We are currently developing a sensor based on the ADuCM355. Each sensor needs a unique ID so we decided to read the 16-byte / 128-bit serial id. Unfortunately our host system only has room for 32-bit sensor id's. Now we noticed that of the 128 bits some bits are equal on our chips, others are different (n=5):
UID:041e8712 389cbfc7 0006135a 33373932
UID:3c1e8712 385c7e9f 0004072d 33373932
UID:841e8712 399ccee7 0006024a 33373932
UID:401e8712 39dcbe9f 00062e1f 33373932
UID:881e8712 38dcce7f 00060659 33373932
It looks like the 3rd column is a serial number and the other bits are more or less stable.
My question: if we only use the 3rd set of 32-bits, can we guarantee that the number is unique? It would still allow for >4 billion devices, I don't think that many ADuCM355's will be produced...