I took a 256,000 samples per second stream from the USRP and directly
displayed the I/Q stream to the X/Y display of a (slightly modified)
oscope.py.
With a -60dBm unmodulated carrier as RF input, and with the receiver tuned
within 0.1 Hertz of the the carrier, I expected to see a tight constellation
of samples slowly rotating around the origin at the rate of (Fcarrier -
Flocal_oscillator).
To my surprise however, when the carrier & receive frequencies are within
0.1 hz of each other, the displayed constellation of I/Q only lives around
[x=0, y=Amplitude].
I've been scratching my head on exactly why this is and have concluded it
has something to do with the finite length of the Hilbert transformer - we
don't really have true input phase because the Hilbert transformer is of
finite length.
Can a Wise One put into concise words what's causing this phenomena?
Then, a second, even more baffling observation: Every few seconds, the
phase of the constellation seems to slip by 180 degrees and the
constellation jumps to [x=0, y=-1] for a while. It seems to me that this
phase jump is at the rate at which the constellation SHOULD have slowly
rotated, i.e. the constellation rotation angle is somehow constrained to
being only around those two points. In other words, instead of the phasor
slowly rotating at 0.1Hz, it jumps between [0,1] and [0,-1]. Thinking about
this duzi, I conclude that this is somehow related the the above insight,
but as for the phase JUMP, maybe someone with good fundamental insight can
share it in simple words for a simple man?
(PS: my modifications to oscope.py was limited to allowing manual scaling of
both axis in XY mode, so i'm pretty sure i haven't messed up something
significant in the xy display that could cause the above phenomena)