matlab: fix DFT for periodic signals

Signed-off-by: Thorsten Liebig <Thorsten.Liebig@gmx.de>

matlab/DFT_time2freq.m

@@ -1,32 +1,47 @@
-function f_val = DFT_time2freq( t, val, freq )
-% f_val = DFT_time2freq( t, val, freq )
+function f_val = DFT_time2freq( t, val, freq, signal_type )
+% f_val = DFT_time2freq( t, val, freq, signal_type )
 %
 % computes the DFT at the given frequencies
-% f_val: single-sided spectrum
+%
+% parameter:
+% t  : time vector
+% val: data vector
+% freq: DFT frequency vector
+% signal_type: 'pulse' (default), 'periodic'
+%
+% return values:
+% f_val:  single-sided spectrum
 %
 % example:
 %   t=linspace(0,1,100);
 %   t_val=0.9*sin(2*pi*3*t); % sine wave; amplitude 0.9; frequency 3 Hz
 %   f=linspace(1,5,101);
-%   f_val=DFT_time2freq( t, t_val, f );
+%   f_val=DFT_time2freq( t, t_val, f, 'periodic' );
 %   interp1(f,abs(f_val),3)
 % ans = 0.8910
 %   plot( t, t_val )
 %   plot( f, abs(f_val) )
-%
-% See also FFT_time2freq
 
 if numel(t) ~= numel(val)
     error 'numel(t) ~= numel(val)'
 end
 
-dt = t(2)-t(1);
+if nargin<4
+  signal_type = 'pulse';
+end
 
 f_val = zeros(1,numel(freq));
 for f_idx=1:numel(freq)
     f_val(f_idx) = sum( val .* exp( -1i * 2*pi*freq(f_idx) * t ) );
 end
 
-f_val = f_val * dt;
+if strcmpi(signal_type, 'pulse')
+  dt = t(2)-t(1);
+  f_val = f_val * dt;
+elseif strcmpi(signal_type, 'periodic')
+  f_val = f_val / length(t);
+else
+  error 'unknown signal type'
+end
  
 f_val = f_val * 2; % single-sided spectrum

matlab/FFT_time2freq.m

@@ -1,6 +1,8 @@
 function [f,val] = FFT_time2freq( t, val )
 % [f,val] = FFT_time2freq( t, val )
 %
+% Note: This function can only be used for pulse signals
+%
 % See also DFT_time2freq
 
 dt=t(2)-t(1); % timestep

matlab/ReadUI.m

@@ -36,15 +36,22 @@ if (nargin<2)
 end
 
 AR_order = 0;
+SignalType = 'pulse';
 
 for n=1:2:numel(varargin)
     if (strcmp(varargin{n},'AR')==1)
         AR_order =  varargin{n+1};
+    elseif strcmpi(varargin{n},'SignalType')
+        SignalType = varargin{n+1};
     else
         warning('CSXCAD:ReadUI', ['"' varargin{n} '" is an unknown argument']);
     end
 end
 
+if strcmpi(SignalType,'periodic') && AR_order>0
+    error 'auto-regressive model not compatible with periodic signals'
+end
+
 if (ischar(files))
     filenames{1}=files;
 else
@@ -66,6 +73,9 @@ for n=1:numel(filenames)
     end    
     
     if (nargin<3) || isempty(freq)
+        if strcmpi(SignalType,'periodic')
+           warning 'ReadUI: periodic signal type not supported by FFT'
+        end
         [UI.FD{n}.f,UI.FD{n}.val] = FFT_time2freq( t,val );
     else
         UI.FD{n}.f = freq;
@@ -82,15 +92,15 @@ for n=1:numel(filenames)
             end
             if (EC~=0)
                 warning('CSXCAD:ReadUI','AR estimation failed, skipping...')
-                UI.FD{n}.val = DFT_time2freq( t, val, freq );
+                UI.FD{n}.val = DFT_time2freq( t, val, freq, SignalType );
             end
         elseif (AR_order<=0)
-            UI.FD{n}.val = DFT_time2freq( t, val, freq );
+            UI.FD{n}.val = DFT_time2freq( t, val, freq, SignalType );
         else
             [val_ar t_ar UI.FD{n}.val EC] = AR_estimate( t, val, freq, AR_order);
             if (EC~=0)
                 warning('CSXCAD:ReadUI','AR estimation failed, skipping...')
-                UI.FD{n}.val = DFT_time2freq( t, val, freq );
+                UI.FD{n}.val = DFT_time2freq( t, val, freq, SignalType );
             end
         end
     end   

matlab/calcPort.m

@@ -19,6 +19,7 @@ function [port] = calcPort( port, SimDir, f, varargin)
 %   'RefPlaneShift': for transmission lines only, See also calcTLPort for
 %                    more details
 %   'SwitchDirection': 0/1, switch assumed direction of propagation
+%   'SignalType':    'pulse' (default) or 'periodic'
 %
 % output: 
 %   % output signals/values in time domain (TD):