Cell Relay Archive

[Alan Chan <achan@designpro.org>]

I did read Annex C of I.432 and it is a very good example. I just had problems
understanding the principle. Some good fellows referred me to some extremely
useful references:

1. Kim, S.C and B.G.Lee, " Synchronization of shift register
   generators in distributed sample scramblers", IEEE Trans.
   Commun., vol.42, no.3, pp-1400-1408, March 1994

2.  "Scrambling Techniques for Digital Transmission" by Byeong 
    Gi Lee and Seok Chang Kim; Springer-Verlag.

I also found out other references by the above authors. They clearly explain the
criteria for choosing sampling time, correction time and correction vector. They
also provide alternative descrambler methods which are simplier than that
illustrated in Annex C. For instance, the contiguous mode greatly simplifies the
logics to implement the descrambler.

Yes, we are targeting PON as specified in ITU-T Recommendation G.983.1. DSS is
specified for downstream transmission only. The upstream uses a simple
self-synchronizing polynomial because of the presence of overhead bytes in each
cell for clock and data recovery. I futilely searched for books about DSS until
I read the above IEEE papers. Maybe you correctly pointed out that DSS is rarely
used except in PON.

Marc Delvaux wrote:
> 
>   Because the DSS operation is quite difficult to grasp, the I.432 goes to great
> length in showing how it may be implemented (at least the part about how to 
> recover and use the scrambler bits).  You did read Annex C of I.432, didn 't you?  
> The modified HEC delineation is not different from the usual one and is not that 
> daunting.
> 
>   I was wondering if you had problems with the principle itself, or with the
> implementation. If I remember correctly, you already have HEC delineation (I 
> contacted your company for the PCI to UTOPIA device) so the HEC part should be 
>well known.
> 
>   May I ask which application is targeted, as far as I know the DSS is not really
> popular. The advantage it provides versus standard HEC is quite limited but is
> significantly more difficult to implement (and somebody probably had a bad trip 
> when he decided to use two non time adjacent bits).  The only application I know 
> of where for some APON system that was treated as a clear channel interface.
>