1
$\begingroup$

I've need to remove a DPL/DCS signal from a recording. Roughly speaking, a DCS signal is a low frequency (67.15Hz) square wave sent at the same time, over the same carrier, as speech. Because it is a square wave, it has many strong harmonics that overlap with speech. Obviously, the speech must be preserved.

On the surface, this seems like a problem that might be solved the same way as removing 60Hz power-line noise: lots of notch filters. However, power-line noise tends to be weaker and comes from a source that is much closer to a sine wave.

So, my question is: is there a better way to do this?

This link contains more info about DCS: http://onfreq.com/syntorx/dcs.html It mentions "Since DCS creates audio harmonics well above 300 Hz (i.e. into the audible portion of the band from 300 to 3000 Hz), radios must have good filters to remove the unwanted DCS noise." Ha!

EDIT: DCS is not a "square wave". I think the correct way to explain it is to say it's a NRZ binary code clocked at 67.15 Hz.

|improve this question
$\endgroup$
  • $\begingroup$ Rather than notch filters, you could try to estimate constant-amplitude sinusoids at each harmonic, and subtract them out. This is equivalent to taking a single, very long FFT of your signal (ideally with length an exact number of DCS cycles), and zeroing out just the bins corresponding to the DCS fundamental and harmonics. $\endgroup$ – dpwe Jul 24 '14 at 21:37
  • $\begingroup$ I have some guesses as to how I might do that, but it sounds hard (even ignoring minor frequency drift and other problems like that). Do you have a reference to something that explains a technique like this in more detail? $\endgroup$ – Bjorn Roche Jul 25 '14 at 14:01
  • $\begingroup$ I have written software which will remove a set of frequencies from an audio stream, in c++, which I use to remove 60 Hz power line contamination (essentially a 60 Hz square wave) from other signals, which might be useful. Interested? $\endgroup$ – Jim V Oct 17 '15 at 23:19
2
$\begingroup$

If you know that it is a square wave, and the frequency is roughly constant, I would think about synthesizing an inverted copy of it and adding it to the signal. The remainder would be your speech, plus some residual harmonics of the square wave frequency, which might be dealt with by a 1-channel adaptive filter (see Sec. IIC of this link: http://www.cs.cmu.edu/~aarti/pubs/ANC.pdf). It's possible that the adaptive filter will do it all by itself--I have used them with good results on periodic interfering signals with attention to the delay setting involved. In order to get the parameters of the inverted signal, you would do one of two things:
1. Fit a two-parameter square wave to the signal using least squares minimization (two parameters being the initial time delay and the amplitude, or add other parameters as you see fit). 2. Use an FFT to estimate phase and amplitude of the fundamental sine wave, then use the theoretical ratio of the amplitude of a square wave to its fundamental (isn't it 2/pi?) to reconstruct the square wave from that.

Which of these two approaches you pick depends on the bandwidth of the speech and its amplitude relative to your DCS signal. Good Luck!

|improve this answer
$\endgroup$
  • $\begingroup$ As I clarified, DCS isn't a perfect square wave, but I think your answer is very helpful anyway -- particularly the part about the adaptive filter. $\endgroup$ – Bjorn Roche Jul 30 '14 at 20:29
  • $\begingroup$ Even if it isn't a square wave, the process of estimation and subtraction may give good results if it has a known structure. Its parameters can be updated adaptively also. $\endgroup$ – girolamous Jul 31 '14 at 14:44
1
$\begingroup$

DCS should NOT be transmitted as a square wave.
The transmit side is supposed to have a low pass filter on the DCS signal before it gets mixed with the transmit audio. A properly functioning DCS transmitter should therefore NEVER transmit harmonics of the DCS signal above 300Hz.

Receive side, a (steep) high pass filter with a cutoff of 300Hz should be all you need.

If you are still getting harmonics from the DCS signal in your received audio then you should look at the transmit side. It is either over modulating the DCS (level is too high) or the transmit side low pass filter is ineffective. You may also have a little of both - DCS modulation level set too high and the filter not quite up to the job.

Given that you have a recording that has the DCS in it, I would assume the recording was made from a line-level output from the radio's demodulator. If the transmitter was working properly, all you need to do is to run your recording through a high pass filter with a 300 Hz cutoff.

|improve this answer
$\endgroup$
  • $\begingroup$ Thanks. Unfortunately, we can't do anything about the transmitter, which is definitely not filtering out all harmonics above 300 Hz. $\endgroup$ – Bjorn Roche Sep 30 '15 at 16:27
1
$\begingroup$

I have written software which will remove a set of frequencies from an audio stream, in c++, which I use to remove 60 Hz power line contamination (essentially a 60 Hz square wave) from other signals, which might be useful. the source code is included below.

Notes:

  • The included code was cut out of larger stuff with a text editor. Maybe a line or two is missing?

  • I run it on Ubuntu 15.04. This particular program sucks in the audio from either the line-in or mic input. The right channel is copied to the right channel output un altered. The left channel has the first 50 even harmonics subtracted and copied and copied to the left channel output.That is, the frequencies 600, 180, 300, ..., 3060 Hz are removed. The width of the notch filter at each frequency is about 1 Hz. If you run a signal generator through the left channel, you will notice some distortion within a few Hz of each notch. For fun, I put the output of an FM radio into the line-in and plugged the headphones in, and distortion on the left channel was not that bad at all, given the amount of processing going on.

  • jackd and the development library are needed:

    sudo apt-get install jackd

    sudo apt-get install libjack-dev

  • For some bizarre reason, you must do two crazy things...

Edit (via sudo gedit or whatever) /etc/security/limits.conf to contain the line @audio - rtprio 99

If the username that runs the program is XXXX, you have to do

sudo usermod -a G audio XXXX

  • The main program has the constants 4096 used, as the sample size of a block of data sucked in at a time. If the sound card on the machine cannot handle this, or the machine is otherwise is slow, use 1024. This should always work.

There. I hope this is sufficiently self contained.

File freqs.h

#ifndef FREQS_H_INCLUDED
#define FREQS_H_INCLUDED

#include <stdio.h>
#include <stdlib.h>
#include <memory.h>
#include <math.h>

class freq_info
{public:
double      Hz;     // The frequency of interest
double      w;      // 2 * pi * Hz
double      dt;     // delta time; time between samples
int         npts;       // number of points
double      tot_time;   // total time of all points
int         npts_1_cycle;
int         npts_max_cycles;
double      *cos_vals;
double      *sin_vals;
double      projection_sin;
double      projection_cos;

double      data_X_sin;
double      data_X_cos;
int         npts_in_product;

inline void zero() { memset( (void*)this, 0, sizeof(*this)); }
inline freq_info() { zero(); }
inline ~freq_info()
{
    if(cos_vals!=0) delete [] cos_vals;
    if(sin_vals!=0) delete [] sin_vals;
}

int init( double frequency, double time_between_samples, int number_of_points);

void projection_init();

void data_X_trigs_range( float *data, int data_idx_beg, int data_idx_end);

void projection_fini();
int  produce_projection ( float *data, unsigned int npts_in);
};

class freq_group
{public:

// Following 3 should be same for all frequencies of interest
double      dt;     // delta time; time between samples
int     npts;       // number of points
double      tot_time;   // total time of all points

int     min_npts_max_cycles;  // minimum of npts_max_cycles for all
                              //    FI[i] entries

double      info_freq_min;
double      info_freq_max;

int     n_freqs;    // # frequencies
freq_info   *FI;        // Info for all frequencies of interest.

inline void zero() { memset( (void*)this, 0, sizeof(*this)); }
inline freq_group() { zero(); }

inline ~freq_group()
{
   if(FI!=0) delete [] FI;
}

void set_time_and_points(double time_between_samples, int number_of_points);

int setup_freqs( double * freqs_of_interest, int n_freqs_of_interest);

int remove_projections1(float *data_in, float *data_out, unsigned int npts_in);

};

File jack_interface.h

#ifndef JACK_INTERFACE_H_INCLUDED
#define JACK_INTERFACE_H_INCLUDED

#include <stdio.h>
#include <errno.h>
#include <unistd.h>
#include <stdlib.h>
#include <string.h>
#include <jack/jack.h>

//When we are finally running, this holds a bunch of info...
class jack_interface
{public:
    enum {max_channels=4};
    char            jack_interface_name[32];
    unsigned int    sample_size;
    unsigned int    samps_per_sec;
    jack_client_t   *jclient;
    int     n_channels;             // or, number of ports
    jack_port_t*    in_ports [max_channels];
    jack_port_t*    out_ports[max_channels];

inline void zero() { memset( (void*)this, 0, sizeof(*this) ); }
inline jack_interface() { zero(); }

int create( char*       interface_name,
        unsigned int    requested_sample_size,
        int     n_channels,
        int         (*process) (jack_nframes_t sample_size, void *arg) );
};
#endif JACK_INTERFACE_H_INCLUDED

File freqs.cpp

#include "freqs.h"
inline void ZAP(char *msg=0)
{
    if(msg!=0) printf("%s\n", msg);
    printf("Fatal error.\n");
    exit(1);
}

void freq_info::projection_init()
{
    projection_sin  = 0.0;
    projection_cos  = 0.0;
    data_X_sin  = 0.0;
    data_X_cos  = 0.0;
    npts_in_product = 0;
    return;
}
void freq_info::data_X_trigs_range(float *data, int data_idx_beg, int data_idx_end)
{
    static  char    name[]  =   "freq_info::data_X_trigs_range(...)";
    int     i;

    if( data_idx_beg<0  || data_idx_end<0   || data_idx_end<data_idx_beg    ||  data_idx_end>=npts)

    {
        printf("%s\nInput ranges out of range.\n", name);
        ZAP();
    }
    for(i=data_idx_beg; i<=data_idx_end; i++)
    {
        data_X_sin += double( data[i] ) * sin_vals[i];
        data_X_cos += double( data[i] ) * cos_vals[i];
    }
    npts_in_product += data_idx_end - data_idx_beg + 1;
    return;
}

void freq_info::projection_fini()
{
    projection_sin = (2.0 * data_X_sin * dt) / ( double(npts_in_product-1) * dt );

    projection_cos = (2.0 * data_X_cos * dt) / ( double(npts_in_product-1) * dt );

    return;
}

void freq_group::set_time_and_points(double time_between_samples, int number_of_points)
{
    static  char    name[]  = "freq_group::set_time_and_points(...)";

    if(FI!=0) delete [] FI;
    zero();

    if(time_between_samples<=0.0 || number_of_points<8)
    {
        printf("%s\nInput parameters too small.\n", name);
        ZAP();
    }

    dt      = time_between_samples;
    npts        = number_of_points;
    tot_time    = double(npts-1) * dt;
   return;
} 

int freq_group::setup_freqs(double *freq_list, int n_freqs_in)
{
    static  char    name[]  = "freq_group::setup_freqs(...)";
    int     rval=0;
    int     i;
    double  freq_min, freq_max;

    n_freqs = n_freqs_in;

    if(freq_list==0 || n_freqs<1)
    {
        printf("%s\nInput args null, zero, or invalid.\n", name);
        rval = 10;
        goto EXIT_SEQ;
    }
    if( dt<=0.0 || npts<8 || tot_time<=0.0 )
    {
        printf("%s\nfreq_group::set_time_and_points NOT called!\n", name);
        rval = 20;
        goto EXIT_SEQ;
    }

    freq_min = 1.0e300;
    freq_max = 1.0e-300;
    for(i=0; i<n_freqs; i++)
    {
        if( freq_min > freq_list[i] )
            freq_min = freq_list[i];

        if( freq_max < freq_list[i] )
            freq_max = freq_list[i];
    }

    if(freq_min<=0.0 || freq_max<=0.0)
    {
        printf("%s\nSpecified frequency zero or negative.\n", name);
        rval = 30;
        goto EXIT_SEQ;
    }
    if( 1.0/freq_min > tot_time)
    {
        printf("%s\nSpecified frequency, %20.10e Hz, too low for time length of sample.\n", name, freq_min);
        rval = 40;
        goto EXIT_SEQ;
    }
    info_freq_min = freq_min;
    info_freq_max = freq_max;

    if( dt > 2.0 * (1.0 / freq_max) )
    {
        printf("%s\nSpecified frequence, %20.10e Hz, too high for sample rate.\n", name, freq_max);
        rval = 50;
        goto EXIT_SEQ;
    }
    if(FI != 0)
    {
        delete [] FI;
    }

    FI = new freq_info[ n_freqs ];
    min_npts_max_cycles = npts+2;
    for(i=0; i<n_freqs; i++)
    {
        rval = FI[i].init( freq_list[i], dt, npts);
        if(rval!=0)
        {
            printf("%s\nError from freq_info::init(...)\n", name);
            goto EXIT_SEQ;
        }
        if(min_npts_max_cycles > FI[i].npts_max_cycles)
        {
          min_npts_max_cycles = FI[i].npts_max_cycles;
        }
   }

   EXIT_SEQ:;
  return rval;
}

int freq_info::init( double frequency, double time_between_samples, int number_of_points)
{
    static  char    name[]  = "freq_info::init(...)";
    int     rval=0;
    static double two_pi = 0.0;

    int     i, max_cycles;
    double  arg;

    if(frequency<=0 || time_between_samples<0.0 || number_of_points<8)
    {
        printf("%s\nInput parameters too small.\n", name);
        rval = 10;
        goto EXIT_SEQ;
    }
    if(two_pi == 0.0)
    {
        two_pi = 8.0 * atan( 1.0 );
    }

    if(cos_vals!=0)
         delete [] cos_vals;

    if(sin_vals!=0)
         delete [] sin_vals;

    zero();

    Hz      = frequency;
    w       = frequency * two_pi;
    dt      = time_between_samples;
    npts        = number_of_points;
    tot_time    = dt * double(npts - 1);

    npts_1_cycle    = int(  (1.0/Hz) / time_between_samples  ) + 1;
    max_cycles = int( tot_time * Hz );
 // Total # full cycles of Hz in the time period
    if(max_cycles<1)
    {
        printf("%s\nSpecified frequency too low for the time span.\n", name);
        rval = 30;
        goto EXIT_SEQ;
    }

    npts_max_cycles = int(  (double(max_cycles)/Hz) / time_between_samples  ) + 1;

    cos_vals = new double[npts+1];
    sin_vals = new double[npts+1];

    for(i=0; i<npts; i++)
    {
        arg = w * double(i) * dt;
        sin_vals[i] = sin( arg );
        cos_vals[i] = cos( arg );
    }


EXIT_SEQ:;
return rval;
}

int  freq_info::produce_projection ( float *data_in, unsigned int npts_in)
{
    static  char    name[]  = "freq_group::produce_projection (...)";
    int     rval=0;
    if( int(npts_in) != npts)
    {
        printf("%s\nMismatch in # of data points.\n", name);
        rval = 10;
        goto EXIT_SEQ;
    }

   projection_init();
   data_X_trigs_range(data_in, 0, this->npts_max_cycles);
   projection_fini();

EXIT_SEQ:;
    return rval;
}

int freq_group::remove_projections1(float *data_in, float *data_out, unsigned int npts_in)
{
    static  char    name[]  = "freq_group::remove_projections1(...)";
    int     rval=0;
    int     i,j,beg,end;
    double  accum;

    if( int(npts_in) != npts)
    {
        printf("%s\nMismatch in # of data points.\n", name);
        rval = 10;
        goto EXIT_SEQ;
    }
    memset( (void*)data_out, 0, sizeof(*data_out) * npts_in);

    for(i=0; i<this->n_freqs; i++)
    {
        FI[i].projection_init();
    }
    for(i=0; i<this->n_freqs; i++)
    {
        FI[i].data_X_trigs_range(data_in, 0, this->min_npts_max_cycles);
    }
    for(i=0; i<this->n_freqs; i++)
    {
        if(FI[i].npts_max_cycles==min_npts_max_cycles)
            continue;
    // have not quite finished a full cycle at this frequency yet...
        beg = min_npts_max_cycles + 1;
        end = FI[i].npts_max_cycles;
        FI[i].data_X_trigs_range(data_in, beg, end);
    }

    for(i=0; i<this->n_freqs; i++)
   {
        FI[i].projection_fini();
   }

    // OK, have the projection of the data onto the various frequencies.
    // Remove it!
    for(i=0; i<npts; i++)
    {
        accum = 0.0;
        for(j=0; j<this->n_freqs; j++)
        {
            accum +=    FI[j].projection_sin * FI[j].sin_vals[i] +
                    FI[j].projection_cos * FI[j].cos_vals[i];
        }
        data_out[i] = float( double(data_in[i]) - accum );
    }

EXIT_SEQ:;
    return rval;
}

File jack_interface.cpp

#include "jack_interface.h"

void jack_interface_shutdown (void *arg)

{
// Do nothing
   return;
}

int jack_interface::create( char*       interface_name,
                unsigned int    requested_sample_size,
                int     n_channels,
                int         (*process) (jack_nframes_t sample_size, void *arg) )
{
    static  char    name[]  = "jack_interface::create(...)";
    int     rval=0;
    const  char     **ports;
    static char     default_interface_name[] = "simple";
    const  char     *server_name = NULL;
    jack_options_t  options = JackNullOption;
    jack_status_t   status;
    char            *use_interface_name;
    unsigned int     use_sample_size;
    static  char    IN1[]   = "in1";
    static  char    IN2[]   = "in2";
    static  char    IN3[]   = "in3";
    static  char    IN4[]   = "in4";
    static  char    OUT1[]  = "out1";
    static  char    OUT2[]  = "out2";
    static  char    OUT3[]  = "out3";
    static  char    OUT4[]  = "out4";
    char*           names_in[max_channels];
    char*           names_out[max_channels];
    int     i;

    // Validate (somewhat) args, etc.
    if(interface_name==0 || interface_name[0]==0)
        use_interface_name = default_interface_name;
    else
        use_interface_name = interface_name;

    if(requested_sample_size==0)
        use_sample_size=1024;
    else
        use_sample_size = requested_sample_size;

    if(n_channels<1 || n_channels>max_channels)
    {
        printf("%s\nn_channels must be between 1 and %d.\n", name, max_channels);
        rval = 10;
        goto EXIT_SEQ;
    }
    if(process==0)
    {
        printf("%s\nprocess function is null.\n", name);
        rval = 20;
        goto EXIT_SEQ;
    }
//--------------------------------------------------------
// open a client connection to the JACK server
    jclient = jack_client_open (use_interface_name, options, &status, server_name);
    if (jclient == NULL)
    {
        printf("%s\njack_client_open() failed, status = 0x%2.0x\n", name, status);
        if (status & JackServerFailed)
        {
            printf ("Unable to connect to JACK server\n");
        }
        rval = 50;
        goto EXIT_SEQ;
    }
    jack_set_buffer_size(jclient, use_sample_size);
    sample_size = jack_get_buffer_size(jclient);
    samps_per_sec = jack_get_sample_rate (jclient);
    jack_set_process_callback (jclient, process, 0);
    jack_on_shutdown (jclient, jack_interface_shutdown, 0);
    names_in [0] = IN1;
    names_in [1] = IN2;
    names_in [2] = IN3;
    names_in [3] = IN4;
    names_out[0] = OUT1;
    names_out[1] = OUT2;
    names_out[2] = OUT3;
    names_out[3] = OUT4;

for(i=0; i<n_channels; i++)
{

    in_ports[i] = jack_port_register (jclient, names_in[i],
      JACK_DEFAULT_AUDIO_TYPE,
      JackPortIsInput, 0);

    if(in_ports[i] == NULL)
    {
        printf("%s\nCannot create jack input port.\n", name);
        rval = 100;
        goto EXIT_SEQ;

    }
}
for(i=0; i<n_channels; i++)
{
    out_ports[i] = jack_port_register (jclient, names_out[i],
       JACK_DEFAULT_AUDIO_TYPE,
        JackPortIsOutput, 0);

    if(out_ports[i] == NULL)
    {
        printf("%s\nCannot create jack output port.\n", name);
        rval = 110;
        goto EXIT_SEQ;
    }
}
// Tell the JACK server to start.  Our
// process() callback will start running now.

if (jack_activate (jclient))
{
    printf ("%s\ncannot activate jack client\n", name);
    rval = 120;
    goto EXIT_SEQ;
}

//  Connect the ports.  You can't do this before the client is
//  activated, because we can't make connections to clients
//  that aren't running.  Note the confusing (but necessary)
//  orientation of the driver backend ports: playback ports are
//  "input" to the backend, and capture ports are "output" from
//  it.

ports = jack_get_ports (jclient, NULL, NULL,   JackPortIsPhysical|JackPortIsOutput);

if (ports == NULL)
{
    printf("%s\nno physical capture ports #1\n", name);
    rval = 130;
    goto EXIT_SEQ;
}
for(i=0; i<n_channels; i++)
{
    if( jack_connect(jclient, ports[i], jack_port_name(in_ports[i]) ) )
    {
        printf("%s\nCannot connect input port # %d\n", name, i);
        rval = 140;
        goto EXIT_SEQ;
    }
}
free (ports);

ports = jack_get_ports (jclient, NULL, NULL, JackPortIsPhysical|JackPortIsInput);

if (ports == NULL)
{
    printf( "%s\nno physical playback ports\n", name);
    rval = 150;
    goto EXIT_SEQ;
}

for(i=0; i<n_channels; i++)
{
    if( jack_connect(jclient, jack_port_name(out_ports[i]), ports[i]) )
    {
        printf("%s\nCannot connect output port # %d\n", name, i);
        rval = 160;
        goto EXIT_SEQ;
    }
}
free (ports);

EXIT_SEQ:;
return rval;

}

File main.cpp

#include <stdio.h>
#include <errno.h>
#include <unistd.h>
#include <stdlib.h>
#include <jack/jack.h>
#include "jack_interface.h"
#include "freqs.h"

enum { nports = 2 };  
static  jack_port_t     *in_ports [ nports ];       
static  jack_port_t *out_ports[ nports ];
static  jack_interface  JI;
static  jack_port_t     *input_portl;
static  jack_port_t     *input_portr;
static  jack_port_t     *output_portl;
static  jack_port_t     *output_portr;
static  unsigned int    samples_per_sec;
static  unsigned int    last_size = 0;
static  unsigned int    sample_size_out;
static freq_group FG;  

int process (jack_nframes_t nframes, void *arg)
{
    jack_default_audio_sample_t *inl, *inr, *outl, *outr;
    input_portl = JI.in_ports[0];
    input_portr = JI.in_ports[1];
    output_portl= JI.out_ports[0];
    output_portr= JI.out_ports[1];
    inl     = (float*)  jack_port_get_buffer (input_portl,  nframes);
    inr     = (float*)  jack_port_get_buffer (input_portr,  nframes);
    outl    = (float*)  jack_port_get_buffer (output_portl, nframes);
    outr    = (float*)  jack_port_get_buffer (output_portr, nframes);

    FG.remove_projections1(last_l_data, outl, nframes);
    memcpy (outr, last_r_data, sizeof (jack_default_audio_sample_t)*nframes);

    return 0;
   }

int main()
{
    static  char client_name[] = "filter5";
    int     rval=0;
    int     i;
    enum { nfreqs = 50 };   // Will be filtering out 50 frequencies
    double Freqs[nfreqs];   // The list of frequencies to eleminate
    double  freq_base = 60.0;
    double  this_freq;

    FG.set_time_and_points(1.0/48000.0, 4096);

    for(i=0; i<nfreqs; i++)
    {
       this_freq = double(2*i + 1) * freq_base;
       Freqs[i] = this_freq;
    }
    FG.setup_freqs(Freqs, nfreqs);
    rval = JI.create(client_name, 4096, 2, process);
    sample_size_out = JI.sample_size;
    samples_per_sec = JI.samps_per_sec;
    sleep(-1);
   return rval;
}
|improve this answer
$\endgroup$
  • $\begingroup$ This code is so badly formatted and non-readable that I don't even want to spend my time fixing it. Would you mind using gist instead and put it nicely in separate files? $\endgroup$ – jojek♦ Oct 19 '15 at 17:19
  • $\begingroup$ No, it isn't fragmented at all. That is the Stack Exchange stuff inserting all sorts of nonsense. $\endgroup$ – Jim V Oct 19 '15 at 18:54
  • $\begingroup$ The help says: Indent four spaces to create an block of preformatted text displayed in a monospaced font:. Each line of code is prefixed by four spaces. And this is the result. Well, I tried. If someone tells me where/how to download the code into four separate files on Stackesxchange (or any place else) I would be happy to do so. $\endgroup$ – Jim V Oct 19 '15 at 19:10
  • $\begingroup$ Seriously impossible? Jeeez... $\endgroup$ – jojek♦ Oct 19 '15 at 19:41
  • $\begingroup$ I don't understand why this was down voted -- it looks very helpful to me. The only thing that would make this a better answer is if you explained a bit of the "magic", i.e., I had a bit of trouble understanding how your filters worked and what filters you were using (are they just standard biquads?) $\endgroup$ – Bjorn Roche Oct 19 '15 at 20:04

protected by jojek♦ Oct 19 '15 at 19:41

Thank you for your interest in this question. Because it has attracted low-quality or spam answers that had to be removed, posting an answer now requires 10 reputation on this site (the association bonus does not count).

Would you like to answer one of these unanswered questions instead?

Not the answer you're looking for? Browse other questions tagged or ask your own question.