Frequency response USRP2 with LFRX and LFTX boards

GNU Radio
1

Hi All

I’m using USRP2 with LFRX and LFTX. I’m trying to read some signals,
ranging from DC up to 2MHz through the USRP2. Unfortunately it does
not behave as it should (approaching freq’s below 1kHz and/or DC).

Using GRC I’ve made two simplified test cases. Case_1 to test TX path
and Case_2 to test RX path to isolate the misbehaviour.
Upon request I could mail the grc blocks/progs.

I observed the following:

case_1 (testing TX path):

grc-blocks:
Constant Source (A), Constant Source (B) → Float To Complex
Float To Complex → USRP2 Sink, Scope Sink, FFT Sink

grc-parameters:
Constant Source (A) → Float, Slider -1.0 … +1.0
Constant Source (B) → Float, Slider -1.0 … +1.0
Float To Complex → no parameters
USRP2 Sink → interp=512, freq=0, gain=0
Scope Sink → sample=32k, XYmode=on
FFT Sink → sample=10k, baseband_freq=0, fft_size=512, refresh=30

Observation:
Playing with variables ChanA and/or ChanB in the range of -1.0 …
+1.0
output from TX_A and TX_B is in the range of -1.0 … +1.0 Volts
depending
on which variable using as input.

Conclusion:
Is what I expect to obtain. Using frequency generator (complex
sinusoid)
gives also the expected results. Using I / Q part of the signal will
put
the value acoordingly either to TX_A or TX_B.

case_2 (testing RX path):

grc-blocks:
USRP2 Source → Scope Sink, FFT Sink

grc-parameters:
USRP2 Source → decim=101, freq=0, gain=0
Scope Sink → sample=32k, XYmode=on
FFT Sink → sample=10k, baseband_freq=0, fft_size=1024, refresh=30

Observation:
Applying external signal to RX_A (Signal Generator 1Vpp 10kHz … 1Hz)
Using 1.0kHz (1Vpp@RX_A), the Scope Sink shows ~ 0.60Vpp, FFT shows
-20dB
Using 100Hz (1Vpp@RX_A), the Scope Sink shows ~ 0.50Vpp, FFT shows
-25dB
Using 10Hz (1Vpp@RX_A), the Scope Sink shows < 0.10Vpp, FFT shows
-40dB
Using 1Hz (1Vpp@RX_A), the Scope Sink shows < 0.02Vpp, FFT shows
-60dB

Measuring outputs of the input amplifier on LFRX board shows NO
frequency
dependet attenuation, besides the basic given one (1:2 factor).
Which means at each otput of the input amplifier we have a signal of
500mVpp with an offset of +1.65V DC (@1Vpp on the input).

Conclusion:
Apparently the internal functions of the FPGA introduce a frequency
depen-
dent attenuation below of 1kHz (filter curve ?).
The input circuit (OP-Amp) does not show this effect. Which means the
input
signal should be available at its full amplitude on the ADC including
DC
levels.
This behaviour does not change when using different decimation rates,
e.g.
remains the same for decim values 100, 101 or 102, in accordance to
doc:
http://gnuradio.org/redmine/wiki/1/USRP2GenFAQ reference:
“What are the USRP2 minimum and maximum decimation/interpolation
…”

Questions:

  1. What kind of circuit (CIC Filter, LowRate HBF, HighRate HBF) is
    introducing
    this frequency response ?

  2. Why does this happen only in the RX path and not also through the TX
    path ?

  3. According to
    http://gnuradio.org/redmine/wiki/1/USRP2GenFAQ reference:
    “How can I use both A/D converters on the USRP2?”
    feeding the signal to RX_B does not give ANY output on the ADC

Any kind of help would be gratly appreciated.
Thanks in advance
Hans

2010/03/29 [email protected]