Using the sensors in consumer megapixel cameras...and other fantasies.

[Matthew Miller]

If it's possible to tap into the DVX100 and effectively aquire an image before it is processed into DV, what else could you tap into?
Both my 5MP and 8MP cameras from sony are able to record mpeg video, the first at 320x240 and the second at 640x480. The framerate for both seems to be a choppy 15-25fps. Mpeg video is really ugly stuff, but the sensor probably sent a clean image to whatever chip inside it did the transcoding to Mpeg.

At their highest resolutions, each camera lets you take a burst of 3 shots in rapid succession, and then a good ten seconds to compress to Jpeg and save onto the tiny little memory stick.
As we all know, these cameras are designed to dump the images or video onto relatively small and slow media. The image and video processors inside probably aren't the worlds fastest or best because they don't need to be. But how fast can those sensors really run?
Now I'm not talking about 2560x1920@30fps. I'm sure 1280x720@24 is probably asking too much from either sensor.
But it is clear that the sensors in both these cameras can sample lower resolutions at the same optical size.

Is this insanity? What about all those new hybrid cameras that do DV video and megapixel still photos. Canon has one or two models that actually uses the entire width of the chip when you shoot anamorphic widescreen DV, because the sensor inside is so close to a 16x9 format.

Are any of these cameras possible candidates for aquiring HD or quasi-HD resolution images straight to a capable computer.

Here's hoping you all feel that the only stupid questions are the ones you never ask.

[Jed Williamson]

some cameras like the canon 10d run DOS for an operating system.

I suppose someone could hack this & load a version of linux or something.

Some guy offered a big reward to hack the xbox a while ago.

6 megapixel SLR digital camera & lenses:

$1,500

P4 Laptop w/7200 rpm HD:

$1,500

720P at 16x9 24P for under $3,000:

Priceless!

[David Warrilow]

Hi,

I recently wondered the same thing. I was using a Canon EOS that was firewire enabled. Some of these still cams have a 35mm sensor as far as I know, that would simultaneously solve some of the "shallow DOF quests" that alot on this board have been pursuing. But I guess it all comes down to what you mentioned - scan rates - can they do HD pixel values @24 or 25 fps? The canon I was using dumped out RAW imanges - that obviously needed correction - but had a great deal more lattitude to fix things up than the Jpegs it fired off at lower resolutions. Now the files were HUGE like 7mb a frame - so @ 25fps - 175mb/s if it could even do that (which I doubt) is not really feesable. But what about some other (smaller) file storage arrangement - like on the DVX thread - Tiff @10 bit or whatever.

I know with some DV cams you can do a 'pass through' - run the cam in still mode and then firewire direct to disc - you get a progressive type image using the full ccd - but it's still 720x576. I tried the same thing with the Canon but couldn't seem to get it to happen i.e. live video over firewire off the chip.

I'm just an infant when it comes to the technical nuances of these things but it seems a digital SLR is so close...and yet so far.

Best,

DW.

Best,

DW.

[Valeriu Campan]

The recently released Canon EOS1 Mark2 is capable to shoot bursts of 40 frammes @ 14fps for a 8Mpixel (3504x2336 pixel) file size before it stops to write the files to a compact flash card. The sensor size is 28.7x19.1mm, a little larger than the full Academy 35mm.
Wonder what happens if you shoot a 1728x1152 or 2544x1696 (JPEG quality 8) file tethered via firewire to a laptop. I don't have access to one of these, but will be interesting to see if anybody can try it.

[Anhar Miah]

I have been thinking in the same lines as David, i think you would have to open up the SLR and built some kind of interface to bypass the cameras DSP to a PC then anyting is possible (within the limits of the CCD/CMOS performance)

But you need a lot more info about the CCDs/CMOSs first, try and find the Datasheet of the sensors, they should contain specs and pin out operation descriptions.


P.S i know that this kind of idea has popped up many times before, its just that is is very difficult to do that not many people can pursue (needs money + alota Skills)

[Matthew Miller]

Alright. It looks like I'm gonna be taking some things apart and doing some heavy reading. I understand electronics well, but I'll definitely be getting into new territory with trying to modify a camera on this kind of level. I build computers on a regular basis as a source of income and because it's a hell of a hobby, and I'm fearless about taking apart equipment and putting it back together.
None of the data I've found so far for these cameras talks about the 'theoretical' limitations that I want to know about. I can't find anything on what the sensors are clocked at or what kind of speed the signal processor is capable of.
I guess me next step will be taking a look inside the equipment in question to see what would be even remotely feasible. I have a few digital cameras at my disposal... if I can convince my girlfriend to let me take apart the one I got her for christmas (insert evil laughter).

With some of these USB2.0 enabled cameras, I wonder if you could make it a matter of hacking the firmware to send a 960x720 or some other low resolution image into the computer at a modest rate of speed.

I realize there is a good chance I won't get very far with this. In the end, I'm just looking for lossless 1280x720@24FPS like every other 'normal' guy out there that doesn't have 60k to spend but wants to be producing high definition right now. I'll probably end up going with one of these industrial imaging solutions like Obin is using. Altasens looks awesome.

[Matthew Miller]

Oh yeah, I had one other ridiculous idea that I'm sure has been brought up here before and would never work. What about using multiple single-CCD(or CMOS) DV camera sensors, aligned side by side to be the size of 35mm film. No wait... that's stupid.
You'd need way too many sensors, probably eight or nine. And there probably aren't many sensors that could be aligned edge to edge. Price would be retarded.

It's one thing to go the route of the Agus35 and let the image from a 35mm lens fall on a ground glass from which the image is recorded by the DV camera, preserving that wonderful DOF.
If you used a medium format or even large format lens, the image on the ground glass would be over twice the size, so you could then use two smaller DV cameras and make each record only some 60% of the projected image. Depending on the orientation of the two cameras you would get roughly a 1440x480 or 960x720 image (bear in mind the overlap). If you could pack three small DV cameras tightly enough in a row, you could get 1440x720 or an ultra retarded 2160x480 (who would ever want that).
I'm talking about these ridiculously thin DV cameras that are out there but still have good optics.
Obviously a large format lens that is meant to throw an image onto giant film would be best for this kind of experiment. 4 tiny DV cameras would give you 1440x960.

[Anhar Miah]

Why not try to reverse engineer first to begin with on a less expensive $80 type cheapo 1 megapixel (or even two) camera and then if you can get that to work move onto the full 35mm cmos sensor?

But here is the rough guideline:

(1) Get donor Camera ( heh heh)

(2) disassemble (carefully without damage to internal compoents)

(3) Gather as much info on sensors and subsystem chips (eg Data sheet, etc. even contact the manufactures)

(4) One you know what pins do what, what clock speed you need and any other requiremenrs are filled then try to get a "live feed" into the computer.

(5) I you have got past number (4) then you have done most of the hard part

(6) Find OR program yourself a capture software that useses lossless/lossy codec depended on users requirements.

(7) take expeiance onto 35mm cmos!

[Wayne Morellini]

<<<-- Originally posted by Jed Williamson : some cameras like the canon 10d run DOS for an operating system. -->>>

Good news.

Cameras use there own OS's, I know of one that is re-programmable, the Digita OS.

http://www.flashpoint.com/home.html

http://digita.mame.net

So this is two, the new Canon XL2 has a SDK aswell.


Thanks

Wayne.

[Jason Keenan]

I've been thinking about the same sort of things, although I don't have anywhere near the right amount of knowledge yet.

My initial questions would be.

---Can you alter the burst speed of digital cameras?

The speeds on different cameras seem to go from slow eg 3fps up to about 10fps. At 3fps you'd need 8 cameras synced to get 24fps but does 3fps mean 3fps or is it an approximate figure. In conversion of film to PAL, the film is sped up to 25fps rather than pulldown for NTSC so there is some flexibility (re changing speed of finished product without being able to tell) but real 24fps would be the ideal obviously. Here is an example of the Nikon D70

Capture Modes: 1) Single, 2) Continuous H (3 fps; up to 3 frames), 3) Continuous L (1.5 fps), 4) Multi-Shot 16 (consecutive 16 frames at 3 fps), 5) Ultra High-speed Continuous (QVGA-size images; 30 fps; number of frames selectable, up to 100 frames), 6) Movie with audio (QVGA-size images; 15 fps; up to 60 sec.)

---What is the burst speed dependant on?

Is it compression speed on a digital camera or is it processor speed and bandwidth. It can obviously push the uncompressed image to the compression unit at at least the burst rate so bandwidth shouldn't be a problem.

---How would the shutter hold up?

Not sure what the shutter is on a reasonable digital camera but how would it hold up if it was constantly in use.

---How does the price hold up?

For a mid range option, lets look at the Fuji Finepix S7000Z. It shoots at 5fps approx. with 6.1mp effective. We're talking approx $880.00 AUD x 5 = $4400.00 if you shoot at 25fps. It shoots bursts of 10 at a time though so unsure if it could be hacked to shoot continuous. If you could deal with 3.1mp you could go down a notch to just over $2,600 just for the cameras.

To keep the cost down, you would probably gather the frames then merge them later using software, which is going to be less of an engineering nightmare than going the dedicated hardware route.

--how do you get it to work?

Do you use a rotating prism that locks at each camera? 25 stops and starts that are perfect could take some serious engineering. Requires some thought. Could you make some sort of multifaceted prism that stays put?



When all is said and done, the reality is that what we want at the end is something that works that you could put together yourself or at least get the bits manufactured to spec reasonably cheaply. Keeping the cameras intact would help just so that you can resell or upgrade if you want to. You also want it to work perfectly at least 95% of the time and with a bit of regular tweaking, 100%.

When you start to increase the number of working parts and push them to exceed their capabilities (like forcing continuous shooting), there are going to be issues. Are they worth it? If you are a geek (like me) then even the thought is fun. If you have a hacker mentality, which it would seem most here do, then the challenge + circumventing horrendous prices is worth the effort.

Raavin

[Wayne Morellini]

I can't give you many detailed answers, but the restriction canbe any where from the slow outputs (fast ones are more expensive) of the sensor chip, compressor, processor, to the storage card.

The good news is that some chips allow you to window a smaller area, or even combine pixels into a single pixel, and output that data at a faster rate according to the ratio of the reduction in pixels. This means that the best megapixel camera might do resolutions above standard, and some other standard resolutions.

The other good news is that things like usb2.0 or firewire can transfer raw data to a disk recording system (but where is the microphone??). The bad news is that it might not be built for maxium speed, the compressor might not be able to keep up, the processor might not be able to take push out enough frames to the interface, and the disk on the other end might not be able to take that much (eek is right), sub2.0 is the worst in this respect . If the compressor can't keep up and you can't reprogram it to do less compression to speed it up, then you will have to output raw (uncompressed big files) frames.

So this restricts you to a maxium of 2Mpixel images off of a single chip camera (that canbe bigger than that), but realistically you should not expect much luck with that (because of USB2.0), 1 megapixel is more likely. But you might not even find a camera that can even do 1 megpixel, but you should be able to get DV res.

So I'm not saying don't try, but just warning you.

You might like to check out our custom HD camera threads, we will probably get better results (depending on the camera releases).

General info:
http://www.dvinfo.net/conf/showthrea...threadid=28779

http://www.dvinfo.net/conf/showthrea...threadid=28781

The actual threads are listed on the first posts.


Thanks

Wayne.

[Gints Klimanis]

A Nikon D70 hacker's group exists. The Nikon capture software can trigger the camera and transfer images at 3 fps (haven't tried this yet, so I'm not sure if this is RAW or high rez JPEG).
With mirror lockup and using the LCD as a viewfinder, we could fantasize about high rate, high resolution video. However, this
particular camera has a USB2 "compatible" interface that actually
runs at USB1 speeds. What's up with the deliberate crippling of this interface?

Nikon isn't in the camcorder video business, but I'd be willing to pay extra for a "high speed" video package that would deliver images over firewire or to flash card. The D70 flash card interface bandwidth is something between 5-6 MBytes/sec, which is about
double the DV25 datarate. I would definitely pay for 60 fps
frame mode video at current DV resolution.

[Jed Williamson]

How about this camera/webcam: Logitech 820

http://www.logitech.com/index.cfm/pr...=6793,detail=1

Out of the box the specs say it can do video at 640x480 at 25fps. or it can take 1600x1200 stills.

I have no idea how to start modifying this but if somebody would say its possible I might try my hand.

amazon has it for $149

there are some sample photos on this site.

www.softmag.no/logitech_clicksmart_820.htm

[Wayne Morellini]

<<<-- Originally posted by Gints Klimanis : A Nikon D70 hacker's group exists...
particular camera has a USB2 "compatible" interface that actually runs at USB1 speeds. What's up with the deliberate crippling of this interface? -->>>

There are a few possible reasons:

A. It cost more to go faster, everything main component has to be faster. Though eventually the lowest cost components may be fast enough to do video.

B. It maybe to stop us from doing real video/HD. Though higher FPS sensor designs cost a lot more money.

C. I forget.

Have they tried windowing, or pixel combining, to speed up the frame rate (these may not be available on the camera)?

Thanks

Wayne.

[Jeff Donald]

Nikon did not want the D70 to equal or approach the capabilities of their D1 series cameras which employ a FireWire interface. This is a way of providing product distinction. The D70 is considered a consumer camera, although many pros are using it.