USRP grounding

We are about to connect the USRP with LFRX and LFTX daughter boards to
a laboratory data acquisition system where noise and ground loops are
an issue. Can anyone confirm my understanding of the grounding on the
USRP, daughter boards, and enclosure?

As I understand it, the USRP is floating with respect to the rest of
the laboratory. The only ways the USRP grounds could be connected to
grounds elsewhere in the lab are through the USB ground (depending on
how the host computer is grounded), or through the shields of the
cables connected to the SMA connectors on the daughterboards (see
figure below).

Here are the details as I understand them, from the circuit diagrams
at http://gnuradio.org/trac/browser/usrp-hw/

On the USRP motherboard:

  • Power ground: The USRP is powered through a transformer. The power
    ground of the USRP motherboard (the ground pin of the power
    regulator IC) is connected to one of the terminals on the output
    side of this transformer. Therefore the USRP power ground is not
    connected to the rest of the lab (except for the coupling through
    the transformer).

  • Digital ground: The ground pins of the digital ICs (and the digital
    ground pins of the AD986X) on the USRP motherboard are connected to
    the power ground (same as the ground pin on the power regulator IC).

  • USB ground: The USB connector ground is connected to the digital
    ground.

  • Analog grounds: The analog ground pins of the two AD986X are
    connected to two separate analog grounds, AGND_1 and AGND_2. Each
    analog ground is connected at one point to the power/digital
    ground, through an inductor.

  • The mounting holes are not connected to any ground.

LFRX and LFTX Daughterboards

  • Power and grounds on both daugherboards are connected to the USRP
    power and grounds through the daughterboard connector.

  • Analog grounds: Each daughterboard has its own analog ground AGND
    that is connected to the separate AGND_1 and AGND_2 on the USRP
    motherboard. On each daughterboard, there is no connection between
    its analog ground and its digital ground.

  • Analog inputs: The signal on the SMA connectors is connected
    to the analog input of the instrumentation amplifier IC through a
    resistor, and to the analog ground through resistors and capacitors.

  • The shield on the SMA connectors on the LFRX and LFTX boards is
    connected to the analog ground.

Enclosure (we have a homemade enclosure, a painted metal box):

  • The USRP motherboard mounting holes are not connected to any
    ground, so it is possible to insulate the enclosure from the
    electronics

  • If the SMA connectors on the enclosure are not insulated from the
    enclosure, then the enclosure is connected to both analog grounds
    (through the SMA connector shields on the daughterboards), and
    there could be loops (see figure)

               SMA  |             USB  |             SMA  |
              shield|           ground |            shield|
    

    Enclosure … | … | … | …
    | | |
    LFRX | AGND | AGND | LFTX
    Daughterboards ±-±-- --±- | --±- —±-+
    | | | | |
    daughterboard | | | | |
    connector | | | | |
    | | | | |
    USRP Motherboard —±–))))–±----±-±-±-))))—±–
    AGND_1 coil digital | coil AGND_2
    and power|
    ground |
    | | +
    )))))))
    Transformer )))))))
    | |
    Lab power

Jonathan Jacky wrote:

We are about to connect the USRP with LFRX and LFTX daughter boards to
a laboratory data acquisition system where noise and ground loops are
an issue. Can anyone confirm my understanding of the grounding on the
USRP, daughter boards, and enclosure?
I know I had to decouple the ground from an external source and the
ground of the LFRX board with a large capacitor (several uF) to get
things
working.

If I did not do this, for some reason the input of the instrumentation
amplifier was sometimes DC-lifted out of its input range.

I didn’t look into it in more detail since the capacitor helped, and I
didn’t need a DC-coupling.
(external source was a videosignal so lowest signal freq was 25 Hz)

Greetings,
Martin