nRF52833 / nRF5340: How many simultaneous AoA transmitters can a single locator handle?

[Pravyuha]

Hi,

For connectionless Direction Finding (AoA), where a locator device scans for extended advertising packets containing a CTE from multiple independent transmitting devices (each transmitting roughly once per 1-2 seconds, unconnected/broadcast mode):

1. Is there a practical limit on how many distinct transmitting devices a single nRF52833 or nRF5340 locator can process AoA for, given they're all broadcasting independently and not connected to the locator?

2. Since this is connectionless (scan-based, not connection-based), does device count scaling behave differently than it would for a connection-oriented procedure? I.e., is the practical limit driven more by radio scan window/duty cycle and IQ sample processing time per packet, rather than a connection count limit?

3. If many transmitters send CTE-bearing advertising packets in close time proximity, is there a risk of the locator missing or incorrectly processing some of them, and if so, is there guidance on advertising interval / randomization to avoid packet collisions between multiple independent transmitters?

4. Would the nRF5340 handle a higher number of simultaneous AoA sources than the nRF52833, given its additional processing core, or is the RADIO peripheral itself (shared design between the two) the actual bottleneck rather than CPU power?

Thanks.

[Amanda Hsieh]

Hi, 

Is there a practical limit on how many distinct transmitting devices a single nRF52833 or nRF5340 locator can process AoA for, given they're all broadcasting independently and not connected to the locator?

Yes, it depends on available time on air, memory usage, CPU usage, etc., and is application-dependent.

2. Since this is connectionless (scan-based, not connection-based), does device count scaling behave differently than it would for a connection-oriented procedure? I.e., is the practical limit driven more by radio scan window/duty cycle and IQ sample processing time per packet, rather than a connection count limit?

Yes, the scaling behavior is fundamentally different in connectionless mode. The practical limit is driven mainly by the number of periodic syncs and single-radio scheduling, with scan duty cycle important for join/discovery, and CPU important mainly for post-IQ processing, not as a substitute for MAX_CONN.

3. If many transmitters send CTE-bearing advertising packets in close time proximity, is there a risk of the locator missing or incorrectly processing some of them, and if so, is there guidance on advertising interval / randomization to avoid packet collisions between multiple independent transmitters?

Yes. With many independent tags, periodic events can overlap in time. The locator has one RADIO, and the SoftDevice Controller schedules periodic sync RX as timing-events that may collide with other timing-events. See SDC — timing when synchronized to a periodic advertiser (collisions) and DF video — overlap and single radio.

You can try:

1. Use longer intervals for static assets, shorter when moving (See DF video).
2. Application policy: terminate sync for known-static tags to free slots and schedule headroom (See DF video).
3. Keep periodic advertising events short (payload, PHY, chain length) to reduce scheduling conflicts (See SDC — periodic advertiser timing).

4. Would the nRF5340 handle a higher number of simultaneous AoA sources than the nRF52833, given its additional processing core, or is the RADIO peripheral itself (shared design between the two) the actual bottleneck rather than CPU power?

The RADIO peripheral (which handles antenna switching and IQ sampling) is the shared hardware bottleneck. There is only one radio on either SoC, so the fundamental radio-time constraint is the same. The nRF5340's dual-core architecture (application core + network core) means the angle calculation algorithms can be offloaded to the application core, freeing the network core for radio operations. This could allow faster IQ sample processing and potentially support a higher throughput of AoA reports. See BT Direction Finding intro On the nRF52833, if angle algorithms run on the same CPU handling the radio, it becomes CPU-limited at high packet rates. See BT Direction Finding intro.

Regards,
Amanda H.

[Pravyuha]

Given periodic advertising intervals of 1-2 seconds and short CTE payloads, roughly how many concurrent periodic syncs is realistic before collision rate becomes a practical problem?