Hello
This message deals with capturing video using a µP DaVinci Centaurus TMS320DM8148. We developed a board based on video digitalizer TVP5158 connected to a µP TMS320DM8148. Inputs are 4 analog PAL channels and these 4 channels are then line-muxed to the µP (digital 8-bit interface).
During the integration, we noticed that we had problems with the definition (blurred edges, stair step artifacts ...), so that it is not good enough for our product. Investigations led us to the fact that only even fields are captured.
Our test has been done with a test bench described here after :
- The first of the 4 analog inputs is connected to a BluRay Disc player, that is also capable of displaying (lossless) pictures on the screen taken from files via a USB stick.
- A picture with a suitable size (PAL-standard active video frame size = 720x576) has been drawn : one line out of every two is red, every other line is blue.
- Using the OMX interface on the Cortex A8 side (Linux O.S.), a chain including only the CAPTURE module is built. The callback Fill_Buffer_Done is used to copy the content of received buffers into files in the Linux File System).
After execution, we can check the content of the files and we can see that we got only red buffers with half the initial frame's size. If colours used for fields are inverted in the reference picture, we can see that captured buffers are then all blue.
Reading the FAQs, I noticed this item that seems quite similar to my own problem :
http://e2e.ti.com/support/dsp/davinci_digital_media_processors/f/716/p/204227/871411.aspx
A. LOG MESSAGES
First of all, I would like to know how I can also get the log messages of Cortex M3 that the integrator joined to the latter request : Tags [Cortex_M3_ISS] ?
[Cortex_M3_ISS] *** Capture Driver Advanced Statistics *** )
B. EXPECTED RESULTS
Secondly, reading the datasheet TI81xx-HDVPSS-UserGuide, and particularly the section dealing with the FVID2 frame format, it is still unclear to me the kind of buffers I should get. What should I expect :
- Twice as many calls to callback FBD, each returned buffer containing just one field ?
- As many calls to callback FBD, each returned buffer containing 2 fields of a whole frame identified by addr[x][y] ?
- The two previous assertions are possible, it is configurable ... ?
Reading carrefully this section, sentence N.3 seems right to me, but the way to manage this remains quite obscure. Could you please confirm and point out any helpful reference.
Thanks a lot in advance for your attention and forthcoming reply. Best regards.
Philippe