The BER curves in the previous posts were plotted with the Fading and AWGN profile varying with each iteration and Eb/No value. Hence the graph had abrupt changes when the channel used to suddenly dive deep blotting out the signal, thus making the comparison for various Eb/No value difficult. Moreover a number of iterations had to be taken and mean was required to find out to get the most suitable reading; this resulted into lot of compuatational time.

Hence to avoid this a constant fading profile and AWGN noise was selected. The seed for the shift registers in the impairements vi were created in the begining and same were used throughout. This again saved some time which was used for generating the randaom seeds everytime for each iteration and for each Eb/No value.
Making the above result reduced the computational time as the only one iteration was required per SNR value; moreover the comparison becomes also simple. A smoother curve :)

Disadvantage- the resolution becomes hard limited by the number of symbols per iteration. Earlier due to random channel per each iteration the mean used to increase the resolution by somewhat value.
Now it goes till 1E-04 and then suddenly dies down

Minor changes--
  1. chart in place of graph, no need to build the array everytime. dont know what was i thinking when i selected graph!- 
  2. the graphs shown below were plotted for 10,000 symbols compared to the ones in the previous post which were plotted for 4,000 symbols.
  3. One of the thing that was discussed with Prof Ranade was that how the separation parameter between two antennas was taken care to result in uncorrelated independent channels. The simulation was carried out by feeding different seeds to the channel shift registers (ie of the fading profile vis). Hence, just to confirm that the channels are not related a cross corelation was carried out between the fading profiles. This showed that the corelation coefficient was much less than 0.5 and hence it can be safely assumed that the channels are unrelated.

Here are the new graphs:

Theoretical Result: (for reference)



This one is without using the Alamouti Scheme. As it can be seen that the graph resembles the original quite nicely until SNR value of 26, then it somewhat remains constant. 


This one is with the Alamouti Scheme



Smooth but not accurate, too steep.
I am not sure what the problem is at this moment.