Timing Recovery problem and Symbol peak issue!

[shahin]

In Timing Recovery we basically try to determine the correct time within a symbol period to take a sample and in the real world communication because we use pulse shaping before transmission and also matched filter at the receiving end we would have to detect the peak point of each symbol. As we look at the eye diagram of a signal after the matched filter we see some peaks are associated with the correct time that we have to take the sample as the symbol BUT we also have some symbol points that are NOT located at peaks.

So my question is how we could determine those symbol points which are not at peaks and I’m wondering if it’s the responsibility of loop filter to just be reluctant to some phase errors occurring on some points that are not peaks?

Sorry if I wasn’t very clear about the problem but whenever I look for the timing recovery they just mention a lot of methods how to detect the peak points or zero crossing points as well as the interpolation methods but nowhere talks anything about the symbols that are not at peaks!

[orlando]

Hello shahin,

You should use the eye diagram to guide you in deciding where in time you should set your sampling instant.

You would want to choose the point, usually close to the middle, where the eye is open the most (vertically) , thus where you have the maximum margin for error.

As noise and ISI is present some errors will occur, but choosing the widest point should help you minimise errors received.

If you have a dynamic system , then this point could be adjusted over time.

i hope that helps.