function [varargout]=mat2wfdb(varargin)

 [xbit]=mat2wfdb(X,fname,Fs,bit_res,adu,info,gain,sg_name,baseline,isquant, isdigital)

 Convert data from a matlab array into Physionet WFDB format file.

 Input Paramater are:

 X       -(required)  NxM matrix of M signals with N samples each. The
                      signals can be of type double.The signals are assumed to be
                      in physical units already and will be converted to
                      ADU.
 fname   -(required)  String where the the header (*.hea) and data (*.dat)
          files will be saved (one single name for both, with no sufix).
 Fs      -(Optional)  1x1 sampling frequency in Hz (all signals must have
          been sampled at the same frquency). Default is 1 Hz.
 bit_res -(Optional)  1xM (or Mx1):scalar determining the bit depth of the conversion for
                      each signal.
                      1x1 : If all the signals should have the same bit depth
          Options are: 8,  16, and 32 ( all are signed types). 16 is the default.
 adu     -(Optional) Describes the physical units (default is 'mV').
          Three input formats:
            - String delimited by forward slashes (e.g. 'V/mV/mmHg'), with
            M-1 slash characters
            - Single string (e.g. 'V'), in which case all signals will 
            have the same physical units.
            - Cell array of strings, where the total units entered has to equal M 
            (number of channels).
 info    -(Optional)  String that will be added to the comment section of the header file.
           For multi-lined comments, use a cell array of strings. Each
           cell will be output on a new line. Note that comments in the
           header file are automatically prefixed with a pound symbol (#)
 gain    -(Required for digital only) Scalar or Mx1 array of floats indicating the difference in sample values 
           that would be observed if a step of one physical unit occurred in the original 
           analog signal. If the 'isdigital' field is 1, this field is mandatory. Otherwise,
           this field is ignored if present. 
 baseline -(Required for digital only) Mx1 array of integers that specifies the sample value for each channel
           corresponding to 0 physical units. Not to be confused with 'ADC zero' which 
           is currently always taken and written as 0 in this function. If
           the 'isdigital' field is 1, this field is mandatory. Otherwise,
           this field is ignored if present. 
 sg_name -(Optional) Cell array of strings describing signal names.

 isquant   -(Optional) Logical value (default=0). Use this option if the
           input signal is already quantitized and you want to remove round-off
           error by mapping the original values to integers prior to fixed
           point conversion. This field is only used for input physical
           signals. If 'isdigital' is set to 1, this field is ignored.

 isdigital -(Optional) Logical value (default=0). Specifies whether the input signal is 
            digital or physical (default). If it is digital, the signal values will be 
            directly written to the file without scaling. If the signal is physical, 
            the optimal gain and baseline will be calculated and used to digitize the signal
            to write the WFDB file. This flag also decides the allowed
            input combinations of the 'gain' and 'baseline' fields.
            Digital signals must have both, and physical signals must have
            neither (as the ideal values will be automatically calculated). 

 Ouput Parameter:

 xbit    -(Optional)  NxM the quantitized signals that written to file (possible
          rescaled if no gain was provided at input). Useful for comparing
          and estimating quatitization error with the input double signal X
          (see examples below).


  NOTE: The signals can have different amplitudes, they will all be scaled to
  a reference gain, with the scaling factor saved in the *.hea file.

Written by Ikaro Silva 2010
Modified by Louis Mayaud 2011, Alistair Johson 2016
 Version 1.0

 Since 0.0.1
 See also wrsamp, wfdbdesc

%%%%%%%%%  Example 1 %%%%%%%%%%%%

 display('***This example will write a  Ex1.dat and Ex1.hea file to your current directory!')
 s=input('Hit "ctrl + c" to quit or "Enter" to continue!');

 %Generate 3 different signals and convert them to signed 16 bit in WFDB format
 clear all;clc;close all
 N=1024;
 Fs=48000;
 tm=[0:1/Fs:(N-1)/Fs]';
 adu='V/mV/V';
 info='Example 1';


 %First signal a ramp with 2^16 unique levels and is set to (+-) 2^15 (Volts)
 %Thus the header file should have one quant step equal to (2^15-(-2^15))/(2^16) V.
 sig1=double(int16(linspace(-2^15,2^15,N)'));

 %Second signal is a sine wave with 2^8 unique levels and set to (+-) 1 (mV)
 %Thus the header file should one quant step equal a (1--1)/(2^8)  adu step
 sig2=double(int8(sin(2*pi*tm*1000).*(2^7)))./(2^7);

 %Third signal is a random binary signal set to to (+-) 1 (V) with DC (to be discarded)
 %Thus the header file should have one quant step equal a 1/(2^15) adu step.
 sig3=(rand(N,1) > 0.97)*2 -1 + 2^16;

 %Concatenate all signals and convert to WFDB format with default 16 bits (empty brackets)
 sig=[sig1 sig2 sig3];
 mat2wfdb(sig,'Ex1',Fs,[],adu,info)

 % %NOTE: If you have WFDB installed you can check the conversion by
 % %uncomenting and this section and running (notice that all signals are scaled
 % %to unit amplitude during conversion, with the header files keeping the gain info):

 % !rdsamp -r Ex1 > foo
 % x=dlmread('foo');
 % subplot(211)
 % plot(sig)
 % subplot(212)
 % plot(x(:,1),x(:,2));hold on;plot(x(:,1),x(:,3),'k');plot(x(:,1),x(:,4),'r')

%%%%%%% End of Example 1%%%%%%%%%