# Pulse-shaping filter and impulse/symbol train

#### ianyappy

Joined Feb 28, 2013
4
I have a query about application of pulse-shaping filters to a impulse/symbol train, for let's say a QPSK modulated signal. Can the symbol train be be oversampled above the symbol rate? If so, are the samples in between symbols zeroed or identical to the preceding symbol? If it is the former, does that mean that simple diagrams such as this are not quite right since the phase is only modulated at one point and not throughout the duration of the symbol? If it is the latter, does this cause inter-symbol interference? Thanks

#### w2aew

Joined Jan 3, 2012
219
I have a query about application of pulse-shaping filters to a impulse/symbol train, for let's say a QPSK modulated signal. Can the symbol train be be oversampled above the symbol rate? If so, are the samples in between symbols zeroed or identical to the preceding symbol? If it is the former, does that mean that simple diagrams such as this are not quite right since the phase is only modulated at one point and not throughout the duration of the symbol? If it is the latter, does this cause inter-symbol interference? Thanks
The picture is an idealized, unfiltered case - designed to dramatically show the phase modulation. In reality, the phase transitions do NOT happen instantaneously. If they did, the signal would occupy a huge amount of BW - bad idea. This is why pulse shaping filters are used - to smooth the phase transition between each state. Yes, if you filter too much, you can get intersymbol interference. The key is to select a filter characteristics that adequately smooths the phase transitions to limit the BW, but does not cause intersymbol interference.

#### ianyappy

Joined Feb 28, 2013
4
Thank you for your reply. So just to clarify, in between symbols, I should regard the phase term as gradually changing to the next symbol throughout the entire symbol duration rather than a quick jump at each symbol?
In addition, I believe you didn't quite answer my first question of whether the samples in between symbols are zeroed or identical to the preceding symbol?

#### w2aew

Joined Jan 3, 2012
219
Yes.

And, in general, the over sampled samples would be the same as the symbol, not zeroed. Let the filter establish the intermediate values.