P2 Tile Driver for PAL Composite Video?

Comments

• Rayman Posts: 16,262

  2026-04-02 16:17

  I've got a VGA tile driver for 1080p. Suppose that could be converted to composite...

  Usages is fairly simple, a bit like P1 tile driver.
  But, the driver itself is not simple...
• Wuerfel_21 Posts: 5,831

  2026-04-02 17:05 edited 2026-04-02 17:05

  @rogloh 's video driver can do text mode with PAL output if you want something that just works, but that one's very complicated.
  My VNG driver set (MegaVGA etc) also does PAL (60Hz variant only) and the code is slightly less complex, but it exclusively displays line buffers that another cog needs to be rendering.

  Other than that I don't think anyone did PAL.

  Putting together a basic driver yourself is only slightly more complex than e.g. VGA. (I think I posted an example of that ages ago? Anyways it's like 5 minutes trivial)

  For the simplest case of non-interlaced non-color, you just need to generate the same old ...->active->front porch->hsync->back porch->.. sequence for regular lines and a long sync pulse for the VSync lines. (though you technically can't call that PAL at all, it'll work). Just need to get the timing right.

  Color requires splitting the back porch into 3 segments: the breezeway, color burst and the remaining back porch that doesn't have a catchy name. The hardware does the modulation for you, you just need to send the right "color" during burst (see below). For PAL (as opposed to NTSC) you also need to alternate the CY register between two values. Slightly tricky because CSC register are NOT buffered like the streamer commands are, so you need to change it at a time where you know that you're in blanking. I recommend doing it shortly after putting the HSync pulse command into the streamer (this ensures the streamer is in front porch)

  These are the CY/CI/CQ register settings I use for NTSC and PAL (slightly cleaned up for clarity). You can't actually reach the correct levels for the summed luma+chroma signal with the 75Ω DAC mode (P_DAC_75R_2V), so these are designed for use with P_DAC_124R_3V mode.
  This is partially result of trial and error to get good colors on a variety of monitors, but everyone else's values (especially on the NTSC side) just look worse or cause issues. And you all know my perfectionism...
  (For monochrome only, you'd simply set the IQSCALE constant to zero, then it'll not bother you)

  CON  ' CSC constant calculation
  
      ONE_IRE = 255.0 / 180.0
  
      BLANKING_LEVEL = ONE_IRE*40.0
      PEDESTAL = 0.0 'ONE_IRE*7.5 ' NTSC only, other standards (including NTSC-J) don't have pedestal
  
      BLANK_LEVEL = round(BLANKING_LEVEL)
  
      B2W_NTSC = ONE_IRE*100.0 - PEDESTAL
      YSCALE_NTSC = 127.0*(B2W_NTSC/255.0)
      IQSCALE_NTSC = YSCALE_NTSC * 1.33
  
      B2W_PAL = ONE_IRE*100.0
      YSCALE_PAL = 127.0*(B2W_PAL/255.0)
      IQSCALE_PAL = YSCALE_PAL * 1.33 ' Can't set correct amplitude (around 1.47) because weird overflow?
  
      CY_NTSC      = (round(0.299*YSCALE_NTSC)&$FF)<<24 + (round(0.587*YSCALE_NTSC)&$FF)<<16 + (round(0.114*YSCALE_NTSC)&$FF)<<8 + round(BLANKING_LEVEL+PEDESTAL)
      CI_NTSC      = (round(+0.5959*IQSCALE_NTSC)&$FF)<<24 + (round(-0.2746*IQSCALE_NTSC)&$FF)<<16 + (round(-0.3216*IQSCALE_NTSC)&$FF)<<8 + round(BLANKING_LEVEL)
      CQ_NTSC      = (round(+0.2115*IQSCALE_NTSC)&$FF)<<24 + (round(-0.5227*IQSCALE_NTSC)&$FF)<<16 + (round(+0.3112*IQSCALE_NTSC)&$FF)<<8 + 128
  
      CY_PAL      = (round(0.299*YSCALE_PAL)&$FF)<<24 + (round(0.587*YSCALE_PAL)&$FF)<<16 + (round(0.114*YSCALE_PAL)&$FF)<<8 + round(BLANKING_LEVEL)
      CI_PAL      = (round(-0.147*IQSCALE_PAL)&$FF)<<24 + (round(-0.289*IQSCALE_PAL)&$FF)<<16 + (round(+0.436*IQSCALE_PAL)&$FF)<<8 + round(BLANKING_LEVEL)
      CQ_PAL_ODD  = (round(+0.615*IQSCALE_PAL)&$FF)<<24 + (round(-0.515*IQSCALE_PAL)&$FF)<<16 + (round(-0.100*IQSCALE_PAL)&$FF)<<8 + 128
      CQ_PAL_EVEN = (round(-0.615*IQSCALE_PAL)&$FF)<<24 + (round(+0.515*IQSCALE_PAL)&$FF)<<16 + (round(+0.100*IQSCALE_PAL)&$FF)<<8 + 128
  
  
      COLOR_BURST_NTSC  = $7C9900_01
      COLOR_BURST_PAL   = $D8A351_01
  
• Wuerfel_21 Posts: 5,831

  2026-04-03 01:15

  @Wuerfel_21 said:
  Color requires splitting the back porch into 3 segments: the breezeway, color burst and the remaining back porch that doesn't have a catchy name.

  To perhaps clarify and because I was looking through some old files and found this one again: The breezeway is the bit of blanking that goes between the sync pulse (left edge) and the color burst.

  
• Christof Eb. Posts: 1,548

  2026-04-03 08:55

  Thank you, @Wuerfel_21 and @Rayman, for your inputs.

  Up to now I have never understood, how colour is coded into a video signal. I have a very rough idea, that it is sent at another time than brightness.
  I have used and modified @ersmith 's VGA tile driver. I think, I was able to understand that. So if I restrict this to BW brightness only it must be possible somehow to use only one DAC and to mix in the sync pulses?

  Hm, now I had a look into some PDF about my monitor, which is a Hyundai L17T. If I understand that right, that particular monitor can also process NTSC.

  OK, found this https://forums.parallax.com/discussion/177879/composite-video-driver and in there @rogloh 's driver. Yes, that is not, what I would call simple....
• Wuerfel_21 Posts: 5,831

  2026-04-03 12:02

  @"Christof Eb." said:
  Up to now I have never understood, how colour is coded into a video signal. I have a very rough idea, that it is sent at another time than brightness.
  I have used and modified @ersmith 's VGA tile driver. I think, I was able to understand that. So if I restrict this to BW brightness only it must be possible somehow to use only one DAC and to mix in the sync pulses?

  >

  The composite video signal does everything all at once. Sync pulses are just negative pulses between the active video (this is why black/blank level is at ~0.3V, so the sync pulse can go down to 0 V). Color is just a 3.58 / 4.43 MHz (NTSC/PAL) carrier wave that rides on top of the brightness. The amplitude determines color saturation and the phase determines the hue. This alternatively expressed as quadrature modulation (where I/Q (or U/V) are the carthesian equivalents to polar hue/saturation). As mentioned, the P2 hardware does it for you.
• Christof Eb. Posts: 1,548

  2026-04-04 06:10

  Aha! Thank You, this is very enlighting for me!