Thread Links Date Links
Thread Prev Thread Next Thread Index Date Prev Date Next Date Index

Re: [8023-10GEPON] Comment #44



All,

Whether we think any of these quantities are peak-to-peak, RMS or whatever, the units (UI) are not.  We should not mix together units and descriptions of the metrics.

Compare the similar tables already in 802.3: 38-10, 58-8, 59-9, 60-10 and 60-11.  They just say just "UI" and "ps".  Table 75C-1 and Table 75C-2 are similar.

If we have consensus that the DJ, RJ and TJ in 802.3av/D3.3 mean what MJSQ mean by these terms, these are not RMS, but they are not peak-to-peak either.  If peak means
" a: the highest level or greatest degree
  b: a high point in a course of development especially as represented on a graph"
Then we take peak-to-peak as highest - lowest, or in this case latest - earliest.  That's what the "peak-to-peak jitter" function on an oscilloscope measures; a different metric to any of TJ, DJ and RJ.

Paraphrasing MJSQ,
TJ is the jitter to 10^-12 BER.  There is more jitter outside this range (although very uncommon) so it can't be "peak-to-peak".

RJ is found from the skirts of the jitter distribution.  It isn't "peak-to-peak" either.

In MJSQ, DJ is found from extrapolating from the skirts of the jitter distribution.  It certainly isn't "peak-to-peak".  It isn't even the peak-to-peak measure of the jitter that's deterministic!

If people feel that simply deleting "p-p" could cause confusion, then something could be added, e.g.

TJ and RJ are in relation to the times where the BER is 10^-12.
and/or
See FC-MJSQ.

This could be as table notes, a NOTE in 75C.1, or regular text.

Adding FC-MJSQ, which is Fibre Channel - Methodologies for Jitter and Signal Quality Specification ANSI INCITS TR-35-2004, to the list of references would be helpful, but we may wish to keep changes to a minimum at this stage in the project.

However, if we simply delete "p-p", 75C will be aligned with clauses 38, 58, 59 and 60.

Piers

> -----Original Message-----
> From: Seiji Kozaki [mailto:Kozaki.Seiji@xxxxxxxxxxxxxxxxxxxxxxxxxxx]
> Sent: 08 June 2009 01:23
> To: STDS-802-3-10GEPON@xxxxxxxxxxxxxxxxx
> Subject: Re: [8023-10GEPON] Comment #44
> 
> Tibi-san,
> 
> Thank you for your response.
> Now, we got a common understanding about this issue as below:
> "We disagree with removal of the "p-p" from the units for any 
> elements 
> of jitter".
> 
> Regards,
> S.Kozaki
> 
> 
> Tibi Galambos さんは書きました:
> > Dear Kozaki-san and Hamano-san
> >
> > Sorry for the possible misunderstanding. As long as the BER 
> is specified, I do agree that the units can in a certain 
> sense be UIp-p, and if that is the customary I have no 
> objection at it. As long as we all agree that the 
> mathematical model (for Rj and Tj) behind it is an unbounded 
> distribution and the "p-p" value is defined by the probability (BER).
> >
> > What I certainly can not agree is the removal of the "p-p" 
> from the units for Dj.
> > So whether the "p-p" is left in or not for Rj and Tj it is 
> fine with me.
> >
> > Thanks for your attention,
> >                             Tibi 
> >
> >
> > -----Original Message-----
> > From: Seiji Kozaki [mailto:Kozaki.Seiji@xxxxxxxxxxxxxxxxxxxxxxxxxxx]
> > Sent: Sun 6/7/2009 6:29 PM
> > To: STDS-802-3-10GEPON@xxxxxxxxxxxxxxxxx
> > Subject: Re: [8023-10GEPON] Comment #44
> >  
> > Tibi-san, and all
> >
> > I agree with Hamano-san's opinion.
> > In the meeting of Sept/2008, TF defined Rj value in UIp-p,
> > which is calculated by using RMS value.
> > Thus, the unit of specifications for Tj and Rj are should be UIp-p.
> >
> > Regards,
> > Seiji Kozaki
> >
> >
> > Hiroshi Hamano ????????:
> >   
> >> Dear Dr. Galambos,
> >>
> >> As indicated in 3av_0809_kozaki_2.pdf, RJ numbers in the 
> Tables are 
> >> not rms, but DJ aligned value to calculate TJ.  Even 
> though RJ does 
> >> not have the peak-jitter nature, I still think DJ, RJ, and 
> TJ numbers 
> >> should all be described in UIp-p.
> >>
> >> Best regards,
> >> Hiroshi Hamano
> >> Fujitsu Labs. Ltd.
> >>
> >> %% Tibi Galambos <Tibi_Galambos@xxxxxxxxxxxxxx>
> >> %% [8023-10GEPON] Comment #44
> >> %% Sun, 7 Jun 2009 09:49:52 +0300
> >>
> >>   
> >>     
> >>> 	The proposal for comment #44 is to remove "p-p" from the headers
> >>> of tables 75C-1 and 75C-2 altogether. This is not correct.
> >>>
> >>> 	The "p-p" attribute has to be removed from the TJ and RJ columns
> >>> only. The headers of both tables have to look as follows:
> >>> 	
> >>> Reference point	DJ (UI p-p)	RJ (UI )	TJ (UI )	
> >>>
> >>> 				
> >>> 	Justification:
> >>>
> >>> 	The jitter budget is built upon the following assumptions:
> >>> a.	Jitter is represented assuming the DJ to have 
> an equi-probable
> >>> bimodal distribution and RJ to be Gausian.
> >>> b.	All sources of random jitter are assumed 
> independent therefore
> >>> RJ rms values can be added by squares.
> >>> c.	All sources of DJ are assumed to be correlated 
> (this is a worst
> >>> case assumption, meaning that all DJ components will be 
> either together
> >>> at max value or together at min value, with equal 
> probability for the
> >>> min and the max to occur)
> >>> 	Under these assumptions, RJ and TJ are defined @ BER while DJ is
> >>> defined by it's peak to peak value and then the following 
> calculation
> >>> holds:
> >>> 			TJ (@ BER) = DJ p-p + RJ (@ BER)
> >>>
> >>>
> >>>
> >>>
> >>> Tibi Galambos 
> >>> Principal Engineer AFE (Analog Front-End)
> >>> FTTH BU
> >>> PMC-Sierra
> >>> Tel: +972-9-9628000 Ext. 473
> >>> Email: tibi_galambos@xxxxxxxxxxxxxx
> >>> <mailto:itibi_galambos@xxxxxxxxxxxxxx>
> >>>     
> >>>       
> >> ---
> >> -----------------------------------------
> >> Hiroshi Hamano
> >> Network Systems Labs., Fujitsu Labs. Ltd.
> >> Phone:+81-44-754-2641 Fax.+81-44-754-2640
> >> E-mail:hamano.hiroshi@xxxxxxxxxxxxxx
> >> -----------------------------------------
> >>   
> >>     
>