very good stuff Steve, we are only reading 1280x720 framesize so your saying we can do extra stuff with the lost lines of resolution to make the image less rolling shutter effected?

Obin,
Actually, I said something different. There are two types of lines: active lines and blanking lines. When you set the region of interest (ROI) to 1280x720, you are setting the active lines. They will clock out based on the pixel clock. Fast clock, fast readout, minimum RS artifacts.

After the active lines you have blanking lines. For high frame rate you try to minimize this. On the SI-1300, the minimum is 15 rows. No readout but integration occurs. The tricky part is to have the vertical blanking time = 1/48th sec so that frame you are talking about tossing right now is just integration time. Maximum potential motion blur.

We may be able to go even faster. The SI-1300 image degrades if we clock too fast but we have run them up to 67fps at 1280x720. This is 14.9msec per frame. 24fps is 41.67msec per frame so the blanking time could equal 26.75msec per frame. That is the goal. Fastest readout and overall exposure max and frame rate of 24 fps.

Don't worry too much about this. I have to work out what the registers get loaded with and I will pass it on to the Robs.

Oh, I see that is aweomse ! Pass it on asap so the Robs can deal with it in the software in an early version!


Rob did you get your camera yet??

It seems I can't understand this stuff very well...
Steve, did you say that a captured frame would have a 1/24s exposing time (1/24s motion blur)?
If yes, what is the reason for that?
Has something to be with a technical limitation?
I ask these questions because a movie frame at normal speed has an exposure of 1/48s, so a 1/48s motion blur...


P.S. Could someone here contact Xilinx people or post resources for FPGA programming and related stuff (tools, boards, prices,etc)?

Definitive motion blur test
 

Here is a way to test the camera for motion blur:
Set up a turn table, or a record player, with a white paper disk and mark on it a line. Take video of that. You will see the frame duration, and the blanking time when you look at the frames.
The line will be blurred for X number of degrees, and the amount of white space between two frames will show the period that the camera is not exposing at all. Can calculate Shutter angle, in film terms.
The rpm of the turntable must be known, obviously.
-Les

Juan,
The motion blur (exposure time) is adjustable from very short to the total frame time (active + vertical blanking) so yes, 1/24th can be programmed.

Xilinx has some free programming tools for smaller designs on their web site. For the record, this is NOT programming even if VHDL or Verilog look like it. This is synchronous hardware design. It requires a different way of thinking. On a clock edge, hundreds of events happen at exactly the same time, propogate and are stable until the next clock edge. Programmers are used to one event happening in time. Even multiple threads are sliced up and happen one at a time. This is not going to be a cakewalk if someone doesn't know hardware design.

Les: Nice test.

software for JPEG2000 to test:

http://www.ece.uvic.ca/~mdadams/jasper/#software

MIT license model

There is also the libj2k, but I-ve been told it is really slow....

http://www.j2000.org/

So ,Steve, does this mean that if I configure exposure to 1/24 then I could add a mechanical shutter to have 1/48 and eliminate the rolling shutter effect completely?

Steve: one other thing I was wondering. If we are running at
1280x720 I assumy we are using the middle 720 of 1024? Not
the top or bottom? (since most lenses are sharpest in the middle)

Juan:
I think you are correct. If you open a shutter during the extended vertical blanking period (1/48 sec), close it, all the pixels will expose for the same duration and at the same time. You read out during the other 1/48th and you don't care about the rolling because the shutter is closed. Give the man a cigar - we have a winner. I think this deserves a drum roll also.

Rob L:
Both the size and location of the window is programmable so yes, it can be centered. Much easier to point the camera also. (Pretty funny when you do a 320x240 window on a 3 Mpix sensor, use a wide angle lens and you can't figure out where the image is coming from).

Leads me to another thought. You have more vertical pixels. You could do image stabilization if you record 820 vertical and use the top and bottom 50 rows to debounce the image. It is free once the software is written.

Nope, not yet. But family life has been keeping me busy the last few days anyway, so I probably wouldn't have been able to do much.

wow...

http://www.taeym.be/new_page_18.htm

wow too...
for more info, see this:
http://www.isgchips.com/Templates/t_quadhdtv.htm

Unfortunately the price will probably be equally WOW. But it
sure is a neat camera. Why is "Star Wars" printed all over those
boards?

Camera arrived
 

The camera just arrived! Looks like I'll need to get a C mount lens though. Nothing came with it.

hey Rob want me to send U mine? I have to ask the crew here but I maybe could do it! i would send our 75mm OR you can jump on ebay NOW and order one with a buy-it-now and have them ship it fast! they are cheap as dirt on ebay

No, don't send yours. I'll buy one, or perhaps buy an adapter so I can use my 35mm SLR lenses.

Guys, don't strain your "wow" on that camera. JVC had one out using that sensor almost two years ago. We priced it recently. I can't remember but somewhere between $30K and $60K for the sensor in low volumes. Lots of NRE to make up on the camera sales so expect $60K-$100K cameras.

I may go buy 50 12mm used lenses and put one with every single camera sale. Sorry Rob, I thought you were covered. If you want one quick, try a security company that pulls old bank cameras. Most of them were c mount.

Zoicks! That would be nearly perfect sensor for a HDTV camera, though. I'll bet there will be more like it over the next few years at lower price points.No problem. I went ahead and bought one off eBay. I also have a friend (one of the founders of Ipix) who may have a lens I can borrow.

The Quad camera was being sold for $60K .
They had a special 8 megapixel LCD display that accepted the 4 DVI inputs for a 'viewfinder'. Like a picture window into reality!

-Les



<<<-- Originally posted by Rob Lohman : Unfortunately the price will probably be equally WOW. But it
sure is a neat camera. Why is "Star Wars" printed all over those
boards? -->>>

Hey, if you can afford the camera, use it with IBM's T221 9.2Mpix 22" LCD monitor. It is only $8.4K so it seems cheap next to the camera. To be fair, I've seen their 3.2Mpix monitor and they are truely awesome.

would a F-C mount adapter like this work?

http://cgi.ebay.com/ws/eBayISAPI.dll?ViewItem&category=30035&item=3822549795&rd=1

Yes, actually I'm planning to buy one of those pretty soon. It won't do me much good right now though; I have no F-mount lenses either.

I have a pretty nice (but slow) Canon EOS lens, but haven't seen any C-mount adapters for EOS. Nikon F-mount seems the way to go for using 35mm SLR lenses with a camera like this, especially if I can build a good GG (ground glass) adapter at some point.

C mount adapters:
Yes they work fine. You will get a narrow FOV because the c mount sensor is smaller. Edmund has them for $65 and bhphotovideo lists one at $29.95 out of stock though.

General Brand
Price : $ 29.95
Shipping Cost >
C-Mount Adapter for Nikon Lens

Mfr # VA304 • B&H # GBCMN

Great way to have excellent optics cheaply.

Yup, which a GG adapter will fix, and also give shallower depth of field. I wonder how practical (and expensive) it would be to design and produce a quantity of C-mount-to-F-mount GG adapters that would work directly with these cameras. That would be really nice.

does a 'narrow FOV' mean that it would negate the properties of a wide-angle lens?

If so you could just add a fisheye to get some of the FOV back, or would the image still be distorted?

That's correct. It would be like putting instamatic film into a 35mm camera -- only the center of the image would be used; the rest would be cut off.
Yes, you could use an extreme wide angle lens to try to get a "normal" FOV, but you would probably end up with a lot of distortion.

Rob:
I would test that distortion because it tends to be worse at the edges of the lens. You will be using the center of the lens. I don't have anything wider than about a 28mm in my Canon bag or I could do a test shot. But, you can get a 6mm c mount lens for $120 or so. If you aren't solving the DOF problem, as Obin found, c mount might be easier.

Thanks for the info! Using 35mm SLR lenses is more of a long-term idea right now. I just ordered a cheapo C-mount lens off eBay for this phase of the project.

Once the system gets to where it's usable I'll decide where I want to go with it. There is something very attractive about having access to a wide variety of high-quality F-mount lenses ...

Just to refresh, from a post within this thread.
About FPGA designs and camera


http://www.elphel.com/model313/index.html

Just a thought.Wouldn´t be useful to have a sticky thread with a compendum of all this technical things to be accessed in a simple way?

I mean chips, software tools, codecs, camera sensors, shutters, raid cards, source code, etc,etc.

@Nordhauser, could you give an idea of the pricing a sensor , Bayer pattern of 1920x1080 active pixels, 24x18 mm would have?
two, three more times the price? I mean only the sensor not the camera head....An IBIS4 1280x1024 costs around 1,000.
I think about same chips we have now (same design) but with bigger pixels...
Don´t know if small pixels but with a wider space between them would be cheaper (I think it would be a waste of space but know nothing about manufacturing procedures)

a 6mm c-mount lens... isn't the same as a 6mm f-mount? in terms of field of view? right? like a 6mm f-mount would be crazy fish-eye?

is there an easy conversion table?

<<<--

Don´t know if small pixels but with a wider space between them would be cheaper (I think it would be a waste of space but know nothing about manufacturing procedures) -->>>

Smaller pixels will give much more noise and less sensitivity. Look at the compact digital cameras and DSLRs side by side for same pixel count

I know,I know, but I said bigger sensor total area, with the same pixel area we have now.Clearer now?

i'm alittle confused by the siliconimaging website... is the SI-1300 camera you guys are using just the circuitboard with the little lens on it, or is it the black box with the connections?

It's the cute little black box.

can you guys tell me where in the thread is mentioned the cost of this cute camera?

or if possible, just quote it.
thanks,

filip

prime lens
 

hi rob

A lot of people like the Nikon primes: 24mm 2.0, 35mm 1.4, 50 mm 1.2, 85mm 1.4, 180mm 2.8 I cheaped out and bought some cannon lens for real cheap for the agus 35

Juan: we are using a Wiki to keep track of project status and
devices and whatnot...

http://www.obscuracam.com/wiki/wiki/

Since it is a Wiki anyone can easily add information / devices
that would be interesting now or in the immediate future.

Thank you very much Rob.

All good people:
I try to separate the sales world from my support of this group. In general, I would like any sales oriented requests to come to me privately.

Since there is a lot of clamor, I will toss out the SI-1300H-CL-S pricing - this is the Micron 1280x720x10bits @ up to 60fps for $2395 minus special offers I have made to this group. This includes the frame grabber, power supply, cables and XCAP-lite software.

Juan:
We do not manufacture sensors, only 'cute cameras'. The engineering costs and pre-production runs on sensors can run millions of $$.