ASK THE EXPERTS
More Answers From Mike Miethig
R & D Manager at Teledyne DALSA
- Email: mike [dot] miethig [at] teledynedalsa.com
- Tel: (519) 886-6000, x2130
What is the difference between FPD Link, GMSL and Camera Link (HS)?
Hi Tommy: FPD-Link is the technology used in Camera Link. This technology sends 4 7:1 serialized data streams + a parallel clock using LVDS levels. Cable skew and data patterns influence distance achieved. Camera Link HS is packet based. The M protocol runs 8b/10b encoding at 3.125 Gbps/lane and achieves a cable bandwidth of 2.1 GByte/s. The X protocol runs at 10.3125 Gbps with scrambling and Forward Error Correction and achieves a data throughput of 1.2 Gbyte/sec per SFP+ connector.
How is data transmitted over camera link affected by using cable lengths longer than the 10 meter limit typically specified? Is it just that the highest frame rates cannot be realized, but lower frame rates are possible? Or is it that the image integrity is compromised?
Hi Kathy: Camera Link uses a clock (1 differential pair) in parallel with the data (4 LVDS differential pairs). The data bit rate is 7x the clock rate and there is certain phase relationship between clock and data that must be maintained for correct data transfer. The data lines are not DC balanced and so data transmission is a function of the data values. The worst case is sending all 1's or all 0's for a long time and expecting the transision between 1 and zero to be correct. As others have mentioned, distance is a function of camera, framegrabber, cable, bit rate and CL configuration. It should be noted that there is no error detection in camera link. Camera Link HS uses 8b/10b or 64/66b encoding to ensure DC balance and achieves 15 meter on copper cables at 3.125 Gbps per lane. The cable is similar to Camera Link but has 7 down lanes for an effective data bandwidth of 2.1 GByte/sec in a single cable. Bit errors are detected and CLHS is single error immune.
Is the IP-Core of the M-Protocol of the Camera Link HS standard capable of 2.1 GB/s per lane, or is it capable of 2.1 GB/s per 7 lanes (so 300 MB/s per lane)? The flyer on the website is not clear about that. (link: http://www.visiononline.org/userAssets/aiaUploads/file/CLHS%20Data%20Sheet%20M%20Protocol.pdf)
Hi Marten: The M-Protocol and the IP core is currently designed to support 3.125 Gbps per lane. An effective bandwidth of 300 MByte/sec of data transfer per lane is achieved for a packet size of about 1k bytes,. The IP core supports 7 lanes which is the maximum supported in the CX4 cable, and results in 2.1 GB/s per cable. For multiple cable applications a second core is used to support the communication query functions of the slave frame grabber. 3.125 Gbps enables copper cable distances of 15 meter to be reliably achieved when driven directly from an FPGA. We use an equalizer chip in the camera to account for varying cable loses and the TLK3134 is used in the current frame grabber.
I'm interested in sending data to camera without ACK from camera. Does the standard support it? I saw there is a message channel, but its direction is camera to application. Thanks, Tal
Camera Link HS (CLHS) has a bidirectional command or message channel generally used to control the camera operating mode or read camera status information. The hardware layer of CLHS guarantees message delivery with data resend on error or hardware acknowledge timeouts. The software layer that runs on the hardware layer is specified as GenCP which has software handshaking and messages are acknowledge at this layer as well. The average effective data rate from frame grabber to camera depends on the design of the camera and the frame grabber but can send 1k bytes at 312 MBytes/sec.