Negative sign means Noise Power > Signal Power

(as Signal Power [dB] - Noise Power [dB] < 0).

It just means a lot of noise. There are situations in which the receiver copes very well even with this noisy environment, in Spread Spectrum Transmission, where the signal is "buried" deep into the noise. Everytime you use GoogleMaps with GPS on your phone, the chip there extracts the signal out of the noise. On a spectrum analyser or on a scope, if u plotted the antenna signal, it would look like noise to you, but the signal is there it has been "spread thin" over a large frequency, such that its spectral density (power for every Hz) becomes lower than that of the noise. Matching filtering (correlation with the transmitted spreading sequence) is able to pull the signal out of the noise.

10log.... is used for power values, 20log... if you use voltage or amplitudes.

In your basic formula η= -10log⁡10(1/y^2 ) does the "1" stands for "signal"? In that case you have normalized the noise values y to the signals values.

No, the values will not be always negative: log(a/b)=log(a)-log(b). So -log(a/b)=log(b)-log(a). If log(b)>log(a) then the expression is positive. By the way, log(1)=0, so it all depends on the value of y in your expression. I understand you are using a normalised version of the expression, as noted above.

Negative sign means Noise Power > Signal Power

(as Signal Power [dB] - Noise Power [dB] < 0).

It just means a lot of noise. There are situations in which the receiver copes very well even with this noisy environment, in Spread Spectrum Transmission, where the signal is "buried" deep into the noise. Everytime you use GoogleMaps with GPS on your phone, the chip there extracts the signal out of the noise. On a spectrum analyser or on a scope, if u plotted the antenna signal, it would look like noise to you, but the signal is there it has been "spread thin" over a large frequency, such that its spectral density (power for every Hz) becomes lower than that of the noise. Matching filtering (correlation with the transmitted spreading sequence) is able to pull the signal out of the noise.

Thank you all for your replies.

Are there any SDR's working with very low SNR's practically. In MATLAB simulation, we can generate the signals with SNR of even -50 dB and perform spectrum sensing. But is it practically feasible in the case of realtime applications.

I can give you easy understandings regarding your topic. Please post here your further query.

As SNR is simply the ratio , and we know that ratio should be anything i.e positive or negative. But for practical case SNR should always be positive value means signal power should always be greater than the noise power. If it is negative than that system has no significance....

No, Vikram Kumar!

The issue is that SNR is expressed in a logarithmic scale, as SNR=20log(rms(signal)/rms(noise)), so, if the rms of the noise is bigger than the rms of the signal the SNR will be negative. If they are the same, then RMS(1)=0, and if signal is bigger, then SNR>0.

@Vikram Kumar "... for practical case SNR should always be positive .."

I bet that your satellite antenna receives very poor signals with negative SNR,but can still reconstruct a TV picture from it.

Thanks to autocorrelation!

SNR (in dB) is 20 log (rms(signal)/rms(noise), so it is the log of a ratio. If the ratio is >1 then SNR is positive, but if the ratio is

@Fernando

Yes, in my programs I use a variable called NSR, if the ratio SNR will be negative. After all you will get only positive numbers.

:-)

@Marc Schmieger

I call that cheating :-))

And it will hide the fact that the noise is bigger than the signal.

Of course, you know what you are doing and how to interpret that, but there will be newbies who would be fooled and interpret the information wrongly.

:-)

There is no cheating if you have both variables available: SNR and NSR. If one doesn`t apply, the other variable will be set to zero. E.g. if SNR < 1, then NSR == |SNR| and SNR == 0.

Just to confuse the audience.