SAS-2 Spread spectrum clock -

Elliott, Robert (Server Storage) Elliott at hp.com
Tue Jul 4 13:04:22 PDT 2006


Attachment #1: <A HREF="r0607042_smime.p7s">smime.p7s</A>

> -----Original Message-----
> From: owner-t10 at t10.org [mailto:owner-t10 at t10.org] On Behalf 
> Of Massimo POZZONI
> Sent: Tuesday, July 04, 2006 2:13 AM
> To: t10 at t10.org
> Subject: SAS-2 Spread spectrum clock - 
> 
> * From the T10 Reflector (t10 at t10.org), posted by:
> * Massimo POZZONI <massimo.pozzoni at st.com>
> *
> The 06-263r2 (SAS-2 Spread spectrum clock) specifies that:
> 
>   - the device trasmitter SSC is 0/-5100 ppm (both SAS and SAS)
>  
>   On the other hand,
>   - expander transmitter is	 0/-5100 ppm if talking to a SATA device
>   - expander transmitter is + 2500/-2500 ppm if talking to a 
> SAS device
> 
>   Let's consider the receiver Clock Recovery (CDR) of the device.
>   Let's assume that the receiver Clock Recovery makes use of the same
>   clock reference used by its transmitter (to save power and area 
>   in the PHY).
> 
>   As a consequence, this clock reference, used by the receiver CDR 
>   to recover the incoming data, will be SSC modulated as the 
>   transmitter.
> 
>   We see now that the receiver CDR is much more stressed than in SATA.
>   In fact its reference clock can be at -5000 ppm, while the data
>   it is receiving can be at +2500 ppm. Total 7500 ppm.
> 
>   My concern is for the Clock Recovery. It has been proven that
>   many problems are already existing in sata devices, where the
>   clock recovery is stressed up to 5000 ppm.
What problems?	I haven't seen any errata to the SATA specification
trying to reduce the ppm range.
There are many accounts on the Internet about problems arising from 
HBAs not supporting SSC as required.  Disk drive vendors that tried 
turning on SSC have had to provide consumers special programs or 
jumpers to turn SSC off so they will work in non-compliant systems.
http://www.intel.com/support/motherboards/server/sb/CS-021230.htm
http://www.xlr8yourmac.com/archives/nov05/111105.html
http://discussions.apple.com/message.jspa?messageID=2379652
http://www.barefeats.com/hard49.html
http://www.seagate.com/support/kb/disc/faq/ssc.html
http://www.wdc.com/en/library/eide/2579-001037.pdf
http://www.hitachigst.com/hdd/support/download.htm
>   Now we are proposing to stress it at 7500 ppm !!!.
> 
>   My proposal:
>   If an expander is transmitting to a SATA device: TX @ 0/-5000
>   If an expander is transmitting to a SAS  device: TX @ +2500/-2500
>   (This is not changed)
> 
>   If a SAS device is transmitting to a SAS expander: TX @ +2500/-2500
>   (This is the new !!)
>   If a SATA device is transmitting to a SAS expander: TX @ 0/-5000 ppm
SAS-1.1 expanders/HBAs are supposed to support receiving +0/-5000 ppm
(I hope they were more compliant than SATA HBAs), but were not
expected to tolerate +2500/-2500 ppm.  SAS disk drive vendors would
prefer to implement just one behavior regardless of the expander/HBA 
to which they are attached, so transmitting +0/-5000 ppm is the
only safe choice.
HBAs would like to have as few choices (e.g. clock trees) as possible.
They must use +0/-0 when attached to a SAS-1.1 phy, and can only 
use +0/-5000 ppm when attached to a SATA phy, so +2500/-2500 ppm
would require a third clock tree that is not really necessary.	
Nothing should break if they did transmit +2500/-2500 ppm to a
SAS-2 phy, since all SAS-2 receivers will be required to 
tolerate it.
>   If the SAS expander cannot tolerate the +2500/-2500 from the
>    SAS device, the SSC is turned off in all the ports of the device.
Disk drive vendors would prefer not to include separate SSC controls 
per phy.  Since there are generally more disk drives than HBAs or 
expanders, it makes sense to keep them as simple as possible.
>    In summary: is it possible to have the same SSC range for both the
>    transmitter and the receiver of a SAS device?
I don't think so.
> 
> Massimo Pozzoni
--
Rob Elliott, elliott at hp.com
Hewlett-Packard Industry Standard Server Storage Advanced Technology
https://ecardfile.com/id/RobElliott



More information about the T10 mailing list