> Hi, John, > > On Aug 28, 2009, at 6:02 , John Ford wrote: > >> We need a packet-enabled PFB and FFT, and then we could discard this >> synchronizer business, but until then, we are stuck using a >> synchronous >> system. We are running at 1.6 Gs/s. > > Do you expect the packet enabled PFB and FFT to alleviate the > synchronization challenges? If so, can you please elaborate on > that. It sounds great, but it runs counter to my intuition.
If the data from each ibob is packaged in a packet, eg. with UDP, the packets can then flow in to the BEE2's 10 GbE ports at their own good time, and we can wait until the complete frame of data from both (all) ibobs is there. You still have to synchronize the 2 (or more) ibobs to each other. No getting around that. But you would not have to synchronize the xaui ports (which are inherently asynchronous, as Jouko points out). This would require that the bee run at a slightly faster clock speed than the ibob, to prevent overrunning the buffers. I think this is essentially how all the packetized correlators work. Just pushing the packetizing upstream of the PFB/FFT stage. John > > Thanks, > Dave >
