Measuring the noise floor of the BASIC_RX on USRP2

I did some experiments tonight, to check the sensitivity of the
BASIC_RX, to try to decide how much
gain I really need in front of it for one of my projects.

I have a signal at about -32dBm, and verified the pk-pk voltage on a
oscilloscope, using a 50-ohm load in front of the scope, to make
certain that I was measuring
what the input circuit of the BASIC_RX would see. Then I plugged the
-32dBm signal into the
BASIC_RX/USRP2, and observed that at my selected bandwidth of
approximately 250KHz, the
FFT display had the wideband noise signal at -20dB. I then unplugged
the input from the BASIC_RX,
which should cause it to “see” the 50-ohm resistor that’s across the
transformer input to the
BASIC_RX (at least, according to the schematics). The result was that
the output on the FFT display
dropped by 28db, giving an input noise level of around -60dbm with no
external signal. Now
granted, that resistor on the BASIC_RX input circuit (why is it
there?) will contribute roughly
-174dBm/Hz of noise, but that is tiny in this context.

Now, it’s my understanding that the maximum power level capability of
the A/D is about 10dBm,
but one should perhaps back off from that by about 1dB or so. Doing
the “math” on the 14-bit
A/D resolution, you’d expect a noise floor around -75dBm or so, which
means that roughly
15dB of “bottom end” on the A/D is “missing” when plugged into a

Is this a correct analysis, or have I missed something somewhere in the

Is there enough noise bumping around inside a USRP2 enclosure to
‘stimulate’ the lower
few bits of the A/D without anything connected?

Principal Investigator
Shirleys Bay Radio Astronomy Consortium