We are currently evaluating a new generation of our hardware in which we are hoping to interface with our custom analog front-ends via Optical links powered by MGT high speed serial links from an FPGA target.
We've been able to work out (theoretically at least) many aspects of the design but one "small" issue is proving hard to correctly pin down mentally.
We have a requirement that we run all devices snychronously. I know we can do clock recovery from Aurora on FPGA but that this will most likely require some clock cleaning before using as an input clock for FPGA code on our front-ends. But this is syntonation. We don't yet have an idea of "time".
Enter TSN and White Rabbit. I know NI has already released products with TSN functionality (something I think is really really cool BTW). On reading about White Rabbit I'm left feeling there's a huge amount of functional overlap between the two. Where White Rabbit seems to have the "edge" is the definition of actual "time" and not just syhcnronicity of individual nodes of hardware. Before we go too far down tha rabbit hole (There's a circular reference for you) I was wondering if there is anyone with more practical experience on the idea of multi-device time synchronisation, clock recovery and so forth WITHOUT being specifically bound to Ethernet protocols.
Any volunteers, any decent information sources?
Nope, I've never needed it but every now and then I think about how I would do it if I had to. Is GPS on all the targets accurate/cost effective/feasible enough? You can buy cheap kits from sparkfun and the like. Look for one that has a PPS output.
Otherwise, maybe have a master broadcast the time over SPI and sync to the last clock?
there is a lot of similarities betwenn the two, you are right. One thing to consider is the following: Time Sensitive Networking is NOT a National-Instruments-only thing, but rather the evolution of the ethernet and maintanined by the IEEE consortium:
which means that TSN will be at some point the new standard of any ethernet network. It also has the possibility to work via wifi. As far as I know, the white rabbit relies on these standards, but takes them a step further:
Open Hardware Repository | Wiki | Wiki
So, to summarize my point: while WR is more accurate (sub nano-second, compared to a few Nano-seconds for TSN - at leat the demo I built with NI hardware), TSN is more future proof and there are a bunch of companies involved:
Analog Devices, Belden/Hirschmann, Bosch Rexroth, B&R Industrial Automation, Cisco, Intel, Hilscher, Kalycito, KUKA, National Instruments, Renesas Electronics, Schneider Electric, SICK AG, TTTech, Xilinx
I hope this helps.
If you need more practival considerations, let me know.
Here is a good place to start:
LabVIEW Time Sensitive Networking (TSN) - Discussion Forums - National Instruments