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

RE: [802.3ae] Questions on section 49.2.5

My responses are bracketed by my initials. Pat

-----Original Message-----
From: Burt Christian [mailto:burt@xxxxxxxxx]
Sent: Monday, November 19, 2001 1:53 PM
To: HSSG Reflector (E-mail)
Subject: [802.3ae] Questions on section 49.2.5


I have the following questions about section 49.2.5

1. In the first paragraph, the ratio between the data-rates
   of XGMII and the PMA service interface is mentioned to be
   16:33. Is this correct or is 32:33, the correct ratio ?

<PAT begin>
16:33 is the correct ratio. Here is the calculation:
The XGMII sends 32 unencoded bits each cycle. 
The PMA sends 16 encoded bits each cycle.

Let Ud be the unencoded data rate;
Cxgmii be the tranfer rate of the XGMII; and
Cpma be the PMA service interface transfer rate.

Cxgmii = Ud/32 transfers/second

Cpma = Ud unencoded bits *  66 encoded bits   * 1 transfer
       -----------------   ------------------   ----------
              s            64 unencode bits     16 bits
     = Ud *33/512 transfers/second

Cxgmii    Ud/32          16
------ = ----------- = ------
Cpma     Ud *33/512      32

<PAT end>

2. With respect to the following sentence in the same paragraph:
    "Where the XGMII and PMA sublayer data rates are not
     synchronized to that ratio, the transmit process will
     need to insert idles, delete idles, or delete sequence
     ordered sets to adapt between the rates."

   The above rate adaptation is intended for adapting for the
   minute frequency offset that can be present if the XGMII
   transmit clock (TX_CLK) and the PMA_TXCLK are not frequency
   locked - correct ? What I am trying to clarify is that this
   rate-adaptation is completely independent of the rate
   adaptation required for interfacing to WIS.

<PAT begin>
The ability to avoid rate adaptation by synchronizing PMA clock
to the XGMII clock on transmit and vice versa on receive applies
only to the LAN PHY. Such synchronization only works for the LAN
PHY where the XGMII and PMA are running at the same nominal 
data rate. With the WAN PHY, they are running at different nominal
data rates and rate adaptation must be performed.

When the clocks have not been synchronized, the same rate 
adaptation mechanism (inserting idles and deleting idles or 
ordered sets) applies to adjusting for the small difference 
when due a PMA and XGMII running at nominal the same data 
rate as they do for the LAN PHY and to rate adaptation for 
the somewhat larger rate mismatch of the WAN PHY. 

In the case of the WAN Phy, one knows that one will be deleting 
idles at the transmitter and adding them at the receiver
(and that the MAC will be inserting extra idles as necessary). This
contrasts with the LAN PHY case where one doesn't know the direction
of the mismatch. One will also need a bit bigger elastic buffer for 
the WAN PHY case. Therefore, the constraints on implementation differ  
slightly for the two cases but the general mechanism is the same.
<PAT end>


Burt S, Christian.
Ceyba Inc.,
Systems Specialist
Tel : (613)599-5797
Fax : (613)599-1247