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

Re: Clause 33 (MDC/MDIO) questions




Isn't what you what to say....

"Any write or read shall be preceeded by an address frame that initializes
the MDIO device to the proper address."

In other words, there is only a requirement that the address be
set-up PRIOR to doing a read or write. There's no requirement that
they be close together in "time". There's also no requirement that
address frames can't be sent zillions of times without reading or 
writing.

An example. Say that two registers on the part contain bits of interest
and that reading these bits quickly is important. One might have some
code that "prepared" to read register one (by setting the address) and then
waited for some critical moment to actually do the read. Some other 
event might cause the same thing to occur on register two, PRIOR to the
read of register one....and so on. This may be a silly example, only reflecting
foolish coding practices, but the standard should not dictate more
than the minimum requirements to allow interoperability. 

As long as the address is set-up prior to doing the read or write, everything
should work fine. We should not imply requirements that don't exist.
We especially don't need to force the inclusion of 555 timers or extra create
error sequences (e.g. address set-up without usage error).

Making the "non-requirement" a recomendation does not help.
It is an absolute requirement that the address frame PRECEED the read/write (perhaps
modified by subsequent operations that also increment). It is absolutely
not a requirement that any operation be FOLLOWED by another.

RR


----- Original Message ----- 
From: "Edward Turner" <Edward_Turner@xxxxxxxxxxxx>
To: <stds-802-3-hssg@xxxxxxxx>
Sent: Friday, October 20, 2000 3:35 AM
Subject: Re: Clause 33 (MDC/MDIO) questions


> 
> 
> 
> Louis,
> 
> I'm going to remove the word 'immediately' from the next draft.
> You can interpret the word 'should' as 'is recommended that', so the sentence
> can also be read as 'It is recommended that the address frame is followed by its
> associated write, ...'.
> There is no IFG defined for the MDIO interface.
> Sending address frames back to back will not break anything.  Each time an
> address frame is received it overwrites the MDIO device's address register.  The
> address used for a transaction will be whatever is held in this address
> register.
> 
> Regards
> Ed
> 
> 
> 
> 
> 
> Louis Lin <louislin@xxxxxxxx> on 19/10/2000 23:39:54
> 
> Sent by:  Louis Lin <louislin@xxxxxxxx>
> 
> 
> To:   stds-802-3-hssg@xxxxxxxx
> cc:    (Edward Turner/GB/3Com)
> Subject:  Clause 33 (MDC/MDIO) questions
> 
> 
> 
> 
> 
> Hi,
> 
> On page 84. second section after the table 33-21.
> 'An address frame should be followed "IMMEDIATELY" by its
> associated write, ..........'
> 
> Can you define "immediately"? is there a max IFG ?
> Or it's just saying the PHY device is expecting a data frame after
> an address frame?
> 
> Will it break anything if we send two address frames without any
> data frame in between?
> I expect that the second register address will overwrite the first
> register address in the PHY's local address table without breaking
> anything.
> 
> Thanks,
> 
> Louis Lin
> 
> 
> 
> 
> 
> 
>