This thread has been locked.

If you have a related question, please click the "Ask a related question" button in the top right corner. The newly created question will be automatically linked to this question.

DM8168 Video Input Format

Is there a way of inputting two channels of video either using RGB or YUV 4:4:4 ?

I can see that VIN[0] can accept an RGB 8:8:8 input, but the other port (VIN[1]) has only got 16 data bits and seems only able to take YUV 4:2:2.

In our application the chrominance bandwidth must be the same as the luminance.

Thank you.

  • Hi,

    It is not possible on TI816x, but it is possible on TI814x.

    Regards,

    Brijesh Jadav

  • Thanks.

    Do you think the DM8148 could H.264 encode two 1080p30 video channels?

  • Yes it is possible to encode 2 ch 1080p30. Note that you would have to clock IVA greater than 420 Mhz to be able to do 2 ch 1080P30 so you will have to select the appropriate 814x device variant capable of clocking IVAHD at high frequency. As you are aware encoder Mhz also depends on the configuration. Pls refer the H264 encoder data sheet for details on Mhz for various configurations.If you dont have access to the TI H264 IVAHD encoder package pls contact your local TI FAE to get the datasheet.The encoder however only supports encoding in 420SP format.

  • Hi,

    It will be better to judge if you let us know your end application. Like what it is trying to do.

  • Hi

    We adopted the 8168 over 2 years ago for a design that receives two video channels (DVI/VGA) at resolutions up to 2560x1600 and 2k x 2k. We know there are hard limits to what the VIP can accept and always believed it was 1920 pixels horizontally. I have read that this limit might be 2k instead.

    Our video sources are 24-bit RGB and we send these to the VIP as 16-bit values in YUV 4:2:2 after conversion using a small FPGA. This all works nicely and we can H.264 compress the inputs (and stream them over Ethernet) using your omx.ti.videnc for standard 1080p resolutions.

    Using 4:2:2 we lose some chrominance detail and end up having minor color fringing (noticeable on high contrast boundaries).

    I know we can input RGB 5:6:5 as RAW data to the VIP and this will (and does) end up in memory, but is this beneficial to the encoder? Can the encoder accept an RGB input? If it has to be YUV can it accept a YUV 4:4:4 format? Would we have to write an RGB 5:6:5 to YUV 4:4:4 conversion process or is there something already available that does this?

    I'm asking these questions from a hardware engineer point of view as our software has been (as is being) done by a 3rd party. Thanks for any answers/feedback on these questions.