MATLAB Central Newsreader - Unique Random Numbers

Unique Random Numbers

Jarrod — Wed, 12 Nov 2008 00:25:03 -0500

I'm creating a random complex signal that consists of a real and imaginary part. The coefficient of each part comes from a predetermined alphabet. Providing the code below:

N=10;
alpha = [-sqrt(3)/2 -1/2 sqrt(3)/2 1/2];
tx = randsrc(1,N,alpha)+j*randsrc(1,N,alpha);

The only problem with this is that I need the real and imaginary parts to always be different. If you examine my alpha closely, it's points on the unit circle; however the points are only accurate if real does not equal imaginary. Any suggestions on how to use the same fix this? Thanks.

Re: Unique Random Numbers

Trevis Crane — Wed, 12 Nov 2008 04:26:03 -0500

A very unelegant way would be to generate your real and imag parts separately

rl = randsrc(1,N,alpha);
im = randsrc(1,N,alpha);

and remove the pairs that are identical

rl(rl == im) = [];
im(rl == im) = [];

You can repeat this iteratively until you have a total number of N pairsthat you need and then create your tx:

tx = rl + j*im;

just a first idea...

"Jarrod " <jrmfzf@gmail.com> wrote in message <gfd7ov$46p$1@fred.mathworks.com>...
> I'm creating a random complex signal that consists of a real and imaginary part. The coefficient of each part comes from a predetermined alphabet. Providing the code below:
>
> N=10;
> alpha = [-sqrt(3)/2 -1/2 sqrt(3)/2 1/2];
> tx = randsrc(1,N,alpha)+j*randsrc(1,N,alpha);
>
> The only problem with this is that I need the real and imaginary parts to always be different. If you examine my alpha closely, it's points on the unit circle; however the points are only accurate if real does not equal imaginary. Any suggestions on how to use the same fix this? Thanks.

Re: Unique Random Numbers

Roger Stafford — Wed, 12 Nov 2008 06:31:02 -0500

"Jarrod " <jrmfzf@gmail.com> wrote in message <gfd7ov$46p$1@fred.mathworks.com>...
> I'm creating a random complex signal that consists of a real and imaginary part. The coefficient of each part comes from a predetermined alphabet. Providing the code below:
>
> N=10;
> alpha = [-sqrt(3)/2 -1/2 sqrt(3)/2 1/2];
> tx = randsrc(1,N,alpha)+j*randsrc(1,N,alpha);
>
> The only problem with this is that I need the real and imaginary parts to always be different. If you examine my alpha closely, it's points on the unit circle; however the points are only accurate if real does not equal imaginary. Any suggestions on how to use the same fix this? Thanks.
---------
If you want your complex values to lie on some curve in the complex plane such as the unit circle, you should be generating your random values as a single random parameter from which the real and imaginary parts are derived, not from two independent random sources.

In the example you give, you should have:

alpha = pi/6*[1,2,4,5,7,8,10,11];
tx = exp(randsrc(1,N,alpha)*j); % <-- Where j is sqrt(-1)

Roger Stafford