FPGA Central

If I use a PERIODIC training sequence to train an equaliser, the PSD of
the
training signal will correspond to the magnitude squared of its Fourier
series,
which is a series of frequency spikes.

How can the equaliser correct for the regions of the channel frequency
response
that are not stimulated by the training sequence, ie the portions
between the
frequency spikes?

[email protected] wrote:

> If I use a PERIODIC training sequence to train an equaliser, the PSD of
> the
> training signal will correspond to the magnitude squared of its Fourier
> series,
> which is a series of frequency spikes.
>
> How can the equaliser correct for the regions of the channel frequency
> response
> that are not stimulated by the training sequence, ie the portions
> between the
> frequency spikes?

If the length of the equalizer is comparable to the periodicities in
your training sequence, then it will converge at the frequency peaks
only. In the valleys between the peaks the response will be pretty
arbitrary.
If the length of the equalizer is less then the period of the training
sequence, it will converge to the envelope of the spectrum.

Vladimir Vassilevsky

DSP and Mixed Signal Design Consultant

http://www.abvolt.com