Reminder: Serial Attached SCSI PHY Study Group Conference call Monday, June24 2:00 pm - 4:00 pm CDT

Mike Jenkins jenkins at lsil.com
Thu Jun 27 19:38:48 PDT 2002


* From the T10 Reflector (t10 at t10.org), posted by:
* Mike Jenkins <jenkins at lsil.com>
*
Paul,

Thanks for the feedback.  I'll insert my responses below.

Regards,
Mike

> 
> Dear Mike,
> 
> This proposal includes not a few loose ends that will jeopardize any chance
> of the scheme operating properly.
> 1) It does not address how good the 50 Ohm resistors are.
> 2) It does not address how good the 10 nF capacitors are.
> 3) It does not address the impedance or bandwidth of the probe.

	This is the method prescribed by IEEE 802.3ae, Clause 47 
	(XAUI), so it's widely used.... i.e., more than a 'scheme'.
	The 50 ohm resistors are typically the inputs to a 'scope 
	or other lab equipment.  I think that will be good enough.  
	For the same reason, there are no probes involved.  (It's 
	straight into 50-ohm lab equipment inputs.)

	The blocking caps would either be part of the TX output 
	under test or part of the "compliance interconnect".  
	In the latter case, their losses and parasitics would 
	be part of the S21, which must be below the prescribed 
	limit.

> 4) Perhaps items 1) through 3) could be absorbed in a combination of
> transmission and return loss specifications.
> 5) Perhaps easier than a probe would be to implement the probe as an
> electrical splitter with return loss and transmission characteristics. The poorer
> the input characteristics of a piece of test equipment, the more effort would
> be required to create the wide band impedance and transmission matching
> networks, or the greater the attenuation to the test equipment port.

	We may be envisioning different things.  Rather than 'in situ' 
	testing, this would be a test on a transmitter not connected 
	to a receiver.

> 6) Even if you get the return loss specification, you must allow for the
> possibility that varying lengths of line in a real system will "rotate" the complex
> impedance in every possible direction. To check this, one might want to try
> using various electrical lengths all with maximum attenuation.
> 7) Lastly, even if the system works all around the circle, what makes you sure
> that it will work in the center? Recall that a classic fix for balky Fibre Channel
> connections was to use a longer cable.
> 8) Wouldn't it be easier to insist on a minimum return loss for the transmitter
> itself?

	I admit that all the theoretical questions have not been 
	fully addressed, although they have been debated at length 
	by the folks who developed XAUI.  Nonetheless, the compliance 
	interconnect method enables future extension of this approach 
	with the extensively developed s-parameter theory.  One 
	such extension which has been discussed is adding the 
	effects of crosstalk, turning the 2-port into a 4-port.

	In any event, I firmly believe that this approach does 
	far more to guarantee link operation than the traditional 
	"near-end eye mask and jitter" approach.  From a practical 
	viewpoint, the greater high-frequency energy in the near-end 
	waveform at these higher datarates, reduces near-end 
	measurements to not much more than an exercise in measuring 
	one's fixture.  And since these higher speeds often employ 
	TX emphasis, near end measurements based on decreasing-
	down-the-line jitter budgets are theoretically unjustified.  
> 
> Regards,
> 
> Paul
> __________________
> 
> Mike Jenkins wrote:
> 
> > * From the T10 Reflector (t10 at t10.org), posted by:
> > * Mike Jenkins <jenkins at lsil.com>
> > *
> > Alvin and all,
> >
> > Regarding the AI #6 assigned to me at the June SAS meeting in
> > Minneapolis (propose a compliance interconnect model for SAS),
> > an initial proposal has been uploaded as:
> >
> >   ftp://ftp.t10.org/t10/document.02/02-240r0.pdf
> >
> > Regards,
> > Mike
> >
> > ...
> > > 6.    Jenkins to develop test load model proposal: Enhancement to define a
> > >       compliant interconnect by which transmitter compliance can be tested
> > >       to assure receiver parameters. Status?
> > ...
> > --
> > ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
> >  Mike Jenkins               Phone: 408.433.7901            _____
> >  LSI Logic Corp, ms/AH260     Fax: 408.954.4764        LSI|LOGIC| (R)
> >  1525 McCarthy Blvd.       mailto:Jenkins at LSIL.com        |     |
> >  Milpitas, CA  95035         http://www.lsilogic.com      |_____|
> > ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
> > *
> > * For T10 Reflector information, send a message with
> > * 'info t10' (no quotes) in the message body to majordomo at t10.org

-- 
~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
 Mike Jenkins               Phone: 408.433.7901            _____     
 LSI Logic Corp, ms/AH260     Fax: 408.954.4764        LSI|LOGIC| (R)   
 1525 McCarthy Blvd.       mailto:Jenkins at LSIL.com        |     |     
 Milpitas, CA  95035         http://www.lsilogic.com      |_____|    
~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
*
* For T10 Reflector information, send a message with
* 'info t10' (no quotes) in the message body to majordomo at t10.org




More information about the T10 mailing list