over-sampling at the receiver

Hi there,

I'm facing some trouble understanding this concept.
I read in some papers that over-sampling at the receiver side increase the signal's samples correlation.
How can this happen i don't know. as far as i know, increasing the number of samples doesn't affect the correlation.

please help.
regards,

Oversampling means that you sample at a higher rate than the signal rate. Since the signal changes slowly wrt. the sampling rate, neighboring samples are correlated.

But if you normalize the correlation delay to the sampling rate, then oversampling just gives you a greater resolution of the same autocorrelation function. So in that sense, nothing changes.

Thanks,

The purpose of my question was to make sure of the following:
In a very low SNR environment the fluctuating of the noise does not allow the non-coherent detection (energy detector) to give good results. therefore we try to identify the signal from the noise by some correlation schemes (because the noise is random variable Gaussian iid).
In this sense can we say that over sampling increases the correlation between sampling. especially if we use the eigenvalues of the correlation matrix to catch this increased correlation.

regards,

I do not follow your line of reasoning. What I think you may be doing is this.

The desired signal is band limited and deterministic. You have a noise signal with greater bandwidth than the desired signal. If you sample at the rate corresponding to the desired signal, there will be aliasing for the noise signal. This decreases your in-band SNR. If you increase your sampling rate there is less aliasing.

After that, the correlation functions as a low-pass filter, which reduces your out-of-band noise. You can achieve the same effect by low-pass filtering before sampling. That way you can avoid oversampling.

over sampling can increase the correlation.for more detail check out the following paper ,,Spectrum-Sensing Algorithms for Cognitive Radio Based on Statistical Covariances
Yonghong Zeng, Senior Member, IEEE, and Ying-Chang Liang, Senior Member, IEEE"
you will get ur answer

Hi there,
thanks for the reply.
Actually the same author in different publication (maximum eigenvalue detection) made me ask this question.

It is not clear why the narrow band signals have high correlation between its samples (the data samples should be independents). Unless if the channel is considered as frequency selective fading.(please tell me if I'm missing the type of the channel)

On the other hand, I've checked other literature s (Wireless Technologies Circuits, Systems and Devices, for Krzysztof Iniewski) which says: Page 148
"Note that over-sampling would correlate noise samples; thus, detection could
be always reduced to Nyquist sampling."

I know that if we increasing the number of sensed samples the detection performance will increase, but this is another story, isn't it?

actually, I've even done simulation for the author's paper :"Maximum Eigenvalue Detection: Theory and Application", but the MED (which based on the theory of correlation increase by oversampling) detection gives me the same results as energy detection.

best regards,

Eng.Abbasi said:

It is not clear why the narrow band signals have high correlation

Click to expand...

You should read up on the relation between the spectrum and autocorrelation function. The power spectral density (PSD) is simply the fourier transform of the autocorrelation function. A narrow bandwidth therefore implies a "broad" autocorrelation function.

"Note that over-sampling would correlate noise samples; thus, detection could
be always reduced to Nyquist sampling."

Click to expand...

The conclusion is correct, but saying it like that is strange and misleading. Over-sampling does not change anything about the signal properties. It _only_ changes your resolution. Oversampling is therefore pointless -- there is no new information. (Albeit it can be advantageous for certain low-complexity implementations.) It may even be counter-productive because it can cause ill-conditioning.

I know that if we increasing the number of sensed samples the detection performance will increase, but this is another story, isn't it?

Click to expand...

Absolutely.