RSSI, proximity and distance

What RSSI is Known fact

When iOS reports a discovered peripheral it includes the received signal strength indicator, "the current received signal strength indicator (RSSI) of the peripheral, in decibels" in Apple's words. Apple suggests it can be used to determine proximity. It is a measurement made by the phone's radio of how much power arrived, expressed in dBm, typically between about −30 (very close) and −100 (barely receivable). In Core Bluetooth a value of 127 means unavailable.

Why it is not distance Known fact

Received power depends on far more than distance:

• Transmit power. Devices transmit at different, sometimes adaptive, power levels. Some advertisements include a transmit-power field; most of the devices we study are not known to include it.
• Antenna orientation. Turning a phone or a pair of glasses can change the received power by more than the difference between one and three metres.
• The human body. Water absorbs 2.4 GHz energy efficiently. Glasses on a head, a phone in a pocket, and a person standing between the two all reduce the signal substantially.
• Reflections. Indoors, signals arrive by several paths and interfere with each other, so RSSI fluctuates from one packet to the next even when nothing moves.
• Receiver differences. Different iPhone models, cases, and even how the phone is held change the measurement.

Inference Given all this, a formula that turns a single RSSI reading into metres would produce a confident-looking number that is often wrong by a factor of two or more. Presenting such a number would be a claim the data cannot support.

What the app does instead Known fact

RSSI is shown as one of four coarse categories: very strong (≥ −50 dBm), strong (≥ −65), moderate (≥ −80) and weak (below −80). Thresholds are editable in the research harness and will be revisited after Block B of the validation protocol, which records RSSI at 0.5, 1, 2, 5 and 10 metres in a quiet room and a busy space. The app also keeps a smoothed value (an exponential moving average with α = 0.3) and a short rolling average to damp packet-to-packet noise. It never displays metres.

What we still have to measure Unknown

• Whether the four categories separate at all for the glasses we study, or whether body attenuation compresses everything into "moderate" and "weak" at normal social distances.
• The packet-to-packet variance at fixed distance, which determines how much smoothing is needed before a category is stable.
• Whether any of the devices include a transmit-power field, which would make path-loss estimates slightly less unreliable.

Related pages

• How Bluetooth detection works
• False positives and false negatives
• FAQ