[Tinyos-devel] CTP and LPL on tmote

[Andreas Köpke]

Omprakash Gnawali wrote:
> On Sat, Dec 6, 2008 at 12:03 AM, Andreas Köpke <koepke at tkn.tu-berlin.de> 
wrote:
> > Hi guys, looks like we isolated the problem.
> >
> > Using two nodes about 3m apart, one sender, one receiver, one message
> > every 1.5s on average we get the following delivery ratios:
> >
> > DC        mean    90% confidence interval
> > 5%        0.097   0.086   0.11
> > 10%      0.21     0.126   0.295
> > 20%      0.26     0.18   0.34
> > 50%      0.55     0.46   0.63
> > 100%    0.995   0.98    1.00
> >
> > Once we switch on LPL, the packets are hardly detected by the receiver.
>
> I don't remember things being this bad.
>
> Is your guess that if we used test programs with something like
> periodic packet transmission (lets say every 1s) and a base station,
> we will have similar problem? I would have a hard time believing that
> is the case but I haven't done this test in a while.
>
> - om_p

The setup is simple: 
- The sender sends unicast packets to the receiver.
- Both use the same duty cycle. 
- The next packet transmission is started randomly after the sendDone event of    
  the previous packet. (uniform in interval [0:2999] ms including borders)
- the delivery ratio is computed using gaps between the transmitted sequence 
  numbers offline using a trace file
- duplicates are removed (sometimes UniqueReceive 
  lets duplicates slip, but that is not a big issue)
- no changes to current CVS version of LPL

The posted result is a subset of my measurements that will take another day or 
so to complete. But it looks like only the confidence intervals will become 
tighter.

So yes, everybody should be able to reproduce the problem using the current 
CVS version.

Best, Andreas