Hello, I would like to ask why would we need the upsampling process for bits in Transmitter side? We do upsample then pass it through matched filter then modulate it, but why and how to choose the upsampling factor ? To be more precise, what would happen if we pass the bits through the matched filter without upsampling? This is in QPSK transmitter and can be generalized for all other digital communication schemes. Thank you --------------------------------------- Posted through http://www.DSPRelated.com
The significance of Upsampling at QPSK transmitter ?
Started by ●September 26, 2017
Reply by ●October 11, 20172017-10-11
On Tue, 26 Sep 2017 10:04:08 -0500, "moaado" <123839@DSPRelated> wrote:>Hello, > >I would like to ask why would we need the upsampling process for bits in >Transmitter side? >We do upsample then pass it through matched filter then modulate it, but >why and how to choose the upsampling factor ? >To be more precise, what would happen if we pass the bits through the >matched filter without upsampling? > >This is in QPSK transmitter and can be generalized for all other digital >communication schemes. > >Thank youUsually in a transmitter you need good control of the shape of the transmitted spectrum, including out-of-band emissions, and adjacent energy suppression. As you mentioned, pulse-shaping for matched filtering is also done for some of the same reasons but also for maximization of SNR at the receiver. All of these tasks are often difficult or impossible with only one sample per symbol in order to meet transmit mask requirements (i.e., the out-of-band emission specs). If you have, for example, a 20% root-raised-cosine filter shape, this requires multiple samples per symbol in order to achieve the proper shaping. Try it with one sample per symbol, then two, then four, etc., and you will see the differences.
Reply by ●October 16, 20172017-10-16
On Thu, 12 Oct 2017 02:24:21 +0000, Eric Jacobsen wrote:> On Tue, 26 Sep 2017 10:04:08 -0500, "moaado" <123839@DSPRelated> wrote: > >>Hello, >> >>I would like to ask why would we need the upsampling process for bits in >>Transmitter side? >>We do upsample then pass it through matched filter then modulate it, but >>why and how to choose the upsampling factor ? >>To be more precise, what would happen if we pass the bits through the >>matched filter without upsampling? >> >>This is in QPSK transmitter and can be generalized for all other digital >>communication schemes. >> >>Thank you > > Usually in a transmitter you need good control of the shape of the > transmitted spectrum, including out-of-band emissions, and adjacent > energy suppression. As you mentioned, pulse-shaping for matched > filtering is also done for some of the same reasons but also for > maximization of SNR at the receiver. All of these tasks are often > difficult or impossible with only one sample per symbol in order to meet > transmit mask requirements (i.e., the out-of-band emission specs). > > If you have, for example, a 20% root-raised-cosine filter shape, this > requires multiple samples per symbol in order to achieve the proper > shaping. Try it with one sample per symbol, then two, then four, etc., > and you will see the differences.Just to add to what Eric said -- you could, in theory, generate QPSK as a square wave or other pulses, and then do the pulse shaping in the analog domain. In fact, when they taught me back in the 1980's, that's how it was taught (if not, perhaps, actually done). The reason it's not done that way now is simply because -- today at least -- it's cheaper and more repeatable to do digitally. -- Tim Wescott Wescott Design Services http://www.wescottdesign.com
Reply by ●October 17, 20172017-10-17
On Monday, October 16, 2017 at 1:12:22 PM UTC-4, Tim Wescott wrote:> On Thu, 12 Oct 2017 02:24:21 +0000, Eric Jacobsen wrote: > > > On Tue, 26 Sep 2017 10:04:08 -0500, "moaado" <123839@DSPRelated> wrote: > > > >> > >>I would like to ask why would we need the upsampling process for bits in > >>Transmitter side? > >>We do upsample then pass it through matched filter then modulate it, but > >>why and how to choose the upsampling factor ? > >>To be more precise, what would happen if we pass the bits through the > >>matched filter without upsampling? > >> > >>This is in QPSK transmitter and can be generalized for all other digital > >>communication schemes. > >> > > > > Usually in a transmitter you need good control of the shape of the > > transmitted spectrum, including out-of-band emissions, and adjacent > > energy suppression. As you mentioned, pulse-shaping for matched > > filtering is also done for some of the same reasons but also for > > maximization of SNR at the receiver. All of these tasks are often > > difficult or impossible with only one sample per symbol in order to meet > > transmit mask requirements (i.e., the out-of-band emission specs). > > > > If you have, for example, a 20% root-raised-cosine filter shape, this > > requires multiple samples per symbol in order to achieve the proper > > shaping. Try it with one sample per symbol, then two, then four, etc., > > and you will see the differences. > > Just to add to what Eric said -- you could, in theory, generate QPSK as a > square wave or other pulses, and then do the pulse shaping in the analog > domain. In fact, when they taught me back in the 1980's, that's how it > was taught (if not, perhaps, actually done). The reason it's not done > that way now is simply because -- today at least -- it's cheaper and more > repeatable to do digitally. >Tim, don't they do this digitally with a windowed-sinc function. like if the bandwidth is B and they're trying to squeeze bits through this bandwidth at a rate of 2B, wouldn't the continuous-time signal (before modulating a carrier) be: i(t) + j q(t) where i(t) = SUM{ i[n] sinc( 2Bt - n) w(t - n/(2B)) } n q(t) = SUM{ q[n] sinc( 2Bt - n) w(t - n/(2B)) } n and if its OQPSK (which seems kewl to me as of late): i(t) = SUM{ i[n] sinc( 2Bt - n) w(t - n/(2B)) } n q(t) = SUM{ q[n] sinc( 2Bt - n - 1/2) w(t - (n - 1/2)/(2B)) } n where i[n] and q[n] are +/- 1 or (-1)^b[n] where b[n] is a bit, 0 or 1. is that what they sorta do for 4-PAM or QPSK? just curious. r b-j
