I am reading a digital signal processing book "The Scientist and Engineer's Guide to Digital Signal Processing" 2nd Edition by "Steven W. Smith" I could not understand the following concept in its second chapter QUOTE Each random number has a value between zero and one, with an equal probability of being anywhere between these two extremes. Figure 2-10a shows a signal formed by taking 128 samples from this type of random number generator. The mean of the underlying process that generated this signal is 0.5, the standard deviation is, and the 1 / (12)^0.5 = 0.29 distribution is uniform between zero and one. UNQUOTE I couldn't derive standard deviation = 1 / (12)^0.5 my derivation is as follows Mean = Sum(xi, i=0 to n-1) / n = 0.5 variance= (1/n)*Sum( (xi-Mean)^2) {i=0 to n-1} = (1/n)*Sum( (xi^2 + Mean^2 - 2*xi*Mean)) {i=0 to n-1} = (1/n)*(Sum(xi^2) + Sum(Mean^2) - Sum(2*xi*Mean)) {i=0 to n-1} = (1/n)*Sum(xi^2) + (1/n)*Sum(Mean^2) -(1/n)*Sum(2*xi*Mean) {i=0 to n-1} = (1/n)*Sum(xi^2) + (Mean^2) - (2*Mean)(1/n)*Sum(xi) {i=0 to n-1} = (1/n)*Sum(xi^2) + (Mean^2) - (2*Mean^2) {i=0 to n-1} = (1/n)*Sum(xi^2) - (Mean^2) {i=0 to n-1} Now we know the value of mean (which is 0.5) but how can we know the value for Sum(xi^2)

# Help needed in Finding of Standard deviation of Random Noise

Started by ●March 6, 2004

Reply by ●March 6, 20042004-03-06

Abdul Nasir Khan wrote:> I am reading a digital signal processing book > > "The Scientist and Engineer's Guide to Digital Signal Processing" > 2nd Edition > by "Steven W. Smith" > > > I could not understand the following concept in its second chapter > > QUOTE > Each random number has a value between zero and one, with an equal > probability of being anywhere between these two extremes. Figure 2-10a > shows a signal formed by taking 128 samples from this type of random > number generator. The mean of the underlying process that generated > this signal is 0.5, the standard deviation is, and the 1 / (12)^0.5 = > 0.29 distribution is uniform between zero and one. > UNQUOTE > > I couldn't derive > > standard deviation = 1 / (12)^0.5 > > my derivation is as follows > > Mean = Sum(xi, i=0 to n-1) / n = 0.5 > > variance= (1/n)*Sum( (xi-Mean)^2) {i=0 to n-1} > = (1/n)*Sum( (xi^2 + Mean^2 - 2*xi*Mean)) {i=0 to n-1} > = (1/n)*(Sum(xi^2) + Sum(Mean^2) - Sum(2*xi*Mean)) {i=0 to n-1} > = (1/n)*Sum(xi^2) + (1/n)*Sum(Mean^2) -(1/n)*Sum(2*xi*Mean) {i=0 to > n-1} > = (1/n)*Sum(xi^2) + (Mean^2) - (2*Mean)(1/n)*Sum(xi) {i=0 to n-1} > = (1/n)*Sum(xi^2) + (Mean^2) - (2*Mean^2) {i=0 to n-1} > = (1/n)*Sum(xi^2) - (Mean^2) {i=0 to n-1} > > > Now we know the value of mean (which is 0.5) but how can we know the > value for Sum(xi^2)The claim of uniform distribution allows you to reorder the sequence to make an arithmetic series. You are effectively calculating the moment of inertia of a stick about its center of gravity. Does that help? Jerry -- Engineering is the art of making what you want from things you can get. �����������������������������������������������������������������������

Reply by ●March 6, 20042004-03-06

Abdul Nasir Khan wrote:> I am reading a digital signal processing book > > "The Scientist and Engineer's Guide to Digital Signal Processing" > 2nd Edition > by "Steven W. Smith" > > > I could not understand the following concept in its second chapter > > QUOTE > Each random number has a value between zero and one, with an equal > probability of being anywhere between these two extremes. Figure 2-10a > shows a signal formed by taking 128 samples from this type of random > number generator. The mean of the underlying process that generated > this signal is 0.5, the standard deviation is, and the 1 / (12)^0.5 = > 0.29 distribution is uniform between zero and one. > UNQUOTE > > I couldn't derive > > standard deviation = 1 / (12)^0.5 > > my derivation is as follows > > Mean = Sum(xi, i=0 to n-1) / n = 0.5 > > variance= (1/n)*Sum( (xi-Mean)^2) {i=0 to n-1} > = (1/n)*Sum( (xi^2 + Mean^2 - 2*xi*Mean)) {i=0 to n-1} > = (1/n)*(Sum(xi^2) + Sum(Mean^2) - Sum(2*xi*Mean)) {i=0 to n-1} > = (1/n)*Sum(xi^2) + (1/n)*Sum(Mean^2) -(1/n)*Sum(2*xi*Mean) {i=0 to > n-1} > = (1/n)*Sum(xi^2) + (Mean^2) - (2*Mean)(1/n)*Sum(xi) {i=0 to n-1} > = (1/n)*Sum(xi^2) + (Mean^2) - (2*Mean^2) {i=0 to n-1} > = (1/n)*Sum(xi^2) - (Mean^2) {i=0 to n-1} > > > Now we know the value of mean (which is 0.5) but how can we know the > value for Sum(xi^2)You have a "uniform distribution". The mean and variance of a uniform distribution between 'a' and 'b' are: 1/2 * (a + b) <-- mean 1/12 * (b - a)^2 <-- variance See this site: http://mathworld.wolfram.com/UniformDistribution.html Good luck, OUP

Reply by ●March 7, 20042004-03-07

Abdul, You're going about this the wrong way. Theoretically, uniform random variables have a continuous probability density function and so you should compute its variance using that definition (look this up in any probability book): Variance(x) = integral of (x - 0.5)*(x - 0.5) dx where the integration is over the interval zero to one. You can use the samples to estimate the variance of your samples, but you cannot use them to compute the variance of the "underlying process". Mike Abdul Nasir Khan wrote:> I am reading a digital signal processing book > > "The Scientist and Engineer's Guide to Digital Signal Processing" > 2nd Edition > by "Steven W. Smith" > > > I could not understand the following concept in its second chapter > > QUOTE > Each random number has a value between zero and one, with an equal > probability of being anywhere between these two extremes. Figure 2-10a > shows a signal formed by taking 128 samples from this type of random > number generator. The mean of the underlying process that generated > this signal is 0.5, the standard deviation is, and the 1 / (12)^0.5 = > 0.29 distribution is uniform between zero and one. > UNQUOTE > > I couldn't derive > > standard deviation = 1 / (12)^0.5 > > my derivation is as follows > > Mean = Sum(xi, i=0 to n-1) / n = 0.5 > > variance= (1/n)*Sum( (xi-Mean)^2) {i=0 to n-1} > = (1/n)*Sum( (xi^2 + Mean^2 - 2*xi*Mean)) {i=0 to n-1} > = (1/n)*(Sum(xi^2) + Sum(Mean^2) - Sum(2*xi*Mean)) {i=0 to n-1} > = (1/n)*Sum(xi^2) + (1/n)*Sum(Mean^2) -(1/n)*Sum(2*xi*Mean) {i=0 to > n-1} > = (1/n)*Sum(xi^2) + (Mean^2) - (2*Mean)(1/n)*Sum(xi) {i=0 to n-1} > = (1/n)*Sum(xi^2) + (Mean^2) - (2*Mean^2) {i=0 to n-1} > = (1/n)*Sum(xi^2) - (Mean^2) {i=0 to n-1} > > > Now we know the value of mean (which is 0.5) but how can we know the > value for Sum(xi^2)