Think of it in terms of independent variables. Right now, for each data point, you have 2 observers, one positioning so 2x1. You need to have a third observer, so now you will be 3x1. If you change the positioning, you will have 3x2 or more and you won't be able to get information about observer one in position 2. This is drastically increasing the number of data points you need. IMO, stick with your original experimental design.
Of course in retrospect if you believe your original design was faulty, by all means scrap that and start again.