How much of NRF51822 64-bit DEVICEID is actually set?

[Gavin]

I am using DEVICEID as an identifier (I know it's not necessarily unique) but when I break the two 32-bit numbers out into bytes, I am finding that the three most-significant bytes of DEVICEID[1] are zero. Is this a freak occurence, or are only five of the eight bytes set?

[endnode]

I believe that Device ID is in fact 48-bit number (6 bytes) matching Advertising Address (MAC) format defined by BT SiG. Beside Nordic's Reference Manual (where are really 2 32-bit words allocated) there is no single word about length of this ID. Note that when used as Adv. Address most significant bits are used to signal address type making the ID effectively only 46-bit. I'm not sure if it is granted that highest bits of DEVICEID are set to "Radnom Static" type so I treat them explicitly in the FW (applicable only if you are changing DEVICEADDR during run-time, otherwise it is set automatically).