Document ID
APT-DS-CXP-HOST-0001
Revision
1.0
IP version
2.2.1
Release date
November 2026
Chapter 2: Functional Description

2.4Video bus

The video output is an AXI4-Stream in the AMD UG934 video convention, with nothing added and nothing left out:

  1. Pixel orderPixel 0, the first in time, sits in the least significant bits. Pixels are packed tight, with no padding between them.
  2. Component orderA pixel is its components packed tight, in the UG934 order of its layout.
  3. Component widthEvery component has VID_BPC bits on the bus. A narrower link depth is MSB-aligned with zero LSBs; a wider one loses its LSBs.
  4. Paddingtdata is padded with zeros at the top to a whole number of bytes; bits a format does not use are zero.
  5. Framingtuser[0] marks the first beat of a frame, tlast the last beat of every line.

Three parameters size the bus: VID_PPC (2 or 4 pixels per beat), VID_BPC (8 to 16 bits per component) and VID_CPP (1 to 3 components per pixel). The IP GUI derives VID_CPP and the tdata width from the formats you switch on.

Table 2-2.tdata width in bits by components per pixel (VID_CPP), pixels per beat (VID_PPC) and bits per component (VID_BPC). The IP GUI derives it from the format switches. Shaded: the default.
VID_CPPVID_PPCBPC 8BPC 10BPC 12BPC 14BPC 16
1
Mono, Bayer, Raw
21624243232
43240485664
2
adds YCbCr 4:2:2 and 4:2:0
23240485664
4648096112128
3
adds RGB and YCbCr 4:4:4
24864728896
496120144168192
Default build, 4 pixels x 16 bits: video_out_tdata[63:0], Bayer 12-bit (PIXEL_TYPE 0x0A)D3 [11:0]63051D2 [11:0]47035D1 [11:0]31019D0 [11:0]15030Build 2 / 8 / 3: video_out_tdata[47:0], RGB 8-bit (PIXEL_TYPE 0x18)R147B139G131R023B015G070Pixel 0, the first in time, sits in the least significant bits. Components narrower than VID_BPC are MSB-aligned with zero LSBs.tuser[0] marks the first beat of a frame, tlast the last beat of every line.
Figure 2-2.Two video beats: the default 64-bit build carrying Bayer 12-bit, and a 2 / 8 / 3 build carrying RGB 8-bit.

AMD video IP uses the same three numbers under other names (Samples per Clock, Maximum Data Width, Number of Video Components). With equal numbers on both sides the buses connect directly. Where an AMD IP has no 14-bit option, build the core with VID_BPC = 16 and the 14-bit data arrives MSB-aligned in 16.