TITLE: Radar FM-CW S-band Data [UMass]


Stephen J. Frasier
Microwave Remote Sensing Laboratory
University of Massachusetts,
Amherst,MA 01003
Tel: 413-545-4582 Fax: 413-545-4652     


The University of Massachusetts FM-CW radar was deployed at IHOP 2002 to
provide verification information of wave structures within and atop the ABL.
This vertically pointing, high resolution (2.5 m range resolution) radar has
shown its capability in discerning small scale wave motions like internal
gravity waves and Kelvin-helmholtz waves. The radar was operated almost
continuously from 13 May to 13 June. The radar was located at the "Auxilliary
site" approximately 20 km east of the NCAR S-Pol radar site in the Oklahoma


The S-band FMCW radar operates at 2.9 GHz and provides 2.5 m resolution
profiles of radar reflectivity up to 2.5 km AGL. It employs two 8' parabolic
dish antennas and a 250 W TWT amplifier. The antennas are mounted on the
flat-bed of a truck and the radar electronics are in the cab of the truck.
The radar obtains individual profiles at approximately 50 ms intervals. The
radar implements an internal calibration loop by injecting into the receiver
an attenuated and delayed sample of the high power amplifier output, via a
surface acoustic wave (SAW) delay line. This produces a continuous echo at an
apparent range of 1500 m (corresponding to a 10 microsec. delay), permitting
radar system monitoring.

FM-CW Radar System Characteristics  (Mode 1 / Mode 2)
center frequency  :  2.94 GHz
bandwidth         :  60 MHz / 30 MHz 
transmit power    :  250 W
range resolution  :  2.5 m / 5 m
height coverage   :  2.5 km / 5 km
max. velocity     :  0.5 m/s (unambiguous)
antenna gain      :  34 dB
antenna beamwidth :  3 degrees
polarization      :  single, linear


The radar was operated in two modes.
Mode 1 :
Range = 2.5 km
Resolution = 2.5 m
Maximum unambiguous Doppler velocity = 0.5 m/s

Mode 2 :
Range = 5 km
Resolution = 5 m
Maximum unambiguous Doppler velocity = 0.5 m/s

The radar was operated remotely from the University of Massachusetts and
quick view images in the form of .png files were made available on the

Signal processing : During each sweep of the radar, the down-converted signal
is digitized by a 16-bit A/D convertor, which samples the analog signal. 

Each profile contains 2048 or 1024 points depending on the mode.

For each raw profile :

1. the DC is removed from each profile. 
2. a Hanning window is applied to minimize the effect of sidelobes
3. signal is converted to frequency domain by FFT
4. "Coherent on receive" applied to synthetic echo from acoustic delay line  
   a. determine phase of delay line echo
   b. determine associated delay
   c. calculate appropriate phase for each frequency corresponding to delay
   d. apply conjugate phase to each frequency sample
Subsequent processing is in 100 profile chunks (5 sec. averaging)

For each height over a 100 profile block :

5. calculate mean (clutter) complex echo for each height
6. remove clutter for each height
7. calculate average power
8. compute signal to noise ratio (snr) using estimate of noise floor  
   (noise floor is estimated from several images)
9. calculate single-lag covariance and use this to :
    a. compute correlation coefficient (rho)
    b. compute velocity (vel) from phase 
       (correlation coefficient must be greater than 0.2 for valid estimate )
For the entire image :

10. remove noise from power computation  
11. compute equivalent reflectivity of calibration line
12. compute reflectivity, and hence structure function parameter (cn2)


VERSION 1.0 September, 2002 
File type : NetCDF
File naming convention : YYYYMMDDhhmm 
Duration : Most files are one hour files

The following information is stored in each netCDF file :

  Header Information :
   1. date : 80 bytes, "Start time and date in UTC"
   2. rangegates : floating point, "Number of range gates"
   3. rangeresolution : floating point, "Range resolution in m (2.5 m or 5 m)"
   4. profiles : floating point, "Number of profiles (after averaging)"
   5. sec_per_profile : floating point, "Averaging period (5 s)"
   6. max_vel : floating point, "Maximum unambiguous velocity (0.5 m/s)"

 Processed data:
  (Data is scaled to minimize file size and must be rescaled appropriately)

   1. time_stamps : Array of (short) integers 
                    Number of elements = 'profiles'
      -- "Time in minutes since start time"
      -- Must be divided by 100.0

   2. cn2 : 2-dimensional array of (short) integers
            Number of elements = 'profiles' x 'rangegates'
      -- "Log Cn^2  Estimated structure function parameter (in dB)"
      -- Must be divided by 1000.0

   3. vel : 2-dimensional array of (short) integers
            Number of elements = 'profiles' x 'rangegates'
      -- "Doppler velocity in m/s"
      -- negative velocities denote movement away from the radar (upward)
      -- Must be divided by 1000.0

   4. rho : 2-dimensional array of (short) integers
            Number of elements = 'profiles' x 'rangegates'
      -- "Pulse pair correlation coefficient"
      -- Must be divided by 1000.0

   5. snr : 2-dimensional array of (short) integers
            Number of elements = 'profiles' x 'rangegates'
      -- "Signal to noise ratio (SNR in dB)"
      -- Noise was estimated from several images as 0.14 for Mode 1
         and 0.28 for Mode 2  


  Radar was operated almost continuously from 13 May to 13 June. Radar was
  down temporarily (15-16 May) to remove some of the artifacts (undesired
  harmonics). Radar was operated in Mode 1 (2.5 km range) in the night and
  Mode 2 during the day (5 km range) when the convective layer was higher.
  Radar was down again on 30 May to fix spurious data in lowest range gates.
  The cause for spurious data has been attributed to AC power supply
  fluctuations. Subsequently, the radar was operated only in Mode 1 for

Spurious noise-like signals between heights of 500 and 800 m are present in
many files. These do not actually represent any atmospheric phenomenon, but
are instrumental artifacts most probably due to instabilities in the
high-voltage power supply of the radar's travelling wave tube amplifier. 
Sometimes this feature is present throughout an one-hour file, making its
detection difficult.  These features appear till a height of about 800 m in
the 2.5 km operation mode and 1.6 km in the 5 km operation mode. In some data
files, this feature appears as a horizontal band existing across the entire
image. To help correctly identify these regions, some sample images have been
posted at
http://abyss.ecs.umass.edu/tep/experiments/ihop2002/interpretation.html. So
far, we have been unsuccessful at removing the noise from the raw data.

Data can be viewed using any plotting software. IDL plotting routines
developed by the University of Massachusetts are available on request.