Is it possible to syncronize large number of USRPs (e.g. N210) in time
and
frequency for building an large MIMO array.
My imagination would be using an Ethernet switch hubbing all the USRPs
and
connect to a host PC vial 10Gbps ethernet. All the USRPs share a common
10MHz/1PPS input.
The communications between all the USRPs with the host PC is via TCP
(I’m
not sure if UHD support TCP now) so that there is no packet loss.
We previously build a 2x2 MIMO using USRP2s with two ethernet adapters
on
host PC, which is quite successful.
Just keep in mind, on one ethernet network there is no real parallel
transfer, everything is sequential, at least on the shared parts like
the
cable from the switch to the PC. Can you live with this latency? Of
course
you can try to force a given sequence, with known timing, and jitter
should
be small when the network is separated from other networks and has the
pure
purpose to serve your USRPs…
I have to do with industrial test/measurement equipment where those
latency/jitter effects are a real PITA, and I am amazed how easily in
many
RF applications those problems are “workarounded”.
Is it possible to syncronize large number of USRPs (e.g. N210) in time
and
frequency for building an large MIMO array.
My imagination would be using an Ethernet switch hubbing all the USRPs
and
connect to a host PC vial 10Gbps ethernet. All the USRPs share a common
10MHz/1PPS input.
The communications between all the USRPs with the host PC is via TCP
(I’m
not sure if UHD support TCP now) so that there is no packet loss.
We previously build a 2x2 MIMO using USRP2s with two ethernet adapters
on
host PC, which is quite successful.
Thanks for you reply Rahlph.
I our cases, the off-line measurement is enough which is not sensitve to
jitter issue.
What we want to make sure is that at TX side, the waveform to different
antenna can be transmitted at the same time; and at the RX side, some
time
stamp is on the waveform so we kown exactly at which time point the
multiple waveform streams is synchronized.
The other requirement is no sample loss over the ethernet connections is
allowed, especially for large amount of USRPs that packet collision
might
be quite often. I guess if UHD support TCP, this could be solved.