EAN-4000-OEM-MIPI-Cameras
© SightLine Applications, Inc.
13
The 4000-OEM uses the MIPI as a transport layer. The data packets are received by the processor and
treated as bytes. These bytes are then assumed to be in the format set in the acquisition parameters
persisted on the OEM. The next describe the MIPI packet format used to transport video data from
different supported cameras.
All MIPI data elements are 8-bits. Since multiple pixel values are grouped together to create data
that is divisible by 8, 10-bit color video will have a group containing the upper 8-bits of 4 pixels,
followed by a single 8-bit value that contains the lower 2-bits of the 4 pixel group. This results in
5x8-bit values to represent 4x10 bit pixels. SightLine camera support uses a full 16-bits (into 2x8) for
camera pixel values that are 14-bits. This makes the encoding simpler and there is enough
bandwidth in the MIPI bus.
B1 MIPI Packet Format
Table B1: MIPI Packet Formats
Camera Format
MIPI Type
ID,Lanes,Gear
Pkt Width
Capture Setting Note
14/16-bit Grayscale YUV422_8
0x1E, 2, 8
x2
Gray, bits=16
IR Camera
8-bit Grayscale
YUV422_8
0x1E, 2, 8
x1
Gray, bits=8
IR Camera
16-bit YCbCr (HD)
YUV422_8
0x1E, 2, 8
x2
YUV, bits=16
BT.1120
8-bit YCbCr (SD)
YUV422_8
0x1E, 2, 8
x2
YUV, bits=8
BT.656, NTSC/PAL
8-bit Bayer
YUV422_8
0x1E, 2, 8
x1
Bayer, bits=8
AGS720P Color
Data conversion for current (non MIPI) cameras:
•
YUV422_8 packet. The packet width is optionally doubled to support 2x8-bit (16-bit data).
•
16-bit data order is low byte, then high byte (little endian). We recommend providing for byte-
swap in custom FPGAs.
•
Number of MIPI Tx Lanes = 2 (4-lanes for high resolution cameras coming soon).
•
The data packets are received by the OEM processor and treated as bytes. They are then assumed
to be in the format set in the acquisition parameters persisted on the OEM (G16, YCbCr16,G8).
•
The pixel width entered in the acquisition parameters setting is not doubled but is inferred by the
other capture parameters (bits=16/8).
•
The number of rows is set by the number of packets sent between FS (Frame Start) and FE (Frame
End). Each packet is a row of data.
•
Normally there is no blanking in the MIPI data, only active pixels are included. OEM system
processing provides ROI (Region of Interest) within a larger frame if you want to include blanking in
your packets. However, this adds capture overhead and reduces processing rates on the system.
•
For a diagram of CSI-2 packets see page 17 in the
Camera Serial Interface CSI-2 and CSI-3
from Mipi Alliance.
