Number of samples needed to determine standard deviation

[Guest]

I work at a company that develops measuring equipment and I want to document the performance of our instruments from a number of test. One of the properties a want to quantify is the standard deviation through a series of measurements. Furthermore to prove that I did not just get lucky and picked a well performing instrument I want to test a number of instruments and find the standard deviation of the performance between the instruments. So if I have to perform several test of several instruments, and let several emplyees do that (to rule out the influence of the employee) the number of test needed grows explosively. As testing an instrument is a time consuming task, I do not want to test more instruments than nesesary, while being "quite" certain that the standard deviation i get is correct. I have been browsing a litte around, but I have not been able to determine how many tests I must perform in order to have enough datapoint to be sure that the std. deviation i get from my samples is reliable. Many of the examples I have seen requires that you know the std. deviation in order to calculate the sample size, but this is not possible as the std. dev. is the parameter I want to determine.

So how many test do I you in general need to determine the std. deviation? 5,10,20...? Ofcourse the more test you make the closer your calculated std. dev. is to the real value, but there must be a point where you can be "quite" confident that your calculated value is reasonably reliable, and the added effort of collecting more data is not in proportion to the added precision gained, but how do you determine when to stop?
Can someone please point me in the right direction?

[Guest]

Well it depends on the accuracy, the more samples you do the more accurate your results would be. This is where confidence intervals come into play. I would tell you more but a simple google search will do.