HDMI 1.4/2.0 Dual/Quad pixel data format for AXI-Streaming

[jdehaven]

Tables 3-6 and 3-7 in PG236 show the data format for 420 YUV 8-bit video in AXI Streaming Dual and Quad pixels-per clock scenarios.  A customer claims that the Quad pixel per clock figure (3.6) is actually showing 8 pixels instead of 4, and the Dual pixel per clock figure (3-7) is actually showing 4 pixels instead of 2.

The following two figures, 3-8 and 3-9 show yuv420 data in both native and AXI Streaming formats (Dual pixel mode).  Both of these figures correctly show that there are there are only two pixels per clock , not 4 as in figure 3.7.

Can you please confirm that figures 3.6 and 3.7 are incorrect?  Attached are screenshots of the figures in question.

[samk]

We are looking at this internally and will close this thread once we have confirmation.

[xud · Accepted solution]

@jdehaven 

In order to make line 0 in AXI4 Stream contains 2 Y samples per clock, and line 1 only contains Y samples, the native interface should be :

• 2PPC case, YUV420 has 4 Y components in every clock. When it is remapped to AXI4-Stream, then each clocks got 2 Y components.
• 4PPC case, YUV420 has 8 Y components in every clock. When it is remapped to AXI4-Stream, then each clocks got 4 Y components.

The image belows shows how 2ppc data looks in native interface, and AXI4 Stream interface. 4ppc should be similar.Also HDMI cable natively contains two pixels per clock in TMDS data. When it maps to our IP 2ppc native interface, two TMDS clock cycles of pixels (eg. Pixel 00/01/02/03) are mapped to native interface.

[davefrazer]

It then seems that figure 3.9 is incorrect, and following the TMDS transmission we should get one chroma component on one line, and the other component on the next line.  The illustration shows both chroma components on one line, and none on the next.