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

RE: PCS Tx Diagram (Clause 48)




Rich,
Please see my comments below:

> -----Original Message-----
> From: Rich Taborek [mailto:rtaborek@xxxxxxxxxxxxx]
> Sent: Tuesday, November 21, 2000 10:17 PM
> To: HSSG
> Subject: Re: PCS Tx Diagram (Clause 48)
> 
> 
> 
> Boaz,
> 
> See my responses below:
> 
> Boaz Shahar wrote:
> > 
> > Rich,
> > Couple of remarks about the transmit state machine in D1.1:
> > 
> > Figure 48-7: Tx Source State Machine
> > 
> > 1.Sending /A/ or /K/ as the first column in the IDLE 
> pattern should be done
> > with probability ~0.5 for each. So, it seems enough to 
> choose each of them
> > based on the value of code_sel (code_sel=1: Send A, 
> code_sel=0: Send /K/).
> > Thus, I think that the qualification by A_CNT_0 should be 
> eliminated.
> 
> Qualification by A_CNT_0 is required to insure that ||A|| is 
> sent with a
> minimum spacing. In the case of a short normal or erred 
> packet, A_CNT_0
> qualification prevents the issuance of ||A|| with less than 
> the minimum
> required spacing.

Two points:
1.Normal short packet is 72 bytes include preamble. This is 18 columns which
is more than the mimimum spacing anyway.
2.The MAC starts to transmit packets only after de dkewing completion in the
other side. After de skewing completion, you do not need the minimum spacing
any more. So in a case of erronous frame, I see no problem. 


>  
> > 2.The path from the "SEND_K" box to the "SEND_DATA" box 
> means that there
> > might be an IDLE pattern which is composed from a single 
> /K/ column? I think
> > that this branch should not appear. The "B" point should 
> always be the next
> > one, because it is associated with the first /R/ column 
> that must exist in
> > the IDLE pattern
> 
> That path covers the case of the anything other than Idle coming over
> the XGMII/RS. It will be sent as is.

So an IDLE pattern composed of first /K/ column and then something from the
MAC is acceptable?

> 
> > Figure 48-8 PCS transmit idle and Signal state diagram
> > 
> > 1.A general remark: When a SIGNAL event occurs, the machine which is
> > generating code_sel and A_CNT_0 are freeze? I'm asking that 
> because if it
> > not so, then it may happen that the A_CNT_0=1 event occurs 
> exactly when
> > TX=SIGNAL occurs. If this happens, then the /A/ column is 
> not transmitted:
> > Instead, the Encode(SIGNAL) is transmitted.
> 
> Yes. This is correct. SIGNAL may occur anytime. Please note 
> that SIGNAL
> and SEQUENCE are not part of the P802.3ae Baseline proposal set and I
> included then for two purposes:
> 
> 1) Show how these 10 Gigabit Fibre Channel Ordered-Sets operate and
> interact with all 10GE Ordered-Sets. A 10GFC liaison request 
> is to only
> reserve 10GE code space for these Ordered-Sets but not include them in
> P802.3ae state diagrams.
> 
> 2) Show how these Ordered-Sets may be used to transport Link Status
> information per taborek_2_1100.pdf

I see. As a matter of fact it is interesting. Can you provide an
explanation/pointer for FC operation?

>     
> > 2.Whats the motivation behind SIGNAL event: Is this the RF? 
> If so, it is not
> > coordinated with your presentation in Tampa, where you show 
> the RF comes
> > always after the /A/ column
> 
> See above for SIGNAL history. The protocol we came up with in 
> Tampa for
> Link Status is neither SIGNAL nor SEQUENCE. I've coined a name for the
> protocol: PULSE, to differentiate it from SIGNAL and 
> SEQUENCE. Note that
> virtually all the Link Status work occurred in Tampa whereas D1.1 was
> available about a week prior to Tampa. This should account for any
> discrepancies. 
>  

I see. So: Any decisions made in Tampa regarding the LSS?

Thx.,
Boaz

> -- 
> 
> Best Regards,
> Rich
>    
> > 
> > Figure 48-9: Transmit Sequence
> > 
> > 1.If I understand whats written there-You send a sequence 
> until A_CNT_0=1,
> > then you send /A/, then you send pseudo random /K/ /R/ pattern until
> > A_CNT_0=1 again, and then you transmit the sequence again. 
> What is the
> > motivation?
> > 
> > Thx.,
> > Boaz
> 
> -- 
> 
> Best Regards,
> Rich
>                                       
> ------------------------------------------------------- 
> Richard Taborek Sr.                 Phone: 408-845-6102       
> Chief Technology Officer             Cell: 408-832-3957
> nSerial Corporation                   Fax: 408-845-6114
> 2500-5 Augustine Dr.        mailto:rtaborek@xxxxxxxxxxx
> Santa Clara, CA 95054            http://www.nSerial.com
> 
>