% UTP_AO_COHERE a set of UTPs for the ao/cohere method
%
% M Hewitson 06-08-08
%
% $Id: utp_ao_cohere.m,v 1.44 2011/07/22 12:29:58 mauro Exp $
%

% <MethodDescription>
%
% The cohere method of the ao class computes the coherence between two
% time-series AOs.
%
% </MethodDescription>

function results = utp_ao_cohere(varargin)
  
  % Check the inputs
  if nargin == 0
    
    % Some keywords
    class   = 'ao';
    mthd    = 'cohere';
    
    results = [];
    disp('******************************************************');
    disp(['****  Running UTPs for ' class '/' mthd]);
    disp('******************************************************');
    
    % Test AOs
    [at1,at2,at3,at4,at5,at6] = eval(['get_test_objects_' class]);
    
    % Exception list for the UTPs:
    [ple1,ple2,ple3,ple4,ple5,ple6] = get_test_ples();
    
    % Get default window from the preferences
    prefs = getappdata(0, 'LTPDApreferences');
    defaultWinType = char(prefs.getMiscPrefs.getDefaultWindow);
    
    % Run the tests
    results = [results utp_01];    % getInfo call
    results = [results utp_02];    % Vector input          (only with two objects)
    results = [results utp_03];    % Matrix input          (not possible)
    results = [results utp_04];    % List input            (only with two objects)
    results = [results utp_05];    % Test with mixed input (not possible)
    results = [results utp_06];    % Test history is working
    results = [results utp_07];    % Test the modify call works
    results = [results utp_08];    % Test input data shape == output data shape
    results = [results utp_09];    % Test output of the data
    results = [results utp_10];    % Test the basic usage against MATLAB mscohere
    
    results = [results utp_11(mthd, [at1 at1], ple1)];    % Test plotinfo doesn't disappear
    
    results = [results utp_12];    % Test basic symmetry properties of cohere (C)
    results = [results utp_13];    % Test basic symmetry properties of cohere (MS)
    results = [results utp_14];    % Test basic symmetry properties of cohere (C)
    results = [results utp_15];    % Test basic symmetry properties of cohere (MS)
    results = [results utp_16];    % Test basic relationship (MS) <-> (C)
    results = [results utp_17];    % Test units handling: complex cohere
    results = [results utp_18];    % Test units handling: magnitude-squared cohere
    results = [results utp_19];    % Test data lengths
    results = [results utp_20];    % Test with single window
    results = [results utp_21];    % Test number of averages: requested/obtained
    results = [results utp_22];    % Test number of averages: correct number
    results = [results utp_23];    % Test number of averages: syntax
    results = [results utp_24];    % Test the basic usage against MATLAB mscohere
    results = [results utp_25];    % Test Kaiser win and olap: (C)
    results = [results utp_26];    % Test Kaiser win and olap: (MS)
    results = [results utp_30];    % Special cases: same input
    
    disp('Done.');
    disp('******************************************************');
    
  elseif nargin == 1 % Check for UTP functions
    if strcmp(varargin{1}, 'isutp')
      results = 1;
    else
      results = 0;
    end
  else
    error('### Incorrect inputs')
  end
  
  %% UTP_01
  
  % <TestDescription>
  %
  % Tests that the getInfo call works for this method.
  %
  % </TestDescription>
  function result = utp_01
    
    
    % <SyntaxDescription>
    %
    % Test that the getInfo call works for no sets, all sets, and each set
    % individually.
    %
    % </SyntaxDescription>
    
    try
      % <SyntaxCode>
      % Call for no sets
      io(1) = eval([class '.getInfo(''' mthd ''', ''None'')']);
      % Call for all sets
      io(2) = eval([class '.getInfo(''' mthd ''')']);
      % Call for each set
      for kk=1:numel(io(2).sets)
        io(kk+2) = eval([class '.getInfo(''' mthd ''', ''' io(2).sets{kk} ''')']);
      end
      % </SyntaxCode>
      stest = true;
    catch err
      disp(err.message)
      stest = false;
    end
    
    % <AlgoDescription>
    %
    % 1) Check that getInfo call returned an minfo object in all cases.
    % 2) Check that all plists have the correct parameters.
    %
    % </AlgoDescription>
    
    atest = true;
    if stest
      % <AlgoCode>
      % check we have minfo objects
      if isa(io, 'minfo')
        
        % SET 'None'
        if ~isempty(io(1).sets), atest = false; end
        if ~isempty(io(1).plists), atest = false; end
        % Check all Sets
        if ~any(strcmpi(io(2).sets, 'Default')), atest = false; end
        if numel(io(2).plists) ~= numel(io(2).sets), atest = false; end
        % SET 'Default'
        if io(3).plists.nparams ~= 9, atest = false; end
        % Check key
        if ~io(3).plists.isparam('nfft'), atest = false; end
        if ~io(3).plists.isparam('win'), atest = false; end
        if ~io(3).plists.isparam('olap'), atest = false; end
        if ~io(3).plists.isparam('type'), atest = false; end
        if ~io(3).plists.isparam('order'), atest = false; end
        if ~io(3).plists.isparam('navs'), atest = false; end
        if ~io(3).plists.isparam('times'), atest = false; end
        if ~io(3).plists.isparam('split'), atest = false; end
        if ~io(3).plists.isparam('psll'), atest = false; end
        % Check default value
        if ~isequal(io(3).plists.find('nfft'), -1), atest = false; end
        if ~strcmpi(io(3).plists.find('win'), defaultWinType), atest = false; end
        if ~isequal(io(3).plists.find('olap'), -1), atest = false; end
        if ~isequal(io(3).plists.find('type'), 'C'), atest = false; end
        if ~isequal(io(3).plists.find('order'), 0), atest = false; end
        if ~isequal(io(3).plists.find('navs'), -1), atest = false; end
        if ~isEmptyDouble(io(3).plists.find('times')), atest = false; end
        if ~isEmptyDouble(io(3).plists.find('split')), atest = false; end
        if ~isequal(io(3).plists.find('psll'), 200), atest = false; end
        % Check options
        if ~isequal(io(3).plists.getOptionsForParam('nfft'), {-1}), atest = false; end
        if ~isequal(io(3).plists.getOptionsForParam('win'), specwin.getTypes), atest = false; end
        if ~isequal(io(3).plists.getOptionsForParam('olap'), {-1}), atest = false; end
        if ~isequal(io(3).plists.getOptionsForParam('type'), {'C', 'MS'}), atest = false; end
        if ~isequal(io(3).plists.getOptionsForParam('order'), {-1 0 1 2 3 4 5 6 7 8 9}), atest = false; end
        if ~isequal(io(3).plists.getOptionsForParam('navs'), {-1}), atest = false; end
        if ~isequal(io(3).plists.getOptionsForParam('times'), {[]}), atest = false; end
        if ~isequal(io(3).plists.getOptionsForParam('split'), {[]}), atest = false; end
        if ~isequal(io(3).plists.getOptionsForParam('psll'), {200}), atest = false; end
      end
      % </AlgoCode>
    else
      atest = false;
    end
    
    % Return a result structure
    result = utp_prepare_result(atest, stest, dbstack, mfilename);
  end % END UTP_01
  
  %% UTP_02
  
  % <TestDescription>
  %
  % Tests that the cohere method works with a vector of AOs as input. (only
  % with two objects in the vector)
  %
  % </TestDescription>
  function result = utp_02
    
    % <SyntaxDescription>
    %
    % Test that the cohere method works for a vector of AOs as input.
    %
    % </SyntaxDescription>
    
    try
      % <SyntaxCode>
      avec = [at1 at5];
      out  = cohere(avec);
      % </SyntaxCode>
      stest = true;
    catch err
      disp(err.message)
      stest = false;
    end
    
    % <AlgoDescription>
    %
    % 1) Check that the number of elements in 'out' is equal to 1.
    % 2) Check that each output AO contains the correct data.
    %
    % </AlgoDescription>
    
    atest = true;
    if stest
      % <AlgoCode>
      % Check we have the correct number of outputs
      if numel(out) ~= 1, atest = false; end
      
      TOL = 1e-13;
      
      % Get shortest vector
      lmin = min([length(at1.y), length(at5.y), length(at6.y)]);
      % Set Nfft
      Nfft = lmin;
      % Get default window
      if strcmpi(defaultWinType, 'kaiser')
        win  = specwin(defaultWinType, Nfft, find(ao.getInfo('cohere').plists, 'psll'));
      else
        win  = specwin(defaultWinType, Nfft);
      end
      % Compute magnitude squared coherence estimate with MATLAB
      % out: at1->at5
      [cxy, f] = mscohere(at1.y(1:lmin), at5.y(1:lmin), win.win, Nfft/2, Nfft, at1.fs);
      if any(abs(out.y-cxy > TOL)), atest = false; end
      if any(abs(out.x-f   > TOL)), atest = false; end
      % </AlgoCode>
    else
      atest = false;
    end
    
    % Return a result structure
    result = utp_prepare_result(atest, stest, dbstack, mfilename);
  end % END UTP_02
  
  %% UTP_03
  
  % <TestDescription>
  %
  % Test that the cohere method doesn't work for a matrix of AOs as input.
  %
  % </TestDescription>
  function result = utp_03
    
    % <SyntaxDescription>
    %
    % Test that the cohere method doesn't work for a matrix of AOs as input.
    %
    % </SyntaxDescription>
    
    try
      % <SyntaxCode>
      amat = [at1 at2;at5 at6];
      out  = cohere(amat);
      % </SyntaxCode>
      stest = false;
    catch err
      stest = true;
    end
    
    % <AlgoDescription>
    %
    % 1) Nothing to check.
    %
    % </AlgoDescription>
    
    atest = true;
    if stest
      % <AlgoCode>
      % </AlgoCode>
    else
      atest = false;
    end
    
    % Return a result structure
    result = utp_prepare_result(atest, stest, dbstack, mfilename);
  end % END UTP_03
  
  %% UTP_04
  
  % <TestDescription>
  %
  % Tests that the cohere method works with a list of AOs as input.
  %
  % </TestDescription>
  function result = utp_04
    
    % <SyntaxDescription>
    %
    % Test that the cohere method works for a list of AOs as input.
    %
    % </SyntaxDescription>
    
    try
      % <SyntaxCode>
      out = cohere(at1,at5);
      % </SyntaxCode>
      stest = true;
    catch err
      disp(err.message)
      stest = false;
    end
    
    % <AlgoDescription>
    %
    % 1) Check that the number of elements in 'out' is equal to 1.
    % 2) Check that each output AO contains the correct data.
    %
    % </AlgoDescription>
    
    atest = true;
    if stest
      % <AlgoCode>
      % Check we have the correct number of outputs
      if numel(out) ~= 1, atest = false; end
      
      TOL = 1e-13;
      
      % Get shortest vector
      lmin = min([length(at1.y), length(at5.y)]);
      % Set Nfft
      Nfft = lmin;
      % Get default window
      if strcmpi(defaultWinType, 'kaiser')
        win  = specwin(defaultWinType, Nfft, find(ao.getInfo('cohere').plists, 'psll'));
      else
        win  = specwin(defaultWinType, Nfft);
      end
      % Compute magnitude squared coherence estimate with MATLAB
      % out: at1->at5
      [cxy, f] = mscohere(at1.y(1:lmin), at5.y(1:lmin), win.win, Nfft/2, Nfft, at1.fs);
      if any(abs(out.y-cxy > TOL)), atest = false; end
      if any(abs(out.x-f   > TOL)),   atest = false; end
      % </AlgoCode>
    else
      atest = false;
    end
    
    % Return a result structure
    result = utp_prepare_result(atest, stest, dbstack, mfilename);
  end % END UTP_04
  
  %% UTP_05
  
  % <TestDescription>
  %
  % Test that the cohere method doesn't work with an input of matrices
  % and vectors and single AOs.
  %
  % </TestDescription>
  function result = utp_05
    
    % <SyntaxDescription>
    %
    % Test that the cohere method doesn't work with an input of matrices
    % and vectors and single AOs.
    %
    % </SyntaxDescription>
    
    try
      % <SyntaxCode>
      out = cohere([at5 at6], [at5 at1; at6 at1], at6);
      stest = false;
      % </SyntaxCode>
    catch err
      stest = true;
    end
    
    % <AlgoDescription>
    %
    % 1) Nothing to check
    %
    % </AlgoDescription>
    
    atest = true;
    if stest
      % <AlgoCode>
      % </AlgoCode>
    else
      atest = false;
    end
    
    % Return a result structure
    result = utp_prepare_result(atest, stest, dbstack, mfilename);
  end % END UTP_05
  
  %% UTP_06
  
  % <TestDescription>
  %
  % Tests that the cohere method properly applies history.
  %
  % </TestDescription>
  function result = utp_06
    
    % <SyntaxDescription>
    %
    % Test that the result of applying the cohere method can be processed back
    % to an m-file.
    %
    % </SyntaxDescription>
    
    try
      % <SyntaxCode>
      out  = cohere(at5,at6);
      mout = rebuild(out);
      % </SyntaxCode>
      stest = true;
    catch err
      disp(err.message)
      stest = false;
    end
    
    % <AlgoDescription>
    %
    % 1) Check that the last entry in the history of 'out' corresponds to
    %    'cohere'.
    % 2) Check that the re-built object is the same as 'out'.
    %
    % </AlgoDescription>
    
    atest = true;
    if stest
      % <AlgoCode>
      % Check the last step in the history of 'out'
      if ~strcmp(out.hist.methodInfo.mname, 'cohere'), atest = false; end
      % Check the re-built object
      if ~eq(mout, out, ple2), atest = false; end
      % </AlgoCode>
    else
      atest = false;
    end
    
    % Return a result structure
    result = utp_prepare_result(atest, stest, dbstack, mfilename);
  end % END UTP_06
  
  %% UTP_07
  
  % <TestDescription>
  %
  % Tests that the cohere method can not modify the input AO.
  %
  % </TestDescription>
  function result = utp_07
    
    % <SyntaxDescription>
    %
    % Test that the cohere method can not modify the input AO.
    % The method must throw an error for the modifier call.
    %
    % </SyntaxDescription>
    
    try
      % <SyntaxCode>
      % copy at1 to work with
      ain = ao(at1);
      % modify ain
      ain.cohere(at5);
      % </SyntaxCode>
      stest = false;
    catch err
      stest = true;
    end
    
    % <AlgoDescription>
    %
    % 1) Nothing to check.
    %
    % </AlgoDescription>
    
    atest = true;
    if stest
      % <AlgoCode>
      % </AlgoCode>
    else
      atest = false;
    end
    
    % Return a result structure
    result = utp_prepare_result(atest, stest, dbstack, mfilename);
  end % END UTP_07
  
  %% UTP_08
  
  % <TestDescription>
  %
  % Test the shape of the output.
  %
  % </TestDescription>
  function result = utp_08
    
    % <SyntaxDescription>
    %
    % Test that the cohere method keeps the data shape of the input object. The
    % input AO must be an AO with row data and an AO with column data.
    %
    % </SyntaxDescription>
    
    try
      % <SyntaxCode>
      out1 = cohere(at5, at6);
      out2 = cohere(at6, at5);
      % </SyntaxCode>
      stest = true;
    catch err
      disp(err.message)
      stest = false;
    end
    
    % <AlgoDescription>
    %
    % 1) Check that the shpe of the output data doesn't change.
    %
    % </AlgoDescription>
    
    atest = true;
    if stest
      % <AlgoCode>
      % Check the shape of the output data
      if size(out1.data.y, 2) ~= 1, atest = false; end
      if size(out2.data.y, 1) ~= 1, atest = false; end
      % </AlgoCode>
    else
      atest = false;
    end
    
    % Return a result structure
    result = utp_prepare_result(atest, stest, dbstack, mfilename);
  end % END UTP_08
  
  %% UTP_09
  
  % <TestDescription>
  %
  % Check that the cohere method pass back the output objects to a list of
  % output variables or to a single variable.
  %
  % </TestDescription>
  function result = utp_09
    
    % <SyntaxDescription>
    %
    % This test is not longer necessary because the cohere method pass back
    % always only one object.
    %
    % </SyntaxDescription>
    
    try
      % <SyntaxCode>
      % </SyntaxCode>
      stest = true;
    catch err
      disp(err.message)
      stest = false;
    end
    
    % <AlgoDescription>
    %
    % 1) Nothing to check.
    %
    % </AlgoDescription>
    
    atest = true;
    if stest
      % <AlgoCode>
      % </AlgoCode>
    else
      atest = false;
    end
    
    % Return a result structure
    result = utp_prepare_result(atest, stest, dbstack, mfilename);
  end % END UTP_09
  
  %% UTP_10
  
  % <TestDescription>
  %
  % Tests that the cohere method agrees with MATLAB's mscohere when
  % configured to use the same parameters.
  %
  % </TestDescription>
  function result = utp_10
    
    % <SyntaxDescription>
    %
    % Test that applying cohere works on two AOs.
    %
    % </SyntaxDescription>
    
    try
      % <SyntaxCode>
      % Construct two test AOs
      nsecs = 10;
      fs    = 1000;
      pl = plist('nsecs', nsecs, 'fs', fs, 'tsfcn', 'randn(size(t))');
      a1 = ao(pl); a2 = ao(pl);
      % Filter one time-series
      f2 = miir(plist('type', 'bandpass', 'fs', fs, 'order', 3, 'fc', [50 250]));
      a1f = filter(a1, plist('filter', f2));
      % make some cross-power
      a4 = a1f+a2; a4.setName;
      % Compute coherence
      Nfft = 2*fs;
      win  = specwin('Hanning', Nfft);
      pl = plist('Nfft', Nfft, 'Win', win.type, 'order', -1, 'type', 'MS');
      out = cohere(a4,a1,pl);
      % </SyntaxCode>
      stest = true;
    catch err
      disp(err.message)
      stest = false;
    end
    
    % <AlgoDescription>
    %
    % 1) Check that output agrees with the output of MATLAB's mscohere.
    % 2) Check that the shape of the output data is equal to the input data
    %
    % </AlgoDescription>
    
    atest = true;
    if stest
      % <AlgoCode>
      % Compute coherence using MATLAB's cohere
      [cxy, f] = mscohere(a4.y, a1.y, win.win, Nfft/2, Nfft, a1.fs);
      if ne(cxy(:), out.y), atest = false; end
      if ne(f,      out.x), atest = false; end
      if ne(out, out, ple2), atest = false; end
      % Check the data shape
      if size(a4.y,1) == 1
        if size(out.y,1) ~= 1, atest = false; end
      else
        if size(out.y,2) ~= 1, atest = false; end
      end
      % </AlgoCode>
    else
      atest = false;
    end
    
    % Return a result structure
    result = utp_prepare_result(atest, stest, dbstack, mfilename);
  end % END UTP_10
  
  
  %% UTP_12
  
  % <TestDescription>
  %
  % Tests symmetry properties of complex-coherence:
  % 1) white noise produced from normal pdf, with a given mean value and
  % sigma (distribution's 1st and 2nd orders)
  % 2) white noise produced from normal pdf, with a given mean value and
  % sigma (distribution's 1st and 2nd orders)
  % 3) complex coherence of the white noise series
  % 4) compare C(x,y) with conj(C(y,x))
  % 5) compare C(x,x) and C(y,y) with 1
  %
  
  % </TestDescription>
  function result = utp_12
    
    % <SyntaxDescription>
    %
    % 1) Prepare the test tsdata:
    %   white noise from normal distribution + offset
    % 2) Assign a random unit
    % 3) Prepare the test tsdata:
    %   white noise from normal distribution + offset
    % 4) Assign a random unit
    % 5) complex coherence of the white noise
    %
    % </SyntaxDescription>
    
    % <SyntaxCode>
    try
      
      % Array of parameters to pick from
      fs_list =             [0.1;1;10];
      nsecs_list =          [100:100:10000]';
      sigma_distr_list =    [1e-6 2e-3 0.25 1:0.1:10]';
      mu_distr_list =       [1e-6 2e-3 0.25 1:0.1:10]';
      
      % Build time-series test data
      
      % Picks the values at random from the list
      fs = utils.math.randelement(fs_list, 1);
      nsecs = utils.math.randelement(nsecs_list, 1);
      sigma_distr = utils.math.randelement(sigma_distr_list, 1);
      mu_distr = utils.math.randelement(mu_distr_list, 1);
      f = [1:5] / 100 * fs;
      A = sigma_distr + sigma_distr*rand(1,1);
      phi = 0 + 2*pi*rand(1,1);
      
      % White noise
      type = 'Normal';
      a_n1 = ao(plist('waveform', 'noise', ...
        'type', type, 'fs', fs, 'nsecs', nsecs, 'sigma', sigma_distr));
      a_n2 = ao(plist('waveform', 'noise', ...
        'type', type, 'fs', fs, 'nsecs', nsecs, 'sigma', sigma_distr));
      a_const = ao(mu_distr);
      a_wave = ao(plist('waveform', 'sine-wave', ...
        'fs', fs, 'nsecs', nsecs, 'f', f, 'A', A, 'phi', phi));
      a_1 = a_n1 + a_const + a_wave;
      a_2 = a_n2 + a_wave;
      
      % Set units and prefix from those supported
      unit_list = unit.supportedUnits;
      % remove the first empty unit '' from the list, because then is it
      % possible that we add a prefix to an empty unit
      unit_list = unit_list(2:end);
      prefix_list = unit.supportedPrefixes;
      a_1.setYunits(unit([cell2mat(utils.math.randelement(prefix_list,1)) cell2mat(utils.math.randelement(unit_list,1))]));
      a_2.setYunits(unit([cell2mat(utils.math.randelement(prefix_list,1)) cell2mat(utils.math.randelement(unit_list,1))]));
      
      % Evaluate the complex coherence of the time-series data
      win_list = specwin.getTypes;
      win_type = utils.math.randelement(win_list(~strcmpi(win_list, 'levelledhanning')), 1);
      win_type = win_type{1};
      if strcmp(win_type, 'Kaiser')
        win = specwin(win_type, 1, find(ao.getInfo('psd').plists, 'psll'));
      else
        win = specwin(win_type, 1);
      end
      olap = win.rov;
      detrend = 0;
      scale_type = 'C';
      n_pts = nsecs*fs/10;
      
      C12 = cohere(a_1, a_2, ...
        plist('Win', win.type, 'olap', olap, 'Nfft', n_pts, 'order', detrend, 'type', scale_type));
      C21 = cohere(a_2, a_1, ...
        plist('Win', win.type, 'olap', olap, 'Nfft', n_pts, 'order', detrend, 'type', scale_type));
      C21_cc = conj(C21);
      C11 = cohere(a_1, a_1, ...
        plist('Win', win.type, 'olap', olap, 'Nfft', n_pts, 'order', detrend, 'type', scale_type));
      C22 = cohere(a_2, a_2, ...
        plist('Win', win.type, 'olap', olap, 'Nfft', n_pts, 'order', detrend, 'type', scale_type));
      stest = true;
      
    catch err
      disp(err.message)
      stest = false;
    end
    % </SyntaxCode>
    
    % <AlgoDescription>
    %
    % 1) Check that C(x,y) equals conj(C(y,x))
    % 2) Check that C(x,x) equals 1
    % 2) Check that C(y,y) equals 1
    
    % </AlgoDescription>
    
    % <AlgoCode>
    atest = true;
    
    if stest
      if ~eq(C12.data, C21_cc.data, 'dy') || ...
          ~isequal(C11.y, ones(size(C11.y))) || ...
          ~isequal(C22.y, ones(size(C22.y)))
        atest = false;
      end
    else
      atest = false;
    end
    % </AlgoCode>
    
    % Return a result structure
    result = utp_prepare_result(atest, stest, dbstack, mfilename);
  end % END UTP_12
  
  %% UTP_13
  
  % <TestDescription>
  %
  % Tests symmetry properties of complex-coherence:
  % 1) white noise produced from normal pdf, with a given mean value and
  % sigma (distribution's 1st and 2nd orders)
  % 2) white noise produced from normal pdf, with a given mean value and
  % sigma (distribution's 1st and 2nd orders)
  % 3) magnitude-squared coherence of the white noise series
  % 4) compare C(x,y) with C(y,x)
  % 5) compare C(x,x) and C(y,y) with 1
  %
  
  % </TestDescription>
  function result = utp_13
    
    % <SyntaxDescription>
    %
    % 1) Prepare the test tsdata:
    %   white noise from normal distribution + offset
    % 2) Assign a random unit
    % 3) Prepare the test tsdata:
    %   white noise from normal distribution + offset
    % 4) Assign a random unit
    % 5) magnitude-squared coherence of the white noise
    %
    % </SyntaxDescription>
    
    % <SyntaxCode>
    try
      
      % Array of parameters to pick from
      fs_list =             [0.1;1;10];
      nsecs_list =          [100:100:10000]';
      sigma_distr_list =    [1e-6 2e-3 0.25 1:0.1:10]';
      mu_distr_list =       [1e-6 2e-3 0.25 1:0.1:10]';
      
      % Build time-series test data
      
      % Picks the values at random from the list
      fs = utils.math.randelement(fs_list, 1);
      nsecs = utils.math.randelement(nsecs_list, 1);
      sigma_distr = utils.math.randelement(sigma_distr_list, 1);
      mu_distr = utils.math.randelement(mu_distr_list, 1);
      f = [1:5] / 100 * fs;
      A = sigma_distr + sigma_distr*rand(1,1);
      phi = 0 + 2*pi*rand(1,1);
      
      % White noise
      type = 'Normal';
      a_n1 = ao(plist('waveform', 'noise', ...
        'type', type, 'fs', fs, 'nsecs', nsecs, 'sigma', sigma_distr));
      a_n2 = ao(plist('waveform', 'noise', ...
        'type', type, 'fs', fs, 'nsecs', nsecs, 'sigma', sigma_distr));
      a_const = ao(mu_distr);
      a_wave = ao(plist('waveform', 'sine-wave', ...
        'fs', fs, 'nsecs', nsecs, 'f', f, 'A', A, 'phi', phi));
      a_1 = a_n1 + a_const + a_wave;
      a_2 = a_n2 + a_wave;
      
      % Set units and prefix from those supported
      unit_list = unit.supportedUnits;
      % remove the first empty unit '' from the list, because then is it
      % possible that we add a prefix to an empty unit
      unit_list = unit_list(2:end);
      prefix_list = unit.supportedPrefixes;
      a_1.setYunits(unit([cell2mat(utils.math.randelement(prefix_list,1)) cell2mat(utils.math.randelement(unit_list,1))]));
      a_2.setYunits(unit([cell2mat(utils.math.randelement(prefix_list,1)) cell2mat(utils.math.randelement(unit_list,1))]));
      
      % Evaluate the magnitude-squared coherence of the time-series data
      win_list = specwin.getTypes;
      win_type = utils.math.randelement(win_list(~strcmpi(win_list, 'levelledhanning')), 1);
      win_type = win_type{1};
      if strcmp(win_type, 'Kaiser')
        win = specwin(win_type, 1, find(ao.getInfo('psd').plists, 'psll'));
      else
        win = specwin(win_type, 1);
      end
      olap = win.rov;
      detrend = 0;
      scale_type = 'MS';
      n_pts = nsecs*fs/10;
      
      C12 = cohere(a_1, a_2, ...
        plist('Win', win.type, 'olap', olap, 'Nfft', n_pts, 'order', detrend, 'type', scale_type));
      C21 = cohere(a_2, a_1, ...
        plist('Win', win.type, 'olap', olap, 'Nfft', n_pts, 'order', detrend, 'type', scale_type));
      C11 = cohere(a_1, a_1, ...
        plist('Win', win.type, 'olap', olap, 'Nfft', n_pts, 'order', detrend, 'type', scale_type));
      C22 = cohere(a_2, a_2, ...
        plist('Win', win.type, 'olap', olap, 'Nfft', n_pts, 'order', detrend, 'type', scale_type));
      stest = true;
      
    catch err
      disp(err.message)
      stest = false;
    end
    % </SyntaxCode>
    
    % <AlgoDescription>
    %
    % 1) Check that C(x,y) equals C(y,x)
    % 1) Check that C(x,x) equals 1
    % 1) Check that C(y,y) equals 1
    
    % </AlgoDescription>
    
    % <AlgoCode>
    atest = true;
    
    if stest
      if ~isequal(C12.data, C21.data) || ...
          ~isequal(C11.y, ones(size(C11.y))) ...
          || ~isequal(C22.y, ones(size(C22.y)))
        atest = false;
      end
      if atest == false
        fs
        nsecs
        sigma_distr
        mu_distr
        f
        A
        phi
      end
    else
      atest = false;
    end
    % </AlgoCode>
    
    % Return a result structure
    result = utp_prepare_result(atest, stest, dbstack, mfilename);
  end % END UTP_13
  
  %% UTP_14
  
  % <TestDescription>
  %
  % Tests symmetry properties of complex-coherence:
  % 1) white noise produced from normal pdf, with a given mean value and
  % sigma (distribution's 1st and 2nd orders)
  % 2) white noise produced from normal pdf, with a given mean value and
  % sigma (distribution's 1st and 2nd orders)
  % 3) complex coherence of the combination of white noise series
  % 4) compare C(x,y) with 1
  %
  
  % </TestDescription>
  function result = utp_14
    
    % <SyntaxDescription>
    %
    % 1) Prepare the test tsdata:
    %   white noise from normal distribution + offset
    % 2) Assign a random unit
    % 3) Prepare the test tsdata:
    %   white noise from normal distribution + offset
    % 4) Assign a random unit
    % 5) complex coherence of the combination of noise
    %
    % </SyntaxDescription>
    
    % <SyntaxCode>
    try
      
      % Array of parameters to pick from
      fs_list =             [0.1;1;10];
      nsecs_list =          [100:100:10000]';
      sigma_distr_list =    [1e-6 2e-3 0.25 1:0.1:10]';
      mu_distr_list =       [1e-6 2e-3 0.25 1:0.1:10]';
      
      % Build time-series test data
      
      % Picks the values at random from the list
      fs = utils.math.randelement(fs_list, 1);
      nsecs = utils.math.randelement(nsecs_list, 1);
      sigma_distr = utils.math.randelement(sigma_distr_list, 1);
      mu_distr = utils.math.randelement(mu_distr_list, 1);
      f = [1:5] / 100 * fs;
      A = sigma_distr + sigma_distr*rand(1,1);
      phi = 0 + 2*pi*rand(1,1);
      
      % White noise
      type = 'Normal';
      a_n = ao(plist('waveform', 'noise', ...
        'type', type, 'fs', fs, 'nsecs', nsecs, 'sigma', sigma_distr));
      a_const = ao(mu_distr);
      % Sinusoidal signal
      a_wave = ao(plist('waveform', 'sine-wave', ...
        'fs', fs, 'nsecs', nsecs, 'f', f, 'A', A, 'phi', phi));
      a_1 = a_n + a_wave;
      % Linear combination (totally correlated time series)
      a_2 = a_1 + a_const;
      
      % Set units and prefix from those supported
      unit_list = unit.supportedUnits;
      % remove the first empty unit '' from the list, because then is it
      % possible that we add a prefix to an empty unit
      unit_list = unit_list(2:end);
      prefix_list = unit.supportedPrefixes;
      a_1.setYunits(unit([cell2mat(utils.math.randelement(prefix_list,1)) cell2mat(utils.math.randelement(unit_list,1))]));
      a_2.setYunits(unit([cell2mat(utils.math.randelement(prefix_list,1)) cell2mat(utils.math.randelement(unit_list,1))]));
      
      % Evaluate the complex coherence of the time-series data
      win_list = specwin.getTypes;
      win_type = utils.math.randelement(win_list(~strcmpi(win_list, 'levelledhanning')), 1);
      win_type = win_type{1};
      if strcmp(win_type, 'Kaiser')
        win = specwin(win_type, 1, find(ao.getInfo('psd').plists, 'psll'));
      else
        win = specwin(win_type, 1);
      end
      olap = win.rov;
      detrend = 0;
      scale_type = 'C';
      n_pts = nsecs*fs/10;
      
      C = cohere(a_1, a_2, ...
        plist('Win', win.type, 'olap', olap, 'Nfft', n_pts, 'order', detrend, 'type', scale_type));
      stest = true;
      
    catch err
      disp(err.message)
      stest = false;
    end
    % </SyntaxCode>
    
    % <AlgoDescription>
    %
    % 1) Check that the complex coherence equals 1
    
    % </AlgoDescription>
    
    % <AlgoCode>
    atest = true;
    TOL = 1e-12;
    
    if stest
      if any(abs((C.y - 1)) > TOL)
        atest = false;
      end
    else
      atest = false;
    end
    % </AlgoCode>
    
    % Return a result structure
    result = utp_prepare_result(atest, stest, dbstack, mfilename);
  end % END UTP_14
  
  %% UTP_15
  
  % <TestDescription>
  %
  % Tests symmetry properties of complex-coherence:
  % 1) white noise produced from normal pdf, with a given mean value and
  % sigma (distribution's 1st and 2nd orders)
  % 2) white noise produced from normal pdf, with a given mean value and
  % sigma (distribution's 1st and 2nd orders)
  % 3) magnitude-squared coherence of the combination of white noise series
  % 4) compare C(x,y) with 1
  %
  
  % </TestDescription>
  function result = utp_15
    
    % <SyntaxDescription>
    %
    % 1) Prepare the test tsdata:
    %   white noise from normal distribution + offset
    % 2) Assign a random unit
    % 3) Prepare the test tsdata:
    %   white noise from normal distribution + offset
    % 4) Assign a random unit
    % 5) magnitude-squared coherence of the combination of noise
    %
    % </SyntaxDescription>
    
    % <SyntaxCode>
    try
      
      % Array of parameters to pick from
      fs_list =             [0.1;1;10];
      nsecs_list =          [100:100:10000]';
      sigma_distr_list =    [1e-6 2e-3 0.25 1:0.1:10]';
      mu_distr_list =       [1e-6 2e-3 0.25 1:0.1:10]';
      
      % Build time-series test data
      
      % Picks the values at random from the list
      fs = utils.math.randelement(fs_list, 1);
      nsecs = utils.math.randelement(nsecs_list, 1);
      sigma_distr = utils.math.randelement(sigma_distr_list, 1);
      mu_distr = utils.math.randelement(mu_distr_list, 1);
      f = [1:5] / 100 * fs;
      A = sigma_distr + sigma_distr*rand(1,1);
      phi = 0 + 2*pi*rand(1,1);
      
      % White noise
      type = 'Normal';
      a_n = ao(plist('waveform', 'noise', ...
        'type', type, 'fs', fs, 'nsecs', nsecs, 'sigma', sigma_distr));
      a_const = ao(mu_distr);
      % Sinusoidal signal
      a_wave = ao(plist('waveform', 'sine-wave', ...
        'fs', fs, 'nsecs', nsecs, 'f', f, 'A', A, 'phi', phi));
      a_1 = a_n + a_wave;
      % Linear combination (totally correlated time series)
      a_2 = a_1 + a_const;
      
      % Set units and prefix from those supported
      unit_list = unit.supportedUnits;
      % remove the first empty unit '' from the list, because then is it
      % possible that we add a prefix to an empty unit
      unit_list = unit_list(2:end);
      prefix_list = unit.supportedPrefixes;
      a_1.setYunits(unit([cell2mat(utils.math.randelement(prefix_list,1)) cell2mat(utils.math.randelement(unit_list,1))]));
      a_2.setYunits(unit([cell2mat(utils.math.randelement(prefix_list,1)) cell2mat(utils.math.randelement(unit_list,1))]));
      
      % Evaluate the complex coherence of the time-series data
      win_list = specwin.getTypes;
      win_type = utils.math.randelement(win_list(~strcmpi(win_list, 'levelledhanning')), 1);
      win_type = win_type{1};
      if strcmp(win_type, 'Kaiser')
        win = specwin(win_type, 1, find(ao.getInfo('psd').plists, 'psll'));
      else
        win = specwin(win_type, 1);
      end
      olap = win.rov;
      detrend = 0;
      scale_type = 'MS';
      n_pts = nsecs*fs/10;
      
      C = cohere(a_1, a_2, ...
        plist('Win', win.type, 'olap', olap, 'Nfft', n_pts, 'order', detrend, 'type', scale_type));
      stest = true;
      
    catch err
      disp(err.message)
      stest = false;
    end
    % </SyntaxCode>
    
    % <AlgoDescription>
    %
    % 1) Check that the magnitude-squared coherence equals 1
    
    % </AlgoDescription>
    
    % <AlgoCode>
    atest = true;
    
    if stest
      if ~eq(C.y, ones(size(C.y)))
        atest = false;
      end
    else
      atest = false;
    end
    % </AlgoCode>
    
    % Return a result structure
    result = utp_prepare_result(atest, stest, dbstack, mfilename);
  end % END UTP_15
  
  %% UTP_16
  
  % <TestDescription>
  %
  % Tests symmetry properties of complex-coherence:
  % 1) white noise produced from normal pdf, with a given mean value and
  % sigma (distribution's 1st and 2nd orders)
  % 2) white noise produced from normal pdf, with a given mean value and
  % sigma (distribution's 1st and 2nd orders)
  % 3) magnitude-squared coherence M of the combination of white noise series
  % 4) complex coherence C of the combination of white noise series
  % 5) compare abs(C)^2 with M
  %
  
  % </TestDescription>
  function result = utp_16
    
    % <SyntaxDescription>
    %
    % 1) Prepare the test tsdata:
    %   white noise from normal distribution + offset
    % 2) Assign a random unit
    % 3) Prepare the test tsdata:
    %   white noise from normal distribution + offset
    % 4) Assign a random unit
    % 5) magnitude-squared coherence of the combination of noise
    % 6) complex coherence of the combination of noise
    %
    % </SyntaxDescription>
    
    % <SyntaxCode>
    try
      
      % Array of parameters to pick from
      fs_list =             [0.1;1;10];
      nsecs_list =          [100:100:10000]';
      sigma_distr_list =    [1e-6 2e-3 0.25 1:0.1:10]';
      mu_distr_list =       [1e-6 2e-3 0.25 1:0.1:10]';
      
      % Build time-series test data
      
      % Picks the values at random from the list
      fs = utils.math.randelement(fs_list, 1);
      nsecs = utils.math.randelement(nsecs_list, 1);
      sigma_distr = utils.math.randelement(sigma_distr_list, 1);
      mu_distr = utils.math.randelement(mu_distr_list, 1);
      f = [1:5] / 100 * fs;
      A = sigma_distr + sigma_distr*rand(1,1);
      phi = 0 + 2*pi*rand(1,1);
      
      % White noise
      type = 'Normal';
      a_n = ao(plist('waveform', 'noise', ...
        'type', type, 'fs', fs, 'nsecs', nsecs, 'sigma', sigma_distr));
      a_const = ao(mu_distr);
      % Sinusoidal signal
      a_wave = ao(plist('waveform', 'sine-wave', ...
        'fs', fs, 'nsecs', nsecs, 'f', f, 'A', A, 'phi', phi));
      a_1 = a_n + a_wave;
      % Linear combination (totally correlated time series)
      a_2 = a_1 + a_const;
      
      % Set units and prefix from those supported
      unit_list = unit.supportedUnits;
      % remove the first empty unit '' from the list, because then is it
      % possible that we add a prefix to an empty unit
      unit_list = unit_list(2:end);
      prefix_list = unit.supportedPrefixes;
      a_1.setYunits(unit([cell2mat(utils.math.randelement(prefix_list,1)) cell2mat(utils.math.randelement(unit_list,1))]));
      a_2.setYunits(unit([cell2mat(utils.math.randelement(prefix_list,1)) cell2mat(utils.math.randelement(unit_list,1))]));
      
      % Evaluate the complex coherence of the time-series data
      win_list = specwin.getTypes;
      win_type = utils.math.randelement(win_list(~strcmpi(win_list, 'levelledhanning')), 1);
      win_type = win_type{1};
      if strcmp(win_type, 'Kaiser')
        win = specwin(win_type, 1, find(ao.getInfo('psd').plists, 'psll'));
      else
        win = specwin(win_type, 1);
      end
      olap = win.rov;
      detrend = 0;
      n_pts = nsecs*fs/10;
      
      M = cohere(a_1, a_2, ...
        plist('Win', win.type, 'olap', olap, 'Nfft', n_pts, 'order', detrend, 'type', 'MS'));
      C = cohere(a_1, a_2, ...
        plist('Win', win.type, 'olap', olap, 'Nfft', n_pts, 'order', detrend, 'type', 'C'));
      stest = true;
      
    catch err
      disp(err.message)
      stest = false;
    end
    % </SyntaxCode>
    
    % <AlgoDescription>
    %
    % 1) Check that the magnitude-squared coherence equals the square
    % modulus of the complex coherence
    
    % </AlgoDescription>
    
    % <AlgoCode>
    atest = true;
    TOL = 1e-15;
    
    if stest
      if any(abs(M.y - abs(C.y).^2) > TOL)
        atest = false;
      end
    else
      atest = false;
    end
    % </AlgoCode>
    
    % Return a result structure
    result = utp_prepare_result(atest, stest, dbstack, mfilename);
  end % END UTP_16
  
  %% UTP_17
  
  % <TestDescription>
  %
  % Tests handling of units:
  % 1) white noise produced from normal pdf, with a given mean value and
  % sigma (distribution's 1st and 2nd orders)
  % 2) white noise produced from normal pdf, with a given mean value and
  % sigma (distribution's 1st and 2nd orders)
  % 3) complex coherence of the white noise series
  % 4) compares the units of the input and output
  %
  
  % </TestDescription>
  function result = utp_17
    
    % <SyntaxDescription>
    %
    % 1) Prepare the test tsdata:
    %   white noise from normal distribution + offset
    % 2) Assign a random unit
    % 3) Prepare the test tsdata:
    %   white noise from normal distribution + offset
    % 4) Assign a random unit
    % 5) complex cohere of the white noise
    %
    % </SyntaxDescription>
    
    % <SyntaxCode>
    try
      
      % Build time-series test data
      fs = 1;
      nsecs = 86400;
      sigma_distr_1 = 4.69e-12;
      mu_distr_1 = -5.11e-14;
      sigma_distr_2 = 6.04e-9;
      mu_distr_2 = 1.5e-10;
      
      % White noise
      type = 'Normal';
      
      a_n = ao(plist('waveform', 'noise', ...
        'type', type, 'fs', fs, 'nsecs', nsecs, 'sigma', sigma_distr_1));
      a_const = ao(mu_distr_1);
      a_1 = a_n + a_const;
      
      a_n = ao(plist('waveform', 'noise', ...
        'type', type, 'fs', fs, 'nsecs', nsecs, 'sigma', sigma_distr_2));
      a_const = ao(mu_distr_2);
      a_2 = a_n + a_const;
      
      % Set units and prefix from those supported
      unit_list = unit.supportedUnits;
      % remove the first empty unit '' from the list, because then is it
      % possible that we add a prefix to an empty unit
      unit_list = unit_list(2:end);
      prefix_list = unit.supportedPrefixes;
      a_1.setYunits(unit([cell2mat(utils.math.randelement(prefix_list,1)) cell2mat(utils.math.randelement(unit_list,1))]));
      a_2.setYunits(unit([cell2mat(utils.math.randelement(prefix_list,1)) cell2mat(utils.math.randelement(unit_list,1))]));
      
      % Evaluate the coherence of the time-series data
      win = specwin('BH92');
      olap = win.rov;
      detrend = 0;
      scale_type = 'C';
      n_pts = nsecs*fs/10;
      
      C = cohere(a_1, a_2, ...
        plist('Win', win.type, 'olap', olap, 'Nfft', n_pts, 'order', detrend, 'type', scale_type));
      
      stest = true;
      
    catch err
      disp(err.message)
      stest = false;
    end
    % </SyntaxCode>
    
    % <AlgoDescription>
    %
    % 1) Check that (complex coherence yunits) equals [1]
    % 2) Check that (complex coherence xunits) equals [Hz]
    
    % </AlgoDescription>
    
    % <AlgoCode>
    atest = true;
    
    if stest
      if ~eq(C.yunits, unit('')) || ~eq(C.xunits, unit('Hz'))
        atest = false;
      end
    else
      atest = false;
    end
    % </AlgoCode>
    
    % Return a result structure
    result = utp_prepare_result(atest, stest, dbstack, mfilename);
  end % END UTP_17
  
  %% UTP_18
  
  % <TestDescription>
  %
  % Tests handling of units:
  % 1) white noise produced from normal pdf, with a given mean value and
  % sigma (distribution's 1st and 2nd orders)
  % 2) white noise produced from normal pdf, with a given mean value and
  % sigma (distribution's 1st and 2nd orders)
  % 3) magnitude-squared coherence of the white noise series
  % 4) compares the units of the input and output
  %
  
  % </TestDescription>
  function result = utp_18
    
    % <SyntaxDescription>
    %
    % 1) Prepare the test tsdata:
    %   white noise from normal distribution + offset
    % 2) Assign a random unit
    % 3) Prepare the test tsdata:
    %   white noise from normal distribution + offset
    % 4) Assign a random unit
    % 5) magnitude-squared cohere of the white noise
    %
    % </SyntaxDescription>
    
    % <SyntaxCode>
    try
      
      % Build time-series test data
      fs = 1;
      nsecs = 86400;
      sigma_distr_1 = 4.69e-12;
      mu_distr_1 = -5.11e-14;
      sigma_distr_2 = 6.04e-9;
      mu_distr_2 = 1.5e-10;
      
      % White noise
      type = 'Normal';
      
      a_n = ao(plist('waveform', 'noise', ...
        'type', type, 'fs', fs, 'nsecs', nsecs, 'sigma', sigma_distr_1));
      a_const = ao(mu_distr_1);
      a_1 = a_n + a_const;
      
      a_n = ao(plist('waveform', 'noise', ...
        'type', type, 'fs', fs, 'nsecs', nsecs, 'sigma', sigma_distr_2));
      a_const = ao(mu_distr_2);
      a_2 = a_n + a_const;
      
      % Set units and prefix from those supported
      unit_list = unit.supportedUnits;
      % remove the first empty unit '' from the list, because then is it
      % possible that we add a prefix to an empty unit
      unit_list = unit_list(2:end);
      prefix_list = unit.supportedPrefixes;
      a_1.setYunits(unit([cell2mat(utils.math.randelement(prefix_list,1)) cell2mat(utils.math.randelement(unit_list,1))]));
      a_2.setYunits(unit([cell2mat(utils.math.randelement(prefix_list,1)) cell2mat(utils.math.randelement(unit_list,1))]));
      
      % Evaluate the coherence of the time-series data
      win = specwin('BH92');
      olap = win.rov;
      detrend = 0;
      scale_type = 'MS';
      n_pts = nsecs*fs/10;
      
      C = cohere(a_1, a_2, ...
        plist('Win', win.type, 'olap', olap, 'Nfft', n_pts, 'order', detrend,'type', scale_type));
      
      stest = true;
      
    catch err
      disp(err.message)
      stest = false;
    end
    % </SyntaxCode>
    
    % <AlgoDescription>
    %
    % 1) Check that (magnitude-squared coherence yunits) equals [1]
    % 2) Check that (magnitude-squared coherence xunits) equals [Hz]
    
    % </AlgoDescription>
    
    % <AlgoCode>
    atest = true;
    
    if stest
      if ~eq(C.yunits, unit('')) || ~eq(C.xunits, unit('Hz'))
        atest = false;
      end
    else
      atest = false;
    end
    % </AlgoCode>
    
    % Return a result structure
    result = utp_prepare_result(atest, stest, dbstack, mfilename);
  end % END UTP_18
  
  %% UTP_19
  
  % <TestDescription>
  %
  % Tests that differently sized data sets are treated properly
  %
  % </TestDescription>
  function result = utp_19
    
    % <SyntaxDescription>
    %
    % Test that applying cohere works on two AOs.
    %
    % </SyntaxDescription>
    
    try
      % <SyntaxCode>
      % Construct two test AOs
      nsecs = [10000:1:20000];
      fs    = 1;
      pl = plist('fs', fs, 'tsfcn', 'randn(size(t))');
      a1 = ao(pl.pset('nsecs', utils.math.randelement(nsecs, 1)));
      a2 = ao(pl.pset('nsecs', utils.math.randelement(nsecs, 1)));
      len_1 = a1.len;
      len_2 = a2.len;
      % Filter one time-series
      f2 = miir(plist('type', 'bandpass', 'fs', fs, 'order', 3, 'fc', [.050 .25]));
      a1f = filter(a1, plist('filter', f2));
      % Compute cohere
      Nfft = -1;
      win  = 'Hanning';
      pl = plist('Nfft', Nfft, 'Win', win, 'order', -1);
      out = cohere(a2,a1f,pl);
      % </SyntaxCode>
      stest = true;
    catch err
      disp(err.message)
      stest = false;
    end
    
    % <AlgoDescription>
    %
    % 1) Check that cohere used the length of the shortest ao.
    %
    % </AlgoDescription>
    
    atest = true;
    if stest
      % <AlgoCode>
      % Compare the nfft with the length of the input data
      
      if out.x(2) ~= 1/min(len_1,len_2)
        atest = false;
      end
      % </AlgoCode>
    else
      atest = false;
    end
    
    % Return a result structure
    result = utp_prepare_result(atest, stest, dbstack, mfilename);
  end % END UTP_19
  
  %% UTP_20
  
  % <TestDescription>
  %
  % Tests that applying a single window the coherence is 1
  %
  % </TestDescription>
  function result = utp_20
    
    % <SyntaxDescription>
    %
    % Test that applying cohere works on two AOs.
    %
    % </SyntaxDescription>
    
    try
      % <SyntaxCode>
      % Construct two test AOs
      nsecs = [10000:100:20000];
      fs    = 1;
      pl = plist('fs', fs, 'tsfcn', 'randn(size(t))');
      a1 = ao(pl.pset('nsecs', utils.math.randelement(nsecs, 1)));
      a2 = ao(pl.pset('nsecs', utils.math.randelement(nsecs, 1)));
      % Filter one time-series
      f2 = miir(plist('type', 'bandpass', 'fs', fs, 'order', 3, 'fc', [.050 .25]));
      a1f = filter(a1, plist('filter', f2));
      % Compute cohere
      Nfft = -1;
      win  = 'Hanning';
      pl = plist('Nfft', Nfft, 'Win', win, 'order', -1);
      out_c = cohere(a2, a1f, pl.pset('type', 'C'));
      out_ms = cohere(a2, a1f, pl.pset('type', 'MS'));
      % </SyntaxCode>
      stest = true;
    catch err
      disp(err.message)
      stest = false;
    end
    
    % <AlgoDescription>
    %
    % 1) Check that the calculated cohere is 1
    %
    % </AlgoDescription>
    
    atest = true;
    TOL = 1e-12;
    if stest
      % <AlgoCode>
      % Compare the calculated cohere with 1
      
      if any(abs(abs(out_c.y) - 1) > TOL)
        atest = false;
      end
      if any(abs(abs(out_ms.y) - 1) > TOL)
        atest = false;
      end
      % </AlgoCode>
    else
      atest = false;
    end
    
    % Return a result structure
    result = utp_prepare_result(atest, stest, dbstack, mfilename);
  end % END UTP_20
  
  %% UTP_21
  
  % <TestDescription>
  %
  % Tests the possibility to set the number of averages rather than setting the Nfft:
  % 1) white noise produced from normal pdf, with:
  %   a given mean value and sigma (distribution's 1st and 2nd order)
  % 2) cohere of the noise, without detrending, random window, set number of
  %   averages
  % 3) check the effective number of averages
  %
  
  % </TestDescription>
  function result = utp_21
    
    % <SyntaxDescription>
    %
    % 1) Prepare the test tsdata:
    %   white noise from normal distribution + offset
    % 2) cohere of the noise, without detrending, random window, set number of
    %   averages
    %
    % </SyntaxDescription>
    
    % <SyntaxCode>
    try
      % Array of parameters to pick from
      fs_list =             [0.1;1;2;5;10];
      nsecs_list =          [2000:1000:10000]';
      sigma_distr_list =    [1e-6 2e-3 0.25 1:0.1:10]';
      trend_0_list =        [1e-6 2e-3 0.25 1:0.1:10]';
      
      % Build time-series test data
      
      % Picks the values at random from the list
      fs = utils.math.randelement(fs_list, 1);
      nsecs = utils.math.randelement(nsecs_list, 1);
      sigma_distr = utils.math.randelement(sigma_distr_list, 1);
      trend_0 = utils.math.randelement(trend_0_list, 1);
      
      % White noise
      type = 'Normal';
      a_n1 = ao(plist('waveform', 'noise', ...
        'type', type, 'fs', fs, 'nsecs', nsecs, 'sigma', sigma_distr));
      a_n2 = ao(plist('waveform', 'noise', ...
        'type', type, 'fs', fs, 'nsecs', nsecs, 'sigma', sigma_distr));
      
      % Constant signal
      a_c = ao(trend_0);
      
      % Total signals
      a1 = a_n1 + a_c;
      a2 = a_n2 + a_c;
      
      % Evaluate the complex coherence of the white noise time-series data
      win_list = specwin.getTypes;
      win_type = utils.math.randelement(win_list(~strcmpi(win_list, 'levelledhanning')), 1);
      win_type = win_type{1};
      switch win_type
        case 'Kaiser'
          win = specwin(win_type, 1, find(ao.getInfo('psd').plists, 'psll'));
        otherwise
          win = specwin(win_type, 1);
      end
      
      olap = win.rov;
      detrend = 0;
      n_pts = -1;
      scale_type = 'C';
      navs = utils.math.randelement([1:100],1);
      
      % Evaluates the coherence asking for the number of averages
      C = cohere(a1, a2, plist('Win', win.type, 'olap', olap, ...
        'Nfft', n_pts, 'order', detrend, 'type', scale_type, 'navs', navs));
      
      stest = true;
      
    catch err
      disp(err.message)
      stest = false;
    end
    % </SyntaxCode>
    
    % <AlgoDescription>
    %
    % 1) Check that calculated navs are identical to those requested
    %
    % </AlgoDescription>
    
    % <AlgoCode>
    atest = true;
    
    if stest
      % Compare the navs written in the output object with the requested one
      if ne(navs, C.data.navs)
        if ne(find(C.hist.plistUsed, 'navs'), C.data.navs)
          atest = false;
        end
      end
    else
      atest = false;
    end
    % </AlgoCode>
    
    % Return a result structure
    result = utp_prepare_result(atest, stest, dbstack, mfilename);
  end % END UTP_21
  
  %% UTP_22
  
  % <TestDescription>
  %
  % Tests the possibility to set the number of averages rather than setting the Nfft:
  % 1) white noise produced from uniform pdf, with:
  %   a given mean value and sigma (distribution's 1st and 2nd order)
  % 2) cohere of the noise, without detrending, random window, random navs
  % 3) get the number of averages
  % 4) get the nfft used
  % 5) run cohere again, with the nfft used
  % 6) compare the calculated objects
  %
  
  % </TestDescription>
  function result = utp_22
    
    % <SyntaxDescription>
    %
    % 1) white noise produced from uniform pdf, with:
    %   a given mean value and sigma (distribution's 1st and 2nd order)
    % 2) cohere of the noise, without detrending, random window, random navs
    % 3) get the number of averages
    % 4) get the nfft used
    % 5) run cohere again, with the nfft used
    %
    % </SyntaxDescription>
    
    % <SyntaxCode>
    try
      % Array of parameters to pick from
      fs_list =             [0.1;1;2;5;10];
      nsecs_list =          [20 100 1000:1000:10000]';
      sigma_distr_list =    [1e-6 2e-3 0.25 1:0.1:10]';
      trend_0_list =        [1e-6 2e-3 0.25 1:0.1:10]';
      
      % Build time-series test data
      
      % Picks the values at random from the list
      fs = utils.math.randelement(fs_list, 1);
      nsecs = utils.math.randelement(nsecs_list, 1);
      sigma_distr = utils.math.randelement(sigma_distr_list, 1);
      trend_0 = utils.math.randelement(trend_0_list, 1);
      
      % White noise
      type = 'Uniform';
      a_n1 = ao(plist('waveform', 'noise', ...
        'type', type, 'fs', fs, 'nsecs', nsecs, 'sigma', sigma_distr));
      a_n2 = ao(plist('waveform', 'noise', ...
        'type', type, 'fs', fs, 'nsecs', nsecs, 'sigma', sigma_distr));
      
      % Constant signal
      a_c = ao(trend_0);
      
      % Total signals
      a1 = a_n1 + a_c;
      a2 = a_n2 + a_c;
      
      % Evaluate the complex coherence of the white noise time-series data
      win_list = specwin.getTypes;
      win_type = utils.math.randelement(win_list(~strcmpi(win_list, 'levelledhanning')), 1);
      win_type = win_type{1};
      switch win_type
        case 'Kaiser'
          win = specwin(win_type, 1, find(ao.getInfo('psd').plists, 'psll'));
        otherwise
          win = specwin(win_type, 1);
      end
      
      olap = win.rov;
      detrend = 0;
      scale_type = 'MS';
      navs = fix(utils.math.randelement(logspace(0,log10(max(0,a1.len/10)),50),1));
      
      % Calculates the coherence asking for the number of averages
      C1 = cohere(a1, a2, plist('Win', win.type, 'olap', olap, ...
        'Nfft', -1, 'order', detrend, 'type', scale_type, ...
        'navs', navs));
      
      % Calculates the coherence asking for the number of points just evaluated
      C2 = cohere(a1, a2, plist('Win', win.type, 'olap', olap, ...
        'Nfft', find(C1.hist.plistUsed, 'Nfft'), 'order', detrend, 'type', scale_type));
      stest = true;
      
    catch err
      disp(err.message)
      stest = false;
    end
    % </SyntaxCode>
    
    % <AlgoDescription>
    %
    % 1) Check that calculated objects C1 and C2 are identical
    %
    % </AlgoDescription>
    
    % <AlgoCode>
    atest = true;
    
    if stest
      % Compare the output objects
      if ne(C1,C2,ple3)
        atest = false;
      end
    else
      atest = false;
    end
    % </AlgoCode>
    
    % Return a result structure
    result = utp_prepare_result(atest, stest, dbstack, mfilename);
  end % END UTP_22
  
  %% UTP_23
  
  % <TestDescription>
  %
  % Tests the possibility to set the number of averages rather than setting the Nfft:
  % 1) white noise produced from normal pdf, with:
  %   a given mean value and sigma (distribution's 1st and 2nd order)
  % 2) cohere of the noise, without detrending, random window, random navs
  % 3) get the number of averages
  % 4) get the nfft used
  % 5) run cohere again, with the nfft used
  % 6) compare navs, nfft, coheres
  %
  
  % </TestDescription>
  function result = utp_23
    
    % <SyntaxDescription>
    %
    % 1) white noise produced from normal pdf, with:
    %   a given mean value and sigma (distribution's 1st and 2nd order)
    % 2) cohere of the noise, without detrending, random window, random navs
    % 3) get the number of averages
    % 4) get the nfft used
    % 5) run cohere again, with the nfft used
    % 6) run cohere again, with conflicting parameters, and verify it uses
    %     nfft rather than navs
    %
    % </SyntaxDescription>
    
    % <SyntaxCode>
    try
      % Array of parameters to pick from
      fs_list =             [0.1;1;2;5;10];
      nsecs_list =          [1000:1000:10000]';
      sigma_distr_list =    [1e-6 2e-3 0.25 1:0.1:10]';
      trend_0_list =        [1e-6 2e-3 0.25 1:0.1:10]';
      
      % Build time-series test data
      
      % Picks the values at random from the list
      fs = utils.math.randelement(fs_list, 1);
      nsecs = utils.math.randelement(nsecs_list, 1);
      sigma_distr = utils.math.randelement(sigma_distr_list, 1);
      trend_0 = utils.math.randelement(trend_0_list, 1);
      
      % White noise
      type = 'Normal';
      a_n1 = ao(plist('waveform', 'noise', ...
        'type', type, 'fs', fs, 'nsecs', nsecs, 'sigma', sigma_distr));
      a_n2 = ao(plist('waveform', 'noise', ...
        'type', type, 'fs', fs, 'nsecs', nsecs, 'sigma', sigma_distr));
      
      % Constant signal
      a_c = ao(trend_0);
      
      % Total signals
      a1 = a_n1 + a_c;
      a2 = a_n2 + a_c;
      
      % Evaluate the complex coherence of the white noise time-series data
      win_list = specwin.getTypes;
      win_type = utils.math.randelement(win_list(~strcmpi(win_list, 'levelledhanning')), 1);
      win_type = win_type{1};
      switch win_type
        case 'Kaiser'
          win = specwin(win_type, 1, find(ao.getInfo('psd').plists, 'psll'));
        otherwise
          win = specwin(win_type, 1);
      end
      
      olap = win.rov;
      detrend = 0;
      scale_type = 'C';
      navs = fix(utils.math.randelement(logspace(0,log10(max(a1.len/10,0)),50),1));
      
      % Calculates the coherence asking for the number of averages
      C1 = cohere(a1, a2, plist('Win', win.type, 'olap', olap, ...
        'Nfft', -1, 'order', detrend, 'type', scale_type, ...
        'navs', navs));
      
      npts_2 = find(C1.hist.plistUsed, 'Nfft');
      % Calculates the coherence asking for the number of points
      C2 = cohere(a1, a2, plist('Win', win.type, 'olap', olap, ...
        'Nfft', npts_2, 'order', detrend, 'type', scale_type));
      
      npts_3 = fix(npts_2/2);
      % Calculates the coherence asking for the number of points AND the window length
      C3 = cohere(a1, a2, plist('Win', win.type, 'olap', olap, ...
        'Nfft', npts_3, ...
        'order', detrend, 'type', scale_type, ...
        'navs', navs));
      
      stest = true;
      
    catch err
      disp(err.message)
      stest = false;
    end
    % </SyntaxCode>
    
    % <AlgoDescription>
    %
    % 1) Check that calculated objects C1 and C2 are identical
    % 2) Check that C3 used different values
    %
    % </AlgoDescription>
    
    % <AlgoCode>
    atest = true;
    
    if stest
      % Compare the navs written in the output object with the requested one
      if  ne(C1,C2,ple3) || ...
          ne(find(C3.hist.plistUsed, 'Nfft'), npts_3) || eq(C3.data.navs, navs)
        atest = false;
      end
    else
      atest = false;
    end
    % </AlgoCode>
    
    % Return a result structure
    result = utp_prepare_result(atest, stest, dbstack, mfilename);
  end % END UTP_23
  
  %% UTP_24
  
  % <TestDescription>
  %
  % Tests that the cohere method agrees with MATLAB's mscohere when
  % configured to use the same parameters.
  %
  % </TestDescription>
  function result = utp_24
    
    % <SyntaxDescription>
    %
    % Test that the applying cohere works on two AOs.
    %
    % </SyntaxDescription>
    
    try
      % <SyntaxCode>
      % Construct two test AOs
      nsecs = 10;
      fs    = 1000;
      pl = plist('nsecs', nsecs, 'fs', fs, 'tsfcn', 'randn(size(t))');
      a1 = ao(pl); a2 = ao(pl);
      % Filter one time-series
      f2 = miir(plist('type', 'bandpass', 'fs', fs, 'order', 3, 'fc', [50 250]));
      a1f = filter(a1, plist('filter', f2));
      % make some cross-power
      a4 = a1f+a2; a4.setName;
      % Create the transpose of a4 to check the output data shape
      a4 = a4.';
      % Compute coherence
      Nfft = 2*fs;
      % Use different windows size as Nfft
      win  = specwin('Hanning', 1000);
      pl = plist('Nfft', Nfft, 'Win', win.type, 'order', 0, 'type', 'MS');
      out = cohere(a4,a1,pl);
      % </SyntaxCode>
      stest = true;
    catch err
      disp(err.message)
      stest = false;
    end
    
    % <AlgoDescription>
    %
    % 1) Check that output agrees with the output of MATLAB's mscohere.
    % 2) Check that the shape of the output data is equal to the input data
    %
    % </AlgoDescription>
    
    atest = true;
    if stest
      % <AlgoCode>
      TOL = 1e-12;
      
      % Redesign the window
      win = specwin('Hanning', Nfft);
      % Compute coherence using MATLAB's cohere
      [cxy, f] = mscohere(a4.y, a1.y, win.win, Nfft/2, Nfft, a1.fs);
      if  any(abs(cxy(4:end)-out.y(4:end))>TOL), atest = false; end
      if ne(f,      out.x), atest = false; end
      if ne(out, out, ple2), atest = false; end
      % Check the data shape
      if size(a4.y,1) == 1
        if size(out.y,1) ~= 1, atest = false; end
      else
        if size(out.y,2) ~= 1, atest = false; end
      end
      % </AlgoCode>
    else
      atest = false;
    end
    
    % Return a result structure
    result = utp_prepare_result(atest, stest, dbstack, mfilename);
  end % END UTP_24
  
  %% UTP_25
  
  % <TestDescription>
  %
  % Tests handling of units:
  % 1) white noise produced from normal pdf, with a given mean value and
  % sigma (distribution's 1st and 2nd orders)
  % 2) white noise produced from normal pdf, with a given mean value and
  % sigma (distribution's 1st and 2nd orders)
  % 3) complex coherence of the white noise series
  % 4) compares the units of the input and output
  %
  
  % </TestDescription>
  function result = utp_25
    
    % <SyntaxDescription>
    %
    % 1) Prepare the test tsdata:
    %   white noise from normal distribution + offset
    % 2) Assign a random unit
    % 3) Prepare the test tsdata:
    %   white noise from normal distribution + offset
    % 4) Assign a random unit
    % 5) complex cohere of the white noise
    %
    % </SyntaxDescription>
    
    % <SyntaxCode>
    try
      
      % Build time-series test data
      fs = 1;
      nsecs = 86400;
      sigma_distr_1 = 4.69e-12;
      mu_distr_1 = -5.11e-14;
      sigma_distr_2 = 6.04e-9;
      mu_distr_2 = 1.5e-10;
      
      % White noise
      type = 'Normal';
      
      a_n = ao(plist('waveform', 'noise', ...
        'type', type, 'fs', fs, 'nsecs', nsecs, 'sigma', sigma_distr_1));
      a_const = ao(mu_distr_1);
      a_1 = a_n + a_const;
      
      a_n = ao(plist('waveform', 'noise', ...
        'type', type, 'fs', fs, 'nsecs', nsecs, 'sigma', sigma_distr_2));
      a_const = ao(mu_distr_2);
      a_2 = a_n + a_const;
      
      % Set units and prefix from those supported
      unit_list = unit.supportedUnits;
      % remove the first empty unit '' from the list, because then is it
      % possible that we add a prefix to an empty unit
      unit_list = unit_list(2:end);
      prefix_list = unit.supportedPrefixes;
      a_1.setYunits(unit([cell2mat(utils.math.randelement(prefix_list,1)) cell2mat(utils.math.randelement(unit_list,1))]));
      a_2.setYunits(unit([cell2mat(utils.math.randelement(prefix_list,1)) cell2mat(utils.math.randelement(unit_list,1))]));
      
      % Evaluate the coherence of the time-series data
      win = 'Kaiser';
      psll = utils.math.randelement([10:10:200], 1);
      detrend = 0;
      scale_type = 'C';
      n_pts = nsecs*fs/10;
      
      C = cohere(a_1, a_2, ...
        plist('Win', win, 'psll', psll, 'Nfft', n_pts, 'order', detrend, 'type', scale_type));
      
      stest = true;
      
    catch err
      disp(err.message)
      stest = false;
    end
    % </SyntaxCode>
    
    % <AlgoDescription>
    %
    % 1) Check that (complex coherence yunits) equals [1]
    % 2) Check that (complex coherence xunits) equals [Hz]
    
    % </AlgoDescription>
    
    % <AlgoCode>
    atest = true;
    
    if stest
      if ~eq(C.yunits, unit('')) || ~eq(C.xunits, unit('Hz'))
        atest = false;
      end
    else
      atest = false;
    end
    % </AlgoCode>
    
    % Return a result structure
    result = utp_prepare_result(atest, stest, dbstack, mfilename);
  end % END UTP_25
  
  %% UTP_26
  
  % <TestDescription>
  %
  % Tests handling of units:
  % 1) white noise produced from normal pdf, with a given mean value and
  % sigma (distribution's 1st and 2nd orders)
  % 2) white noise produced from normal pdf, with a given mean value and
  % sigma (distribution's 1st and 2nd orders)
  % 3) complex coherence of the white noise series
  % 4) compares the units of the input and output
  %
  
  % </TestDescription>
  function result = utp_26
    
    % <SyntaxDescription>
    %
    % 1) Prepare the test tsdata:
    %   white noise from normal distribution + offset
    % 2) Assign a random unit
    % 3) Prepare the test tsdata:
    %   white noise from normal distribution + offset
    % 4) Assign a random unit
    % 5) complex cohere of the white noise
    %
    % </SyntaxDescription>
    
    % <SyntaxCode>
    try
      
      % Build time-series test data
      fs = 1;
      nsecs = 86400;
      sigma_distr_1 = 4.69e-12;
      mu_distr_1 = -5.11e-14;
      sigma_distr_2 = 6.04e-9;
      mu_distr_2 = 1.5e-10;
      
      % White noise
      type = 'Normal';
      
      a_n = ao(plist('waveform', 'noise', ...
        'type', type, 'fs', fs, 'nsecs', nsecs, 'sigma', sigma_distr_1));
      a_const = ao(mu_distr_1);
      a_1 = a_n + a_const;
      
      a_n = ao(plist('waveform', 'noise', ...
        'type', type, 'fs', fs, 'nsecs', nsecs, 'sigma', sigma_distr_2));
      a_const = ao(mu_distr_2);
      a_2 = a_n + a_const;
      
      % Set units and prefix from those supported
      unit_list = unit.supportedUnits;
      % remove the first empty unit '' from the list, because then is it
      % possible that we add a prefix to an empty unit
      unit_list = unit_list(2:end);
      prefix_list = unit.supportedPrefixes;
      a_1.setYunits(unit([cell2mat(utils.math.randelement(prefix_list,1)) cell2mat(utils.math.randelement(unit_list,1))]));
      a_2.setYunits(unit([cell2mat(utils.math.randelement(prefix_list,1)) cell2mat(utils.math.randelement(unit_list,1))]));
      
      % Evaluate the coherence of the time-series data
      win = 'Kaiser';
      psll = utils.math.randelement([10:10:200], 1);
      detrend = 0;
      scale_type = 'C';
      n_pts = nsecs*fs/10;
      
      C = cohere(a_1, a_2, ...
        plist('Win', win, 'psll', psll, 'Nfft', n_pts, 'order', detrend, 'type', scale_type));
      
      stest = true;
      
    catch err
      disp(err.message)
      stest = false;
    end
    % </SyntaxCode>
    
    % <AlgoDescription>
    %
    % 1) Check that (complex coherence yunits) equals [1]
    % 2) Check that (complex coherence xunits) equals [Hz]
    
    % </AlgoDescription>
    
    % <AlgoCode>
    atest = true;
    
    if stest
      if ~eq(C.yunits, unit('')) || ~eq(C.xunits, unit('Hz'))
        atest = false;
      end
    else
      atest = false;
    end
    % </AlgoCode>
    
    % Return a result structure
    result = utp_prepare_result(atest, stest, dbstack, mfilename);
  end % END UTP_26
  
  %% UTP_30
  
  % <TestDescription>
  %
  % Tests handling of special cases:
  % 1) white noise produced from normal pdf, with a given mean value and
  % sigma (distribution's 1st and 2nd orders)
  % 2) the same noise series
  % 3) cohere of the white noise series
  % 4) compares the output to unity
  %
  
  % </TestDescription>
  function result = utp_30
    
    % <SyntaxDescription>
    %
    % 1) Prepare the test tsdata:
    %   white noise from normal distribution + offset
    % 2) Assign a random unit
    % 3) Prepare the test tsdata:
    %   the same data as 1) and 2)
    % 4) cohere of the series
    %
    % </SyntaxDescription>
    
    % <SyntaxCode>
    try
      
      % Build time-series test data
      fs = 1;
      nsecs = 86400;
      sigma_distr_1 = 4.69e-12;
      mu_distr_1 = -5.11e-14;
      
      % White noise
      type = 'Normal';
      
      a_n = ao(plist('waveform', 'noise', ...
        'type', type, 'fs', fs, 'nsecs', nsecs, 'sigma', sigma_distr_1));
      a_const = ao(mu_distr_1);
      a_1 = a_n + a_const;
      
      % Set units and prefix from those supported
      unit_list = unit.supportedUnits;
      % remove the first empty unit '' from the list, because then is it
      % possible that we add a prefix to an empty unit
      unit_list = unit_list(2:end);
      prefix_list = unit.supportedPrefixes;
      a_1.setYunits(unit([cell2mat(utils.math.randelement(prefix_list,1)) cell2mat(utils.math.randelement(unit_list,1))]));
      
      % Build the second object as a copy of the first
      a_2 = a_1;
      
      % Evaluate the cohere of the time-series data
      win = specwin('BH92');
      olap = win.rov;
      detrend = 0;
      n_pts = nsecs*fs/10;
      scale_type = 'C';
      
      C = cohere(a_1, a_2, ...
        plist('Win', win, 'Nfft', n_pts, 'order', detrend, 'type', scale_type, 'olap', olap));
      
      stest = true;
      
    catch err
      disp(err.message)
      stest = false;
    end
    % </SyntaxCode>
    
    % <AlgoDescription>
    %
    % 1) Check that calculated cohere equals 1
    
    % </AlgoDescription>
    
    % <AlgoCode>
    atest = true;
    
    if stest
      if sum(ne(C.y, 1))
        atest = false;
      end
    else
      atest = false;
    end
    % </AlgoCode>
    
    % Return a result structure
    result = utp_prepare_result(atest, stest, dbstack, mfilename);
  end % END UTP_30
  
end