Bluetooth vs. ZigBee signal strength measurement accuracy

[jamesford]

Hey I hope that some can give me an explanation for my observations.
Its a electronic/physics question.

So at UNI I was rebuilding a positioning system which is based on signal strength.
The system simply collects the RSS (received signal strength) and calculates the probabilities for the position. Sensors are attached to the walls and you then measure the RSS between them to find out the position of the person in the room. So the person does not need to wear an device.

The origin paper which read to understand the basic concept is using ZigBee devices (TI CC2531 with PCB antenna AN043).

I thought why not using Bluetooth. Its the same frequency range, my devices have the same antenna design AN043, sample rate (10 measurements per sec per link) is the same and its cheaper. So I build my system with 25 BLE sensors.

Two things were different:
1. They used frequency hopping on 3 channels - I didn't, just one channels
2. They used every node as sender and receiver - I had 25 senders and 3 receivers.

As result I expected simply less accuracy. But what I got is like random data.

My first experiment was one sender and one receiver. Which worked - like a light barrier.
But as soon as I deployed and used all devices in the room it became strange.
Observing the links (line of sight) between sender and receiver when nobody is in the room shows that all RSS are stable, like expected.
But when u enter the room its like someone would throw a stone into water. All reciever measure a change in RSS even if you are 6 meters away.

The ZigBee system in this case was able to detect the positions because only links which were interfered by the persons presence had a change in RSS.

Do you have any ideas my Bluetooth performce so bad in this device free positioning experiment ?

Is the Bluetooth Signal Strength measurment less accurate ? if, yes whats the reason (hardware)?
Is the frequency hopping maybe so important?

I happy to hear all of your theories.

Another observation is this (img in attachment):
The images shows a historgram (on top is oldest).
Each horizontal line shows a 3 second bucket for signals.
The coloration show the count of signals for the certain RSS.
Strange is that is kind of clusters in 3 lines.
I thought about fresnel zones. Any Ideas ?

Thanks guys.

[w2aew]

The piece of the puzzle that you aren't considering is the modulation type.  Zigbee uses direct sequence spread spectrum, where the low datarate data is encoded (spread) into longer PN code words for transmission.  This makes it more tolerant to certain impairments than the higher datarate GFSK (or pi/4 DQPSK, or 8DPSK) used by the various Bluetooth variants.  Bluetooth achieves much higher data throughput, and some of the impairment tolerance comes from the frequency hopping.

[jamesford]

Thanks, w2aew.

Never thought about this, but it makes sense.

In the meantime I also thought about multi-path tolerance.
BLE seems to be much more tolerant and receives more reflected signals which then have a lower RSSI.
ZigBee is not so tolerant an mainly receives signals directly (maybe looses more signals), which makes the RSSI more stable.

What do you think ?

btw: pretty cool yt channel.