On 11/07/2017 11:35 PM, Carlos Rafael Giani wrote: > I was discussing this with Tim earlier, and there was a little > confusion. He meant V4L2_FIELD_SEQ_TB/BT, not V4L2_FIELD_INTERLACED. > > When I tried out 1080i50 with HDMI in and the i.MX6, I got something > that looks like V4L2_FIELD_SEQ_TB. The question though is: are the field > rows always arranged like this with interlaced HDMI? Or does this depend > on the source? Could it for example be possible that 2 HDMI cameras > deliver both interlaced video, but the first one delivers > V4L2_FIELD_INTERLACED, and the second delivers V4L2_FIELD_SEQ_TB? Could > this happen? HDMI doesn't deliver any of the V4L2_FIELD_* formats, those are a property of the DMA engine and/or possible processing blocks earlier in the video pipeline. Interlaced formats as transmitted over HDMI are sent as a sequence of fields. All defined interlaced formats except for one send the top field first, then the bottom field. The exception (I'm 99% certain of that) is 720x480i (aka NTSC) which sends the bottom field before the top field. The video pipeline will typically DMA each field in a separate buffer (i.e. FIELD_ALTERNATE) or concatenate two fields in one buffer (SEQ_TB or SEQ_BT for 720x480i), although the latter is less common. There is little point to it since it would just increase the latency with little benefit. If the video pipeline contains a deinterlacer, then FIELD_NONE is also an option. Regards, Hans > > > On 2017-11-07 19:40, Tim Harvey wrote: >> Greetings, >> >> I'm trying to understand the various field orders supported by v4l2 >> [1]. Do HDMI sources always use V4L2_FIELD_INTERLACED or can they >> support alternate modes as well? >> >> Regards, >> >> Tim >> >> [1] - https://www.linuxtv.org/downloads/legacy/video4linux/API/V4L2_API/spec/ch03s06.html >
