In BLE, whether due Slave latency can Window widen CI/2 -TIFS can happen which leads to connection lost ?

[Hrishikesha]

For eg:Lets consider

Master Clock Accuracy = 500PPM

Slave Clock Accuracy = 500PPM

ConnInterval = 10ms

CE length = 625us

Slave Latency = 499

STO = 11s

Calculation: Window Widening(WW) = ((masterSCA + slaveSCA)/1000000 )* TimeSinceLastAnchorpoint

WW value at which connection is lost = (CI/2) - TIFS = 5ms - 150us = 4850us

In 0th conn Event: Master Send and received a packet and closes the connEvent

After 0th event slave will be in latency for next 499 connEvent

In 1st ConnEvent : WW = 10^-3 * 10ms(timesinceLastAnchorpoint which is CI) = 10us

.

.

.

.

.

At 486th event only WW reaches > (CI/2) - TIFS, which will cause the slave to lose the connection.(Refer Bluetooth Version 4.2[Vol 6 , Part B],section 4.5.7,Last Para)

But the thing here is slave will not be listening during slave latency, so there is no point to detect a anchor point.Because of slave latency , WW should not reach (CI/2) - TIFS and lose connection.

[JohanSt]

The SoftDevice will try to keep the link alive beyond the point where the window widening becomes larger than (CI/2 - TIFS) in the rare situation of a high slave latency, short connection interval and inaccurate clock.

This can result in a disconnection if two peers drift away such that they use different channels. It may also lead to a disconnection if the connection event counters become different, and a control procedure involving an instant is executed.