Open Source MIPI CSI-2 Camera Receiver on FPGA, IMX219 Raspberry PI Camera USB 3.0 Stream

[@qzorn4440]

wow soo much fantastic CSI-2 video information. 😎 Thank you.

[@Michael-d8f2p]

Is there a way I may contact you directly?

[@anandmoon5701]

Well explained super cool stuff.

[@robert_trirop]

Awesome :)!

[@muhammadshahzaib3813]

nice video and neat stuff. keep it up.

[@sagarwalishetti1176]

This is really nice!!

[@dhaneshdprabhu8482]

Nice vidoe and well explained. How much did it cost for Lattice chip and how much time did it take to make the development board of lattice IC

[@OrenCollaco97]

Hi did you implement the debayering for this yourself? I have been trying to use the V4L2 API to get the raw bayer frame (which I'm able to) but can't seem to get the debayering right. I've tried doing cv::cvtColor(bayer_frame, rgb_frame, cv::COLOR_BayerRG2RGB); after bitshifting for RG10 but seems to get artefacts I can't figure out. I'd appreciate if you could point me to the theory of your pipeline. Nice work btw.

[@OrenCollaco97]

Nvm, you have a GitHub repo for this.

[@circuitvalley]

Debayer in basic form is relatively easy, You can read about it on my site www.circuitvalley.com

[@veeshal100]

Hi, I have a somewhat similar design working on the Lattice Crosslink FPGA(4 Lane MIPI CSI-2 Receiver on FPGA). It is basically based on Lattice's reference design "FPGA-RD-02060-1-1-MIPI DSI_CSI2-OpenLDI-LVDS-Int-Bridge". But the problem we are facing is that the colors in the video looks corrupted/deformed. Can you help me in understanding what can be the real issue here? I can share you some snippets of the video if I can have your mail id. Thanks!

[@browaruspierogus2182]

How fast and how many bits is that ADC IMX219 cannot do more than HD/30fps. Try some crazy 4K sensors with 60fps if you can

[@circuitvalley]

IMX219 Can do way faster than 30 FPS , i have tested 1080p @ 60 FPS, 720p @ 120 FPS, VGA @ 200 FPS, or even 1000FPS at really low resolution , because of limitation of the interface @~4K resolution it can do 15FPS , Here is the test video showing raspberry pi camera going up to 1000 FPS kzbin.info/www/bejne/q4PEeYulYr-ob5I

[@browaruspierogus2182]

@@circuitvalley What do you think is good sensor for 4K@60 also what would be best fpga for it

[@circuitvalley]

@@browaruspierogus2182 There are many new sensors from many different vendors, but problem is may be one way or another you can get hold of the sensor but documentation will not be available. If i need to find a sensor 4K@60 FPS i would look for any consumer electronic camera and try to reverse engineer the protocol. Choice of FPGA Depend on what you trying to do , if it is a Bridge application than Lattice Cross link may do or Lattice ECP5 , Xilinx Zynq is also a good candidate if need to do some processing on SoC.

[@JackToth]

CircuitValley i’m really curious abut the max capabilities of the sony senor in the Pi Camera HQ

[@circuitvalley]

@@JackToth I am working on Interfacing IMX477 to Crosslink NX as well. Check back in few weeks

[@Zeropadd]

💗

[@maikdean8490]

Hi bother I have Samsung module 108 megapixel sensor can I take 2 of them to shoot a single video stream making high resolution?

[@circuitvalley]

If i understood you correctly you are looking for video stitching. Yes that is possible but this exact FPGA is bit too small for that.

[@maikdean8490]

CircuitValley I have Samsung 108 megapixel camera module 2 of them I want to combine the resolution to shoot one video stream can you please help me it’s for astronomy

[@maikdean8490]

CircuitValley because video stitching I don’t know what that means can you please explain more like connecting the pins together?

[@circuitvalley]

@@maikdean8490 I don't have direct solution in my hand, i have too look around. combining output of two image sensor into a one single seamless frame is normally called video stitching. With stitching final image will appears as if its take with one camera. its pretty complex thing to implement. other simpler solution is adjust merge output of both sensors side by side . it very easy compared to stitching. If you have no experience with FPGA even simpler solution is hard to implement. I have no idea how to guide you further from here.

[@maikdean8490]

CircuitValley thank you man for your hard work I do astronomy if you see what I see you won’t believe it but I need high megapixel Count to see mars clouds and also moons rain just try to help me I have 3 108 megapixel Samsung module or GoPro sensor it shoots in 4K if I took 3 of them it’s 9k resolution

[@Fernandoc24]

Where is the SOURCE?

[@circuitvalley]

Here on GitHub github.com/circuitvalley/mipi_csi_receiver_FPGA

[@BuzZ.]

Why do I always end up here?