Re: [Discuss-gnuradio] 4 RX with 2 WBXs?

[Jason Abele]

On Fri, Sep 10, 2010 at 09:38:06PM -0400, Scott Storck wrote:
> On Fri, Sep 10, 2010 at 5:11 PM, Jason Abele <address@hidden> wrote:
> 
> > On Fri, Sep 10, 2010 at 03:59:34PM -0400, Scott Storck wrote:
> >
> > > The other option that would work for me is connecting antenna 0 to WBX 0
> > and
> > > connecting antennas 1-3 to WBX 1 through a commutator to switch them at a
> > > sub-1 kHz frequency.  Can you tell me if any of the unpopulated headers
> > on
> > > the WBX board are digital outputs that could potentially be used to
> > generate
> > > a switching signal and/or low rate ADC inputs that could be used to
> > sample a
> > > switching signal generated elsewhere?
> >
> > You could already switch between the 4 antenna ports (2 on each WBX)
> > using the antenna switches built into the WBX at sub-kHz speeds.  If you
> > really needed to solve the antenna switching problem externally, there
> > are GPIOs available on the some of the 0.1" pin headers, check the
> > schematics here:
> >
> > http://code.ettus.com/redmine/ettus/projects/public/documents
> >
> > Thanks, I'll definitely try to use the internal switching, but I'm not sure
> if it'll work for my application--my original goal was to measure relative
> phase and amplitude of antenna 1 vs 2, 3, 4.
> Looking at the schematics, I see that the several of the WBX's unused
> io_rx/tx signals (io_rx[15:14] and io_tx[15:8]) are available on the
> simple_gdb's J15.  Am I on thre right track?  If so, that will be perfect.
> Marcus mentioned that GPIO support was "spotty" though.  From the USRP FAQ
> page on FPGA registers (
> http://gnuradio.org/redmine/wiki/gnuradio/UsrpFAQFpgaRegs) it seems like it
> would be straighforward to toggle two of the pins to control the
> commutator.  Is there something I'm missing?
> 

You are on the right track here.  I think the statement about GPIO
support being spotty is an artifact of possible bit-rot in the gr-gpio,
but for your use case, I would just make your own version of the wbx
daughterboard code in: <SRCDIR>/usrp/host/lib/db_wbxng.cc

If you could clarify your application a little bit, someone here might
be able to help you be more clever in using the existing antenna
switching and a single USRP to fit your application.  Does it need to be
a measurement of three antennas (which can be switched) vs one fixed
reference antenna?

> 
> > > Finally, what it would take for 2 USRPs to operate a total of 4 WBXs
> > > coherently?  I assume this would at least require connecting Clock Out on
> > > USRP0 to Clock In on USRP1, but is that enough to get all 4 WBX PLLs
> > locked
> > > together?  Also, does using two USRPs simply mean that the max USB data
> > rate
> > > for each USRP is halved, or is the effect more severe?
> >
> > If you want 4 coherent RX using WBX, what you want is 4 USRP2's, 4 WBX's
> > and a GPSDO to supply 10MHz Refclock and PPS signals to all the USRP2's.
> > You would also need 4 separate Gigabit ethernet ports on your host PC.
> > Then, you will have 25Msps from each receiver.
> >
> > That sounds good, but I'm afraid I don't have the budget for that much
> hardware on this project.  I can afford another USRP and 2 more WBXs
> though.  Can you tell me how close would that get me to my goal of 4
> coherenet receivers if I use the first USRP as the reference for the second?
> From what I've read, it sounds like the 4 PLLs will not drift relative to
> each other, but that there will be random phase offsets that must be
> calibrated out after each tune?  If so, that would be fine as I already need
> to do a calibration before each data collect to account for other sources of
> phase offset.
> The signal I'm trying to measure is less than 200 kHz bandwidth, so I can
> afford to decimate by 128 in the USRP.  Hopefully this means that USB
> bandwidth won't be a problem even with two USRPs on the same PC?

So, in principle, with a clock modification to share the same reference
clock between two USRPs, and two independent USB busses, you may be
getting close to what you need.  At decimation 128, you would have the
bandwidth on the USB bus for your four simultaneous receive, but others
may have better insight about the practicality of multiple USRPs on one
USB bus.  Also, with a shared reference clock, the WBX PLLs will not drift
relative to each other once they are locked.

>From there, consider the following:

You may need to consider the USRP timestamp fpga work from Eric
Schneider in order to time-align the samples from two separate USRPs:
http://lists.gnu.org/archive/html/discuss-gnuradio/2010-09/msg00012.html

and you will still need some way of setting the timestamp clocks on both
USRPs simultaneously.

You may need to think carefully about not using fractional-N in the WBX
or resetting the fractional divider state synchronously on two USRPs.
You might wish to modify the WBX daughterboard code to turn off the
frac-n divider and stick to integer-n.

You may need to consider the refclock R divider states on the two WBXs,
but you indicate that the phase offset is something you might be able to
calibrate out after each tune.

Hopefully we can discuss your application further here and see if there
is a way to keep it on one USRP to limit some of the extra work

Jason