Test setup:
o A signal generator producing -46dBm at 38.1MHz
o USRP2 + Basic_RX
o GRC flowgraph:
o An FFT filter, reducing the incoming 250KHz bandwidth to 5KHz
o A power detector:
(I**2 + Q**2)--->single-pole-iir-filter
o A strip-chart scope display
Using the scope display, I measured the detector output difference
between nothing connected (which on the BASIC_RX means that
the A/D is “seeing” the 50-ohm termination across the input-half of
the balun transformer), and my -46dBm signal connected.
That difference amounted to a 65dB difference in detected output power.
Doing the math, does that mean that a “naked”
USRP2 + Basic_RX is sensitive down to roughly -111dBm, or should I
factor in root(5KHz), which brings it up to -92dBm.
–
Marcus L.
Principal Investigator
Shirleys Bay Radio Astronomy Consortium
http://www.sbrac.org
On 10/7/2010 2:38 PM, Marcus D. Leech wrote:
factor in root(5KHz), which brings it up to -92dBm.
I would spec that configuration as being sensitive down to -111dBm, as
that is the input signal power that is equivalent to the noise power
you’re seeing (you can spot signals so long as they’re slightly higher
than -111dBm). Of course, as you do some processing, or change filter
settings, this will change. It also doesn’t mean that you could demod a
signal at -111dBm (you still need some margin of SNR), and if you were
speccing for a customer, you might bump that up a bit to give yourself
some ‘wiggle room.’
That’s at least how I see it. Anybody else care to weigh in?
–
Patrick Yeon
ThinkRF
613-369-5104 x418
On 10/07/2010 07:26 PM, Patrick Yeon wrote:
I would spec that configuration as being sensitive down to -111dBm, as
that is the input signal power that is equivalent to the noise power
you’re seeing (you can spot signals so long as they’re slightly higher
than -111dBm). Of course, as you do some processing, or change filter
settings, this will change. It also doesn’t mean that you could demod
a signal at -111dBm (you still need some margin of SNR), and if you
were speccing for a customer, you might bump that up a bit to give
yourself some ‘wiggle room.’
Well, in this case, there’s no “demodulation” required, since this is
for a radiometer. My working assumption in situations like this is that
you need about 3dB above the inherent noise power of the detector to
reliably detect the presence of excess noise at the input, assuming
very short integration times on the detector.
Incidentally, I also did some numerical work based on the data sheet for
the LTC2284 A/D that’s used in the USRP2. The
NVSD of the LTC2284 is roughly 25.5nV/root(Hz), which converts to
roughly -145dBm/Hz NPSD, or roughly 28.5dB noise
figure (which if this were an amplifier, would be absolutely awful!).
What that means for what I have in mind is that I only
need a little bit of gain ahead of the A/D to achieve the sensitivity
levels I need.
That’s at least how I see it. Anybody else care to weigh in?
–
Marcus L.
Principal Investigator
Shirleys Bay Radio Astronomy Consortium
http://www.sbrac.org
“Marcus D. Leech” [email protected] writes:
Using the scope display, I measured the detector output difference
between nothing connected (which on the BASIC_RX means that
the A/D is “seeing” the 50-ohm termination across the input-half of
the balun transformer), and my -46dBm signal connected.
I see your point and your logic, but this seems to be assuming too much.
What if you adjust your signal generator to -110 dBm? That’s a fairly
low value, but it should be easily acheivable on most RF signal
generators.
That difference amounted to a 65dB difference in detected output power.
Doing the math, does that mean that a “naked”
USRP2 + Basic_RX is sensitive down to roughly -111dBm, or should I
factor in root(5KHz), which brings it up to -92dBm.
If you’re looking at CW, 5 KHz is huge, and probably as you bring down
the bandwidth you’ll get the same output as long as you’re in the
passband.
This is how ARRL does receiver tests; it’s probably at least good to
understand how you’re differing:
http://p1k.arrl.org/~ehare/testproc/testproc.pdf
You could probably convince them to run a set of tests on a USRP-based
radio, sort of a product review without a product and without really
doing the rest of the review.