Stats question?

An engineer has proposed a change to a batchwise manufacturing processthat is expected to improve the yield.Some historical data is available for batches made prior to the change.If the change is expected only to shift the mean (i.e. not change the process variation), what number of batches would be required to confirm a yield improvementof 4percentage points with a Type I and Type II error of 5%? What number of batches would be needed if the shift was only 1 percentage point?