We use LM285BXZ (National Semiconductor) as a reference in a power measurement application. The design intention was to use a fixed voltage reference at 1.2V. In the circuit pin 3 (feedback) was left floating (it is a NC in the fixed parts). We have several hundered products in the field with this flaw and have now seen a small shift in our power measurement. In house we were able to mitigate this by connecting the FB pin to the anode as per the data sheet. The problem is seen infrequently and is difficult to repeat. Can you tell me what the expected behavior is in the event the FB pin is left floating? It is a reasonably noisy RF environment. Any advice is greatly appreciated.]