

Major problems to get frame synchronism; AFIK, burst modes are
manual; you would need to do a major hack to the onboard camera
firmware to get the multiple bodies to "cooperate". 

(C) IOW, "Too many moving parts."

As I see it, the central problem is the file recording data rate -- you
should attack that issue to get your solution. 

As a matter of fact, there is a solution to that issue. One of the defined
RAID modes is data striping.  You would need enough HDDs so that
their aggregate data rate exceeded the target rate of 45MBy per second.
You would also need to engineer an interface that interfaced with the 
camera's flash card interface (either SD or CF) and (with respect to the
camera) emulated a single drive.  Software associated with the multiple
HHDs would do the data striping.  That software would also handle the
replay issues.  Linux is capable of the required data buffering and RAID-0
issues.  If someone has already created the SD or CF emulation hardware
and associated transfer cable, this is possible. 

BTW, USB 2.0 could provide sufficient data transfer rate. 45MBy/sec needs 
360,000,000 bits per second and USB 2.0 provides 480,000,000 bits per
second -- I don't know if the camera body firmware can be set to transfer
the image data over the USB connection in real time -- i.e. while the images
are being generated. 

You should check the data rates of high speed or 'professional' flash
memory cards -- discover the actual data rate in bits per second not just
the (marketing) numbers like 12X, ...

--
Martin

I think you're right, better to focus on data acquisition and storage at this stage. I'm not sure how clear an image a revolving mirror would reflect, especially at slower frame rates. I've thought about prisms to split the light, too, but then you're losing stops. What is the refresh rate limit of the CMOS? Thanks for the ideas MGD!

Attached Link: Flash memory speed

--
Martin