As oversampling produces several baud rate signal sequences (for example, oversampling by a factor of 2 produces two baud rate sequences with an offset of half baud duration), could these sequences be viewed as a kind of diversity? and what's its gain? _____________________________________ Do you know a company who employs DSP engineers? Is it already listed at http://dsprelated.com/employers.php ?
Does oversampling bring diversity gain?
Started by ●May 11, 2007
Reply by ●May 11, 20072007-05-11
On May 11, 11:08 am, "Benjamin Jiajia" <mzh...@marvell.com> wrote:> As oversampling produces several baud rate signal sequences (for example, > oversampling by a factor of 2 produces two baud rate sequences with an > offset of half baud duration), could these sequences be viewed as a kind > of diversity? and what's its gain? >Just like any "diversity"-based communication systems, this depends on the signal and noise model. Is your noise continuous-time white additive Gaussian noise? Or is it a discrete-time model? How do you normalize between the two? I think this is an important part to answering your question. One approach is to examine the power spectrum of the modulation, and of the noise, see if it gives you any gains. This is useful for when you use a matched filter. Write out the output of the matched filter due to the signal, and due to the noise, in the frequency domain. Then you can compare to see how much you really gain (or not). Hope that helps a bit, Julius
Reply by ●May 11, 20072007-05-11
Benjamin Jiajia wrote:> As oversampling produces several baud rate signal sequences (for example, > oversampling by a factor of 2 produces two baud rate sequences with an > offset of half baud duration), could these sequences be viewed as a kind > of diversity? and what's its gain???? A signal is a signal. sampling faster than need be eases the design of certain filters, but it doesn't alter the signal. Jerry -- Engineering is the art of making what you want from things you can get. ¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯
Reply by ●May 11, 20072007-05-11
On Fri, 11 May 2007 11:08:28 -0500, "Benjamin Jiajia" <mzheng@marvell.com> wrote:>As oversampling produces several baud rate signal sequences (for example, >oversampling by a factor of 2 produces two baud rate sequences with an >offset of half baud duration), could these sequences be viewed as a kind >of diversity? and what's its gain?I think if you really want to stretch the usual use of the term "diversity" you could make that argument, although some people might raise an eyebrow at the idea. I wouldn't, actually, but I'm like that. ;) To me the idea of "diversity" in this context is just to use multiple, hopefully reasonably independent, means of accumulating information that increases the probability that the impairments can be overcome. So oversampling does sort of meet that definition since the time separation provides different measurements of the same signal with different noise instances. Time diversity is a legitimate and well-known diversity technique, and if one is worried mostly about noise, then the additional samples may as well be taken interlaced in a single transmission as taken at half the rate in two separate transmissions. Both are instances of time-diversity and the combining will just be the usual processing gain afforded by filtering 2:1. That being said, "diversity" in communications systems is usually done to mitigate fading, not noise, and it is highly unlikely (well, let's say not possible) that adding more samples is going to provide any additional useful information about the signal, since they're all going to be faded the same. "Diversity" opportunities only work if the channel impairments are not well correlated across the diversity inputs. More samples within the same faded signal will be highly correlated with the fading effects in the original samples, so no gain will be realized. Noise is different, since the noise instances in the samples will not be correlated if the noise is white. So, really, it depends on the channel model. Noise? Sure, oversampling improves "diversity". Fading? Not gonna help at all. Eric Jacobsen Minister of Algorithms Abineau Communications http://www.ericjacobsen.org
Reply by ●May 11, 20072007-05-11
> > Noise is different, since the noise instances in the samples will not > be correlated if the noise is white. > > So, really, it depends on the channel model. Noise? Sure, > oversampling improves "diversity". Fading? Not gonna help at all. > >If the signal has passed through a bandpass filter near the Nyquist BW, which is the usual case in a receiver, will not the noise during samples taken of any particular symbol, be highly correlated thus reducing any gain of oversampling? Mark
Reply by ●May 11, 20072007-05-11
On May 11, 3:01 pm, Mark <makol...@yahoo.com> wrote:> > Noise is different, since the noise instances in the samples will not > > be correlated if the noise is white. > > > So, really, it depends on the channel model. Noise? Sure, > > oversampling improves "diversity". Fading? Not gonna help at all. > > If the signal has passed through a bandpass filter near the Nyquist > BW, which is the usual case in a receiver, will not the noise during > samples taken of any particular symbol, be highly correlated thus > reducing any gain of oversampling? > > MarkFor your noise model, that is correct, assuming idealized filters etc. But often the noise is generated by the amplifiers, and within the ADC itself. What about in that case? Julius
Reply by ●May 11, 20072007-05-11
On 11 May 2007 13:01:57 -0700, Mark <makolber@yahoo.com> wrote:> >> >> Noise is different, since the noise instances in the samples will not >> be correlated if the noise is white. >> >> So, really, it depends on the channel model. Noise? Sure, >> oversampling improves "diversity". Fading? Not gonna help at all. >> >> > >If the signal has passed through a bandpass filter near the Nyquist >BW, which is the usual case in a receiver, will not the noise during >samples taken of any particular symbol, be highly correlated thus >reducing any gain of oversampling? > >MarkRemember that this case is just the usual SNR improvement due to the processing gain from oversampling and filtering. Nothing magic going on there, it's just that the 'diversity' point of view is another way to look at it. Eric Jacobsen Minister of Algorithms Abineau Communications http://www.ericjacobsen.org
Reply by ●May 12, 20072007-05-12
>As oversampling produces several baud rate signal sequences (for example, >oversampling by a factor of 2 produces two baud rate sequences with an >offset of half baud duration), could these sequences be viewed as a kind >of diversity? and what's its gain? > > > > >_____________________________________ >Do you know a company who employs DSP engineers? >Is it already listed at http://dsprelated.com/employers.php ? >Do some matlab simulations with and without oversampling. There is no diversity gain unless you see the BER-SNR curve gets steeper with oversampling. I guess there is no diversity gain since there is no diversity signal path, and you could probably only see a shift of the BER-SNR curve with oversampling. This is merely a power gain, which is equivalent to doubling the transmitter power instead of oversampling by two. LBB _____________________________________ Do you know a company who employs DSP engineers? Is it already listed at http://dsprelated.com/employers.php ?
Reply by ●May 12, 20072007-05-12
On May 11, 4:44 pm, julius <juli...@gmail.com> wrote:> On May 11, 3:01 pm, Mark <makol...@yahoo.com> wrote: > > > > Noise is different, since the noise instances in the samples will not > > > be correlated if the noise is white. > > > > So, really, it depends on the channel model. Noise? Sure, > > > oversampling improves "diversity". Fading? Not gonna help at all. > > > If the signal has passed through a bandpass filter near the Nyquist > > BW, which is the usual case in a receiver, will not the noise during > > samples taken of any particular symbol, be highly correlated thus > > reducing any gain of oversampling? > > > Mark > > For your noise model, that is correct, assuming idealized filters etc. > But often the noise is generated by the amplifiers, and within the > ADC itself. What about in that case? > > JuliusOK, oversampleing can be used to reduce the effect of quantizing noise in the A/D and other components after the channel filter but not to reduce the noise generated in the front end. In any well designed receiver, the front end noise dominates. Based on the context of the OPs question I think the architecture we are talking about is A/D conversion after the (analog) chanel filter (where the noise during any one symbol period will correlate) ....perhaps this is an incorrect assumption on my part. Mark Mark
Reply by ●May 12, 20072007-05-12
On May 11, 7:52 pm, Eric Jacobsen <eric.jacob...@ieee.org> wrote:> On 11 May 2007 13:01:57 -0700, Mark <makol...@yahoo.com> wrote: > > > > > > > > >> Noise is different, since the noise instances in the samples will not > >> be correlated if the noise is white. > > >> So, really, it depends on the channel model. Noise? Sure, > >> oversampling improves "diversity". Fading? Not gonna help at all. > > >If the signal has passed through a bandpass filter near the Nyquist > >BW, which is the usual case in a receiver, will not the noise during > >samples taken of any particular symbol, be highly correlated thus > >reducing any gain of oversampling? > > >Mark > > Remember that this case is just the usual SNR improvement due to the > processing gain from oversampling and filtering. Nothing magic going > on there, it's just that the 'diversity' point of view is another way > to look at it. > >This usual SNR improvement processing gain applies only to wideband noise generated after the channel filter and does not apply to narrow band noise generated before the channel filter.. Mark Mark
