Re: [bluetooth] question...

[Niclas Hedhman <niclas@hedhman.org>]

On Friday 17 January 2003 16:10, Alban Villain wrote:
> It is not specific to BT stacks. It concerns data transmission systems in
> general, and wireless systems in particular. It has nothing to see with
> implementation.
>
> Do you agree or not that corrupted baseband packets can pass all the
> baseband tests ?
>   - FEC2/3 is not able to correct or detect all errors ;
>   - CRC may not detect all errors ;
>   - So there is a very little but non zero probability that corrupted
> packets passes through baseband checkings...

I agree on that. In fact, the probability may be higher than one think.
Even a 2^32 chance is pretty often. E.g. I produce 1 million devices that 
sends 1000 packets per hour on average, and 10% of packets contains random 
data, we are looking at a "slip through" every 40 hours or so.

> For your products, did you experience unexplained crashes ?

I wish. I haven't come so far in implementation. But I have faced exactly the 
same problem for non-wireless systems. Typically it boils down to raising the 
"probability" by dozens of magnitudes in the upper level protocols.

My concern, considering your statement, was that the BT stack by itself is not 
fault-tolerant enough to handle corrupted packets at a higher level. That 
sounded worrying, and if it is so by design I will cancel my project right 
away. If it is so by implementation in available stacks, it can be fixed.

Cheers

Niclas

> The time between 2 crashes due to a non detected corrupted packet may be
> very large. I didn't estimate it, but it is possible to do it using
> probabilistic properties of error correction and detection codes used in
> Bluetooth. It may be many months, or years.
>
> Alban

--
To unsubscribe from bluetooth mailing list please visit http://www.opencores.org/mailinglists.shtml