Determining sample size and acceptable practical difference

Hi all,

I need to conduct testing to prove that there is no practical difference between current and proposed manufacturing method. In this case I would be testing object weight. I have tested historical data to determine what is the current process. Based on almost 7,000 samples, Cpk value of current process is 0.732, SD - 0.0004014, data is not normal. Because I have quite a poor process to work with, I can't really accept it going any worst, therefore based on SD and CPK I have calculated acceptable Practical Difference to be 0.000037.

Based on all of the above, Sample Size calculation (Alpha = 0.05, Power = 0.95) sample size required to detect such a difference is 6,120/2 for comparing Two Means.

This is a problem, as I can't afford to test such a high number of samples, my practical limits on sample size are 100-150 units. Also, I'm not looking to compare my test results versus historical data, I will be completing "head to head" Control & Test groups analysis, but I need statistical rational on those groups size and acceptable practical difference.

Any ideas on how I can complete such a testing with small sample size? Given that my current process is quite poor and I can't afford any negative change in it.

Thanks for your help