On Apr 1, 1:50*pm, Ubaid Abdullah <ubaidabdul...@gmail.com> wrote:
> On Apr 1, 6:48*am, ckian...@gmail.com wrote:
>
>
>
> > On Mar 31, 8:11*pm, Ubaid Abdullah <ubaidabdul...@gmail.com> wrote:
>
> > > > Since PAPR should be
> > > > computed before any matched filtering (am I right?), it seemed to
> > > > imply that RC pulse is used at the transmitter pulse shaping and hence
> > > > no matched filtering at the receiver.
>
> > > I dont think so. PAPR is a parameter of importance for power
> > > amplifier. That means we are interested in the PAPR of the final
> > > waveform that comes out of your DAC. You will compute PAPR for the
> > > signal that is input to the PA not at any other stage of the
> > > transmitter.
>
> > > Using RC pulse shaping at the transmitter doesn't imply that you
> > > shouldn't use matched filtering at the receiver. You should perceive
> > > matched filtering as a device that filters the out-of-band noise at
> > > the receiver and applies weights to the samples. You always need
> > > something for this job whatever the pulse shaping filter you use at
> > > the transmitter.
>
> > > Ubaid Abdullahhttp://dspdotcomm.blogspot.com
>
> > I agree with you. But is there any reason to use a RC instead of RRC?
> > What matched filter should one use to maintain ISI free and still
> > filter the out-of-band noise?
>
> You have two requirements for a communication system design.
>
> i). Zero ISI
> ii). Implementation of a matched filter at the receiver.
>
> A raised cosine filter is one that is designed to produce zero ISI.
> But, this zero ISI is needed before detection of the symbols or after
> matched filtering. If you implement raised cosine filter at the
> transmitter, the overall response of transmitting raised cosine filter
> and matched filter doesn't produce zero ISI. That's why RC filter is
> broken down into two parts each with an RRC response. In this case, we
> will implement RRC filter at the transmitter and the same filter at
> the receiver as matched filter to get the overall RC response. This
> gives us zero ISI prior to detection and fulfills both the
> requirements stated above.
>
> Ubaid Abdullahhttp://dspdotcomm.blogspot.com

That is correct, which brings us back to my second question above that
the pulse shaping function at the transmitter should not be a RC
filter, but the paper seems to imply that it is.

How do one deal with the filter delay if cyclic prefix is added after
pulse shaping?

In addition, in a scheme that uses a known seq (PN1) as the cyclic
prefix for symbol sync, e.g. [PN1|Data|PN2] where PN1 = PN2 and [Data|
PN2] is the FFT block, the starting symbols of PN1 after matched
filtering would not be right because of the filter delay, isn't it?