You are here


A tricky one!

Dear comrades, the researchers of Comms group. Have a break from your hard work and have a go on this one. I liked it a lot.

Plug for RobProb on the comms group CVS

Here's my first blog entry. Well done to Jos and Andy for putting together an excellent resource that will no doubt promote and assist collaborative research no end!

I'm going to use this opportunity to plug my RobProb code on the group's CVS. This well documented code provides data types for both normal- and logarithmic-domain probabilities and probability ratios. More specifically, there are data types for logarithmic domain L-values and Logarithmic Likelihood Ratios (LLRs), along with their normal-domain equivalents. These data types are capable of representing infinite values (which are associated with absolute confidences) without the need for clipping at some arbitrary large value. Furthermore, this code provides IT++ vectors of these data types to store frames of LLRs, for examples. Conversion functions both to and from the double and vec data types are provided to retain compatibility with existing code.

Visualization of Bayesian Detection in MIMO Systems

The animated graphic on the right illustrates the possible channel output states of a 2x3 real-valued MIMO system using BPSK modulation. Additionally the Bayesian decision boundary is provided.


EXIT Chart Analysis When Considering Wideband Channels

EXIT charts under wide-band fading conditions can be generated in the following to ways:

    1. Only take a snap-shot of the channel, which would correspond to the AWGN EXIT chart

    2. Assume uncorrelated fading so that the resultant EXIT chart reflects the average behavior of the system.

The first option is not very useful since it does not really account for the Rayleigh fading. The second option appears attractive in the first place but is in my opinion not meaningful to analyze receivers. The reason is that in wide-band communication systems we would like to be able to exploit the freq. domain diversity of the channel. However, if uncorrelated fading is considered full diversity is already obtained in the time domain and the freq. domain diversity in fact only contributes interference. What would be of interest is some sort of EXIT chart representation for block-fading channels. The only way I could think of generating them is as follows:

    1. Generate channel realization and record the EXIT curve

    2. Repeat step 1 until the fading statistics are satisfied

    3. Average all EXIT functions by taking into account the fading PDF

Also it sounds easy, step three is in-fact rather complicated since you need to evaluated the probability of each channel realization. The problem seems to be closely related to the channel capacity versus outage probability issue.

Well done, Jos and Andy!

This is my first blog entry in Comms group website. The new comms group website is fabulous, thanks to the hard work of Jos and Andy.

Subscribe to RSS - blogs