Why do we need to over sample signal in raise cosine filter?

[chessmath2009]

Hi

I would like to do know why do we need to do oversampling in a raise cosine filter. What will happen if we do not do oversampling?

I need to know what this oversampling factor does? the higher is better or the lower?

 

[zorro]

Are you talking about raised cosine filter spectral shaping in digital communications?
If not, please clarify what is the context.

 

[KlausST]

Hi,

In general a true cosine filter is impossible to build. Oversampling makes it easier to come close to the optimum.

With oversamplig one part of the filter is still analog, but the other part is digital. In digital you don't have part value tolerance (like with hardware), therefore you can build "critical" filters with predictable high Q, exact frequency, exact phase shift.

Klaus

 

[chessmath2009]

Yes, I mean raised cosine used in digital communication.

 

[zorro]

So, let's talk about baseband pulse transmission.
Really, oversampling is not needed, but can help reducing the requirements of the symbol synchronizer:
If you don't oversample (i.e. if you take samples once at every symbol) and samples are taken at maximum eye opening you have good reception. Eventually, if Nyquist first criterion is not met, intersymbol interference can be minimzed using a suitable equalizer.
This sampling requires a good timing recovery (symbol synchronization).
If sampling is shifted in time, eye is (partially or completely) closed and you have degradation that can not be completely recovered because there was frequency aliasing in sampling.
Sampling twice per symbol assures that there is no alias (when using raised cosine shaping for any roll-off < 1) and optimum detection can be achieved (with a good algorithm) despite the alignment of the samples with respect to the symbols.
Is it clear? I hope your question was about this subject.
Z