I'd like to discuss a very important quantity of noise. The noise is any behavior that deviates from the average behavior and clearly it controls the signal to noise ratio. So, if we can beat down the noise then it means we'll be able to see a smaller signal. So, if we think about the photogenerated current in the time interval tau, we can write that as i equals q times n times tau. Okay, so essentially N is the number of photogenerated electrons, tau is the time interval and q is the charge on the electron. So, we can now go ahead and write down the average photocurrent. The average photocurrent is nothing more than equal to q and then n and brackets which is essentially the average number of photogenerated electrons divided by tau. So, if the photogeneration of electrons is strictly puts on each event is independent of the others and there is no correlation between event. Essentially, we can write something nice. We can write essentially that the mean which is the average is equal to the variance of the distribution. So, remember that the variance is equal to sigma squared and so this is nothing more than i sub n squared which is equal to the average value of i minus the average value of i and this whole quantity squared and then average value taken. So, this is equal to average value of i squared minus the average value of i whole quantity squared. So, we can go ahead and write down an expression for the variance of the photocurrent. We can essentially write in bracket in squared is equal to the average of i minus average value of i whole thing squared and then average. So, this is equal to q over tau squared and then average value of n minus average value of n and then here this whole thing squared and the average taken. So, as I said if the photogeneration is put on, then we can basically take this expression for N and this is essentially just equal to the average value of the photogenerated electrons. So, now we can write down a simplified expression for the mean square of the noise current or the variance. So, in squared equals q squared, average value of n over tau squared and so if we want we can write this as q times the average value of the photocurrent divided by tau. We can think about relating the sample time to the bandwidth. So, the bandwidth b is equal to one over two times tau and parentheses. So, essentially my mean squared of the noise current is equal to two q I capital I times b. One very important noise source is shot noise. It's basically comes from the fluctuations in the carrier arrival times. It's present in all semiconductor, detectors and it presents a fundamental limitation on performance. So, if we think about typical values of shot noise, we can essentially think about a silicon P-I-N. So, for shot noise the dark current, I dark, essentially equals the square root of two q I bandwidth times bandwidth and if we think about a silicon PIN. A silicon PIN will essentially have values that range from 1.7 times 10 to the minus 15 to 1.8 times 10 to the minus 14 amps.