Hanning window - I/Q signal

Hi everybody,

I have implemented Hann windowing function as follows:

var N = samples.Length / 2;
for (int i = 0; i < samples.Length; i++)
      {
           samples[i] *= 0.5 - 0.5 * Math.Cos((2 * Math.PI * i) / (N -
1));
      }

Samples field consists of varying real and imaginary values. When i apply
Hann window my signal mirrors every time at negative frequency, ergo the
peak should be at 5kHz only but it also appears at -5kHz after the
operation mentioned above.

What do I miss? 
Thanks for your advice!


---------------------------------------
Posted through http://www.DSPRelated.com

On Thursday, February 4, 2016 at 7:41:15 PM UTC+1, r.a.m- wrote:
> I have implemented Hann windowing function as follows:
> 
> var N = samples.Length / 2;
> for (int i = 0; i < samples.Length; i++) {
>    samples[i] *= 0.5 - 0.5 * Math.Cos((2 * Math.PI * i) / (N - 1));
> }
> 
> Samples field consists of varying real and imaginary values.

I'm not sure I understand what samples really is.  Is it a double[]
with interleaved real and imaginary components or is it a Complex[]?
Either way, your code snippet seems wrong.

If samples is a Complex[] you would not divide the array length by
two, would you?

If samples is a double[] with interleaved real and imaginary
components the function you want probably looks more like this:

  var N = samples.Length / 2;
  for (int i = 0; i < samples.Length; i++) {
     double w = 0.5 - 0.5 * Math.Cos((2 * Math.PI * i) / (N - 1));
     samples[2*i+0] *= w;
     samples[2*i+1] *= w;
  }

It is important to use the same scaling factor for real and imaginary
part.  You basically used slightly different factors and applied a
short window on only the first half of the array.  Inconsistent scale
factors between real and imaginary parts can account for some aliasing
(5 kHz mirroring to -5 kHz) to some extent.  In the extreme case you
can multiply the imaginary part by 0 and then the spectrum has to be
symmetric around 0 Hz. ;-)

Cheers!
sg