nRF5340 FICR ID uniqueness

[Kelso]

Quick question.

Per  Unique Device ID  the unique ID in chips from around a decade ago were not guaranteed to be unique.

I'm wondering if this is still the case for a given nRF5340 chip, or if changes in manufacturing approach have affected the way this is handled.

If they're not guaranteed, is there a statistical bound on the likelihood two devices will have the same ID?

-Carl

[tesc]

Hi,

Both DEVICEADDR and DEVICEID are randomly generated in production, and therefore not 100 % guaranteed to be unique. That is correct. It is also correct for nRF5340 devices.

Assuming close to true randomness, the statistical likelihood of two devices sharing the same randomly generated ID can be calculated the same as for the solution to the Birthday problem.

I.e. with n devices using an ID with d possible values, the probability of at least one collision is approximately n^2 / 2d. (This uses square approximation, which works well for probabilities less than or equal to 1/2.) Accurate calculations are possible, but involve large factorials.

2^64 is an incredibly large number. You will therefore see:

• a one-in-a-billion (1:10^9) probability of collision at around two hundred thousand devices
• a one-in-a-million (1:10^6) probability at around six million devices
• a 1 % probability at around six hundred million devices

Note that the above probabilities are for the existence of one or more collisions among all of the devices involved; the probability of one specific device being involved in an ID conflict is much lower.

Regards,
Terje

[Kelso]

Thanks for confirming it's still random as well as the collision estimates.