Transmitter specification at 6G

Allen.N.Kramer at seagate.com Allen.N.Kramer at seagate.com
Tue May 22 14:51:01 PDT 2007


Formatted message: <A HREF="r0705223_f.htm">HTML-formatted message</A>
Attachment #1: <A HREF="r0705223_nameless-3988-3.gif">nameless-3988-3.gif</A>

Once upon a time in an FC Standard long long ago we used to specify TX 
Jitter as DJ, RJ, and Total jitter.
And they all added up quite neatly until one day someone made the 
observation that we where over-limiting the jitter. He observed that if 
one did better on one component of jitter (DJ or RJ), shouldn't he be 
allowed exceed the other parameter as long as he didn't exceed Total 
jitter. At the time, we had the choice of specifying "DJ and Total" or "RJ 
and Total". Which ever was chosen, the opposite jitter attribute would be 
allowed to exceed the limit then applied by specifying the three 
components of jitter: DJ, RJ, and Total jitter.   If "DJ and Total" were 
chosen, receivers would have to tolerate RJ up to near the Total jitter 
number if someone designed a near perfect DJ transmitter. If "RJ and 
Total" were chosen, a receiver would have to tolerate DJ up to near the 
Total jitter number if someone came up with a transmitter with exceedingly 
low RJ.  After a lot of discussion, "DJ and Total" was chosen because the 
consensus at that time was that a receiver could tolerate a little more RJ 
in the presence of well controlled DJ, but that well controlled RJ did not 
guarantee that a receiver could tolerate DJ that was a larger proportion 
of Total jitter. 
The current version of 07-063 uses an "RJ and Total" jitter specification 
at 6Gbp/s.  While receivers have improved dramatically, I still do not 
think that they have improved to the point that the transmit device's DJ 
should not be specifically limited by a "DJ and Total" methodology. If 
receivers can provably tolerate more DJ at the transmitter than is 
currently specified, then we should increase it.  However, I think it 
should still be intentionally limited within TJ, leaving RJ as the jitter 
component that can be increased in the presence of better DJ.  After all, 
we are going to be faced with closed, or near closed eyes and the primary 
component of this 'closing' will be ISI, which is DJ. The ISI basically 
shrinks the amplitude of high transition bits that follow a low transition 
region.  DJ from other sources, such as the transmitter, basically time 
skew these same low amplitude high transition bits from the center of the 
eye, further smearing the eye.	We need not only limit the eye closure 
allowed by ISI, we need also to intentionally limit the displacement of 
these amplitude challenged bits from the center of the eye.  The
‘channel’ 
that interconnects the transmitter to the receiver creates additive DJ to 
the transmit signal from such sources as ISI (mentioned above), imperfect 
terminations, impedance discontinuities at connectors, via’s, packages, 
crosstalk, etc.   These DJ sources also skew the low amplitude high 
transition bits from the center of the eye.  The effects of the majority 
of these sources of DJ on a signal are not easily quantified, and the 
universe of creatively intermixing these sources of DJ onto a signal in a 
’channel’ is certainly not contained in a single S-parameter file of an 
external 10meter SAS cable.   To then allow a transmitter without ‘clearly 
defined DJ limits’ to drive such a channel is ill advised.  The hope of 
developing a 6G standard that specifies all parts of the TX-RX link in a 
manner that results in achieving the near-zero bit error rate desired by 
storage system developers necessitates keeping a "DJ" jitter specification 
on the transmitter. 
It may even require retuning to the method of long ago: intentionally 
limiting “RJ” and “DJ”, since I assume we did not change from a “DJ
and 
Total” jitter specification (that has been used to successfully develop 
several generations of near-zero bit error rate storage systems) to an “RJ 
and Total” specification unless this change addressed a known problem. 
Recovering signals that are ‘closed’ is after all, a new endeavor in 
near-zero bit error rate storage systems, and may require specifying both 
RJ and DJ along with Total Jitter (i.e. the limit on RJ and DJ do not 
necessarily have to add up to Total Jitter; the three attributes {RJ, DJ, 
Total} just need to constrain the measured transmitter performance in a 
manner that increases the opportunity for building a new generation of 
near-zero bit error rate storage systems).
Regards,
Al
************************************
 Allen Kramer 
   Phone: 952-402-2624 
    Fax: 952-402-3471 
     Seagate Confidential
************************************



More information about the T10 mailing list