Oversampling in communication systems

Started by Mas Bas August 15, 2005
Hi all,

I would like to know about the oversampling used in communication 
system. There are 2 technique which commonly used.

1. Using (N-1) zero padding, to get N factor oversampling and finally 
use a digital filter to filled up the zeros.

2. Repeat the sample (N-1) times to get N factor oversampling and also 
finally use a digital filter to adjust the value in oversampling location.

Which one is the most correct one ?

Please advice,

Thanks,

Mas
"Mas Bas" <buzzberry@gmail.com> wrote in message 
news:4300b825$0$18642$14726298@news.sunsite.dk...
> Hi all, > > I would like to know about the oversampling used in communication system. > There are 2 technique which commonly used. > > 1. Using (N-1) zero padding, to get N factor oversampling and finally use > a digital filter to filled up the zeros. > > 2. Repeat the sample (N-1) times to get N factor oversampling and also > finally use a digital filter to adjust the value in oversampling location. > > Which one is the most correct one ? > > Please advice, > > Thanks, > > Mas
Hello Mas, I don't think it is a question of which one is correct and which one is incorrect. They simply are different. They both cause the spectrum to be repeated in the new sample space, with zero padding having a flat response, and the repeated value case offering a bit of rolloff. The 1st case is described by a convolution with a Kronecker Delta, and the second case by a convolution with a rectangular pulse. The frequency response of the Delta is flat, and the pulse's response is a sinc function. Also the repeated value case has more energy, but either way the lowpass filter that is used to select out a single copy of the spectrum, can compensate for the loss of energy or the non flat frequency response. So either way can work. It becomes your choice. If the upsampling is very high, the loss of energy may pose some quantization issues. I hope this helps. Clay S. Turner