Here are my comments/thoughts:
 
Training Complete is generated by the physical layer receiver.  How, why, when is implementation specific and probably will not be defined by the PHY working group (although my comment here is not binding on the WG).
 
Training Completed is not a function of any received dword or primitive, so it does not _require_ dword synchronization to occur.
 
06-515r0 (this is the latest revision) requires SP_DWS Sync Acquired and Training Completed to move forward in the state machine.  The order of these two events is not specified.
 
Regards,
 
Steve

From: owner-t10@t10.org [mailto:owner-t10@t10.org] On Behalf Of Craig Stoops
Sent: Wednesday, January 24, 2007 9:17 AM
To: t10@t10.org
Subject: RE: SAS2 SNW-3 SP29:SAS_Train

Hi Rob and Steve,
 
I see the receiver is supposed to send this message, but I do not see any definition of what the conditions are for generating this message.
 
It also doesn't say that the message is only sent after DWS has obtains synchronization as it does with ALIGN Receieved and TRAIN_DONE Received.
 
It would seem to me, that Training Completed is the same as SP_DWS Sync Aquired. The point being to lock the rx pll, then the dws will sync.
 
I am refering to 06-515R0 as I don't see a revision posted on t10.
 
Can you point me the way to which section has the definition of Training Completed? I particularly want to know if this message generation requires a certain number of patterns and of what type to be seen to qualify it.
 
Thanks,
Craig
ExpertIO, Inc
www.expertio.com
805-428-0839
 
-----Original Message-----
From: owner-t10@t10.org [mailto:owner-t10@t10.org] On Behalf Of Elliott, Robert (Server Storage)
Sent: Wednesday, January 24, 2007 6:53 AM
To: t10@t10.org
Subject: RE: SAS2 SNW-3 SP29:SAS_Train

That state is waiting for the SP receiver to report Training Completed (meaning the receiver itself is satisfied with the incoming data stream), not that it has detected any particular incoming TRAIN or TRAIN_DONE (indicating the state of the other phy).
 
Training Completed is in the list of messages defined earlier in the proposal, but is never used again; it was supposed to have been used here.
 
That correction will be incorporated into sas2r08.
 

From: owner-t10@t10.org [mailto:owner-t10@t10.org] On Behalf Of Craig Stoops
Sent: Tuesday, January 23, 2007 4:24 PM
To: t10@t10.org
Subject: SAS2 SNW-3 SP29:SAS_Train

Steve, Rob and other interested parties,
 
Per the latest 6Gb SNW-3 proposal pasted below, the TRAIN Completed message is not defined in the document. Based on my read of other parts, I think this should be "TRAIN Transmitted message" as the requirement per 6.7.4.2.3.3 is to send at least 1 TRAIN pattern and after achieving receiver sync, then transition to sending TRAIN_DONE.

6.8.4.12.4 Transition SP29:SAS_Train to SP30:SAS_TrainingDone

This transition shall occur if:

a) this state receives a TRAIN Completed message before the TLT timer expires; and

b) dword synchronization is acquired.

Second point - becuase one end device may be capable of acheiving rx sync faster than the other, it is possible (as seen by our modeling) for the slower device to be in SP29 and before that device gains sync the faster device starts sending TRAIN_DONE.

So it can not be a condition of SP29 -> SP30 to receive a TRAIN pattern, but rather either TRAIN or a TRAIN_DONE pattern This is implied by aquiring receiver synchronization.. Currently, as it is undefined, some people may consider the TRAIN Completed message to be TRAIN Received messsage.

I think it should be further clarified, as I beleive was the original intention, that both TRAIN and TRAIN_DONE serve to allow a receiver to gain synchronization. If true, then the slow device will transition from SP29 to SP30 when gaining rx sync even though it is receiving TRAIN_DONE and not TRAIN.
 
 
 
Basically, in summary, rewording 6.8.4.12.4 to:
  This transition shall occur after: 
    a) Transmitting at least 1 complete TRAIN pattern (or said the spec way, after having received at least one TRAIN Transmitted message)
   b) Receiver DWORD synchronization is acquired
 
 
Thoughts?
 
Craig Stoops
ExpertIO, Inc
www.expertio.com
805-428-0839