[HN Gopher] RK3588 - Implementing a Vectorscope for processing v...
       ___________________________________________________________________
        
       RK3588 - Implementing a Vectorscope for processing video in real
       time
        
       Author : zdw
       Score  : 50 points
       Date   : 2025-05-02 16:05 UTC (4 days ago)
        
 (HTM) web link (jas-hacks.blogspot.com)
 (TXT) w3m dump (jas-hacks.blogspot.com)
        
       | kamma4434 wrote:
       | I wonder why you should have a vectorscope in real time if the
       | point is showing histograms to humans.
       | 
       | As the image does not change much frame to frame, I presume that
       | if you compute it and display it every fifth frame or so, nobody
       | would ever notice.
        
         | dragontamer wrote:
         | Well it's clearly some kind of strange requirement to require a
         | relatively low power and specifically the RK3588 instead of any
         | other chip or even Desktop chip.
         | 
         | It's a lot of work for some processor likely using less than 5W
         | of power. Impressive for sure but now I'm curious what
         | application this is for in general.
         | 
         | Who needs to do this but doesn't want to use a far more
         | powerful cell phone or laptop processor?
         | 
         | -------
         | 
         | Not to hate on the work at all! It's clearly a lot of effort to
         | get this to run on such a relatively small chip.
         | 
         | Maybe it's just exercise to learn how to use OpenGL on such a
         | small platform for GPGPU compute? Might be good reason enough
         | to try to accomplish??
        
           | NewJazz wrote:
           | Rk3588 was originally designed with smart screens as one
           | possible use case. That's why it has an HDMI input on board
           | (and several video outputs).
        
           | luma wrote:
           | If the chip is getting the job done and it's a fixed-function
           | device, why would one want to spec a more expensive, more
           | power hungry solution in its place?
           | 
           | I don't think "minimum required hardware" is a "strange
           | requirement", in fact it's essentially ALWAYS the requirement
           | when developing embedded solutions.
        
           | shadowpho wrote:
           | This is probably development for a tool to light up LED to
           | match up with input video. So cost is a big concern and
           | Rk3588 is already one of the beefier/expensiver options
        
           | Aurornis wrote:
           | > Well it's clearly some kind of strange requirement to
           | require a relatively low power and specifically the RK3588
           | instead of any other chip or even Desktop chip.
           | 
           | RK3588 has 4K @ 60Hz HDMI input built in to the chip. Install
           | the Linux distribution, attach HDMI cable, and you can pipe
           | HDMI input into v4l2 or any other destination. No messing
           | with capture cards, powered USB hubs, or 3rd-party drivers.
           | 
           | > Who needs to do this but doesn't want to use a far more
           | powerful cell phone or laptop processor?
           | 
           | It's a fast chip. Fast enough. Convenience of the small all-
           | in-one solution is more important than using the fastest chip
           | you can find.
        
         | dylan604 wrote:
         | wait, are you saying that someone that was trained on how to
         | use a vectorscope would not notice the image not changing frame
         | by frame in real time? that is absolutely ludicrous. i grew up
         | using waveforms/vectorscopes/audio phase meters, and seeing 1/5
         | of the data would drive me crazy. that would be like watching
         | streaming video with buffering problems. we no longer use
         | RealMedia for a reason
        
       ___________________________________________________________________
       (page generated 2025-05-06 23:01 UTC)