Intro:

    We have 2 N210 usrps. One with a sbx daughter board and a second with a wbx daughter board.
We are using the wbx daughter board to take  plasma line data for aeronomy (using juha's software).

Devices:

• Use uhd_find_devices to get devices
  
• use uhd_usrp_probe  --args addr=ipadr  to probe each device.
• Note: We are currently using the 2nd usrp with the wbx daughter board for plasma line observations.
  

usrp2/n210 specs

• external pps reference
• external 10Mhz reference
• fixed 100Mhz clock rate
• 100Ms/s dual adc.
  
• fpga:
• xilinx spartan 3a dsp 3400 fpga
• 2 rx DDC chains in fpga
• 16 bit and 8 bit sample modes (sc8 and sc16)
• Up to 25 Mhz or rf bandwidth with 16 bit samples
• Up to 50 Mhz of rf bandwidth with 8 bit samples
• LED'S:
• A: transmitting
• B: mimo cable link
• C: receiving
• D:firmware loaded
• E: reference lock
• F: cpld loaded
• Ref Clock 10Mhz
• usrp2: 5 to 15dbm
• N210: 0 to 15 dbm
• PPS:
  
• usrp2: 5V pp
• N210: 3.3 to 5V pp
• Only 1 daughter board slot.
• WBX Daugherboard
• 5db noise figure
  
• 2 quadratrure frontends (1 transmit, 1 rcv)
• Defaults to direct conversion.
• Can be used in low IF mode through lo_offset with uhd:tune_request_t
• receive lo. can be set to integer=N tuning for better spur performance.
• rcv antennas: tx/rx or rx2
• Bandwidth: 40 Mhz rx & tx
• SBX Daugtherboard
• 2 quadratrure frontends (1 transmit, 1 rcv)
• Defaults to direct conversion.
• Can be used in low IF mode through lo_offset with uhd:tune_request_t
• receive lo. can be set to integer=N tuning for better spur performance.
• rcv antennas: tx/rx or rx2
• Bandwidth: 40 Mhz rx & tx
• LEDs
• All flash when daughterboard is initialized
• TX LD: transmit synth lock detect
• TX/RX: receiver on Tx/Tx antenna port (no TX)
• RX ld:  receive synth lock detect
• RX1/RX2: receiver on RX2 antenna port
• Tuning
• usrp has 2   tuning stages:
• rf frontend goes rf to if (daughter board)
  
• DSP: goes IF to base band (mother board)
• After tuning, check for lo's locked.. or wait for up to a second.
  

Documentation:

• USRP Hardware driver and USRP manual
  
• USRP N210 product details .. with list of daughter boards
  
• USRP N200/210 data sheet (.pdf)
• data sheet: a/d for our N210: ads62p44   (.pdf)
  
• USRP2 motherboard schematic (.pdf)
  
• wbx daughter board
  
• schematics (.pdf)
• fe_schematic (.pdf)
• adl5387_datasheet (.pdf)  wbx mixer
• sbc daughterboard
• schematics (.pdf)
• adl5380 datasheet (.pdf)   sbx mixer
  

Measurements (in date order):

170131 measuring the basic rx brd response

    We have been using (feb17) the wbx daughter board for usrp datataking with the 430 tx. The wbx has a built in mixer that allows it to mix 260 MHz down to DC. It  generates an I and Q datastream, each of which is sampled by a different AtoD converter. The mixer is convenient, but using 2 digitizers can cause phase errors in the I,Q dataset.

    We have also purchased 2 basic rx daughter boards. They  pass the IF signal through to a single digitizer which does real sampling (at 100 MHz).  This guarantees that there are no IQ phase errors.

How the basic rx, usrp mother board works:

• The basic rx board passes the IF signal through to a single digitizer (real sampling).
• the a/d  runs at 100 MHz
• The 100 MHz sampling digitally down converts the IF input to 0 to 50 MHz. The nyquist zones are:
• 0-50 MHz  not flipped in freq
  
• 50-100 MHz  flipped
  
• 100-150 MHz not flipped
• 150-200 MHz flipped
• 200-250 Mhz  not flipped
• 250-300 Mhz  flipped
• If the input band crosses a nyquist zone (like the 260 MHz IF example below, then part of the band will be aliased at 50 MHz (into the same side since it is real sampling).
  
• The usrp will then digitally mix the requested cfr down to -25 to 25 MHz.
• If the band at  0-50 MHz was flipped by the digital down conversion, then the digital mixer will reflip the band so it is in increasing freq
• there is a cic and halfband filter that then reduces the bandwidth to 25MHz (if we use the maximum clock).
  

Stepping a sine wave through the band.

• Stepping the synth.
• /share/megs/phil/x101/170131/swpfreq.proc was used on vxWorks. It does:
  
• Set synth output to -40 dbm
• loop through the 41  1 Mhz steps
  
• sync to 1 sec tick
• set freq
• wait 2 ticks
• endloop
• so there are 2 seconds of stable sine wave, and 1 sec when the sine wave was changing.
• The usrp was started on a hardware  1 sec tick prior to stepping the synth.
• unfortunately we weren't running ntp on usrp167 computer, so the system clock did not match real time.
• I had to search for the the correct "second" datafile that had the clean sine wave. After this i just stepped by 3 1 second files..

SETUP

• hp 3GHz synth to set to -30 dbm amp. routed through the external output
• 2nd IF noise source selected
• synth output and noise src combined
  
• Tests with filters:
  
•  -> 6db  -> filter -> 6db pad -> usrp
• Test with no filter
• -> 6db pad -> usrp
  

Three tests were done:

• 31jan17  IF cfr =260Mhz, 240-280 MHz IF filter. step 240 to 280
  
• this test crossed a nyquist zone (240-250 zone, 250-280 next zone) so it shows the aliasing
• 31jan17  IF cfr = 230 Mhz  215-245 MHz IF filter. step 210 to 250 MHz
• 16feb17  IF cfr = 125 MHz, no IF FILTER. STEP 105 to 145 MHz
• this is the band we will eventually use.

Processing the data:

• for each test
  
• Find the usrp file with the starting sine wave
• For each 1 MHz freq step
• Read in 20 records of 2^20 points each.
  
• compute power spectrum on each 2^20 len dataset
• average the 20 spectra
• smooth,decimate by 16
• We ended with a 64K long spectra over 25 MHz (giving 381.47 hz resolution
• Convert to DB, scaling by global max value.
  
• For each freq step find all of the points that were above -40db (-35db for 260 test)

Plotting the results:

    Summary of the 3 tests with the band shapes (.ps) (.pdf)

• Page 1 Nyquist zones and the 2 tests..
  
• top: nyquist zones and the tests
• black are nyquist zones that are not flipped, red are zones that get flipped.
• blue lines: the 260 +/- 20Mhz test. You see that 240-250Mhz crosses a nyquist zone .. thus causing aliasing.
• I did the 260 test to show that we should use the 260Mhz IF (with it's filters) with the basic rx board.
  
• green lines: 230 +/- 20 Mhz test. This sits within 1 nyquist zone.
• middle:
• where the digital down conversion leaves the tests.. the 125 test is not shown, it is in the middle of the band.
• bottom:
• where the freq end up after the digital mixing.. the plot is mislabeled.. it should be centered at 0 hz.
• Page 2 230 and 125 Mhz band shapes.
  
• top:  usrp band response. 230 Mhz test with 230Mhz IF filter.
  
• the black line is the response of the 30Mhz IF analog filter that was used.
• the dashed red lines show the 25 Mhz band edges
• The lines to the right and left will be aliased back into the band.
• bottom: usrp band response 125 MHz test, no filter
• I've over plotted what the 230Mhz filter response would be if it's center was shifted to 125 MHz.
• the digital filter aliases are:
  
• 112 -10db
• 111 -18db
• 110 -32db
  

    Plotting the individual tests. The 41 spectra with 10 spectra per frame.

• When plotting the various spectra, i increased the freq for each sine wave by a small amount so you could see harmonics that sat on top of another sine wave.
  

• 31jan17: step 240 to 260 Mhz. 260 +/- 20 Mhz filter (.pdf)
  
• 260 folds down to 40Mhz  and it is flipped.
  
• top frame 240 to 249 Mhz. These are aliased by the sampling/digital down conversion since they sit below the 250 to 300 Mhz nyquist zone.
• 2nd frame: 250 to 259 . not sure why 251,252,253 are aliased. this is at the edge of the 50 Mhz band
• 3rd frame : 260 to 269
• bottom: 270 to 279.  273,274,275 are the alias from the digital filtering.
• 31jan17: step 210 to 250 Mhz. 230 +/- 15 Mhz filter (.pdf)
• 230 folds down to 30 Mhz. it is unflipped.
• the digital filter edges  are at 217.5 and 242.5
  
• top  frame: 215,216,217 are aliases from the digital filter (edge 217.5
• 2nd 220  to 229
• 3rd  230 to 239
• The low level spikes spaced by 2 Mhz are the 1st harmonic from the synth.
  
• bottom 240 to 249. 243,244,245 are aliases from the digital filter.
• 16feb17: step 105 to 145 MHz (.pdf)
  
• 125 cfr folds down to 25 Mhz (unflipped). no filter used. this is  the setup we will probably use.
• The digital filter edges are at 112.5 and 137.5 Mhz
• The vertical scale was switched -50 to -60 db (i had less noise source power).
• The 2 MHz spikes at -40dbm are from the synthesizer (2st harmonic).
• top: 105 to 114 Mhz. 110,111, 112 Mhz are aliased at the digital filter edge.
• 2nd 115 to 124 Mhz
• 3rd: 125 to 134 MHz
• bottom: 135 to 144 Mhz. 138,139,140 are aliased at the edge of the digital filter.

Summary:

• The usrp basic rx board was tested at 260,230 and 125 Mhz. A sine wave was stepped +/- 20Mhz in 1 Mhz steps.
• the 260 IF should not be used with the basic rx board.. too much aliasing
• The 230 MHz filter used for the 230 Mhz test does a good job of anti aliasing.
• 125 Mhz test shows that this is an acceptable cfr to use for observations.
• the band shape of the 230 Mhz filter (but centered at 125 Mhz) will do a good job of anti aliasing.
• start to fall off at 107 Mhz, down 32 db at 100Mhz
• We could ask for a steeper filter.. although i don't think it is necessary.
• It might be nice to ask for a linear phase filter (in case the pulsar people ever use this).
• 2.5 Mhz away from the edge of the 25 Mhz band, the alias power  is down 33 db.

<- page up
home_~phil