TX radiation pattern

1
TX radiation pattern

• In order to make our TX antenna omni directional,
  by rotating it, we have to first figure out what
  is the beam width of a single antenna. Seen from
  the plot below, the antenna beamwidth is 30
  degree

Single Antenna Beam (-30o30o)

• Single Antenna Beam (-3030)
• Antenna Array (Two Antennas)

2
TX radiation pattern

• We test two sets of radiation patterns, based
  on rotating the TX 30-degree each (shown down)
  and 20-degree each (shown right)

• 20-degree is more omni-directional compared to
  30-degree, so during the channel measurement, we
  will make the TX rotating 20 degree each step

3
Thermal noise cancellation Considerations

• Tolerable measurement time

• Equipment noise
• 2.5mV noise from oscilloscope
• Quantization noise Changes as signal peak goes up
• Input referred noise of amp
• Amplifier specification Gain, noise figure...

When the frequency of the pulse generator is
100kHz, even if we use average number equals
4096, the total time for a whole round
measurement will be 5.23hours, which is still
tolerable and can make our result more accurate.

• Other considerations
• High bandwidth
• Relative small number of bits for scope
• Jitter issue (keep within 10ps)

4
Input Referred Noise vs No. of Amplifiers

• When there are more than two amplifiers, the
  noise is going to be dominated by the amplifiers
  but not the oscilloscope, so our choice of two is
  smart and lucky!

5
Different Estimations and Application
Nonparametric Methods

• Simplest periodogram
• Return power per unit frequency
• Good for high SNR and long data
• Not consist estimator
• Improved welch
• Dividing data into overlapping segments
• computing a modified periodogram of each segment
• averaging all the PSD estimates
• Can choose window and overlap percentage
• Variation and resolution trade off
• variance inversely proportional to the number of
  segments
• Good for low SNR
• Modern multitaper
• filtering signal through a filter bank optimal
  FIR BP filters, derived from DPSS
• Time-bandwidth parameter (NW) controls the
  variation and resolution tradeoff
• as NW increase, variation decrease and BW for
  each taper increase
• Pretty computing time-consuming

6
Different Estimations and Application Other
methods

• Parametric methods
• PSD is assumed to be the output of a linear
  system driven by white noise
• Yule-Walker autoregressive method
• Burg method and Covariance method
• Estimating the parameters (coefficients) of the
  linear system which hypothetically "generates"
  the signal
• Better results than Nonparametric methods when
  data is relatively short
• Subspace methods (high resolution methods)
• High resolution and super resolution methods
• Self control signal number and threshold
• Multiple signal classification method and
  Eigenvector method
• Pseudo spectrum estimation
• Based on eigenanalysis of the autocorrelation
  matrix
• Effective detection for low SNR and spectra of
  sinusoidal signals buried in noise

7
Interference deconvolution

• More exploration on the frequency deconvolution
• Add more consideration to the relative power
• Add more consideration to the harmonics of the
  interference.
• Capturing interference with the TEM horn
• For each case, deconvolution is performed by
  considering

-- Measured scope and amplifier spectrum -- Final
deconvolved interference psd
-- Raw spectrum (interference convolve with the
scope and amplifier response)
8
Interference deconvolution

• Interference at Berkeley downtown (after
  deconvolution)
• Spectrum usage percentage
• Many frequency holes are presented, which can
  potentially be used for the cognitive radio groups

• Especially low at 35GHz

9
(No Transcript)