The parallel interface to Gigabit Ethernet SERDES are 8b/10b encoded
for all of the ones that I've seen. With an 8 ns cycle time and
unidirectional traffic they're not the most demanding application but
they are common and should give you a convenient platform for evaluation.
The conversion delay AFAIK is on the order of two to four gate delays.
In reasonable technologies (e.g. 250 nm) this should be a couple of hundred
picoseconds at most and should probably be ahead of the final stage
latches. For most applications requiring interchip traffic the latency
hit should be acceptable.
> ----- Original Message -----
> From: D. C. Sessions <email@example.com>
> To: SI-List <firstname.lastname@example.org>
> Sent: Tuesday, September 21, 1999 4:23 PM
> Subject: [SI-LIST] : 8b/10b
> > Mike Degerstrom wrote:
> > > On Sep 20, 2:24pm, D. C. Sessions wrote:
> > > > Subject: Re: [SI-LIST] : RE: Another decoupling question
> > > > Mike Degerstrom wrote:
> > > > > On Sep 20, 10:39am, D. C. Sessions wrote:
> > > > > > IMHO the best bet for minimizing SSO transients is to use balanced
> > > > > > such as 8b/10b so that there isn't any substantial common-mode
> > > > > > Actually saves pins and cuts jitter too.
> > > > >
> > > > > Not sure where you are coming from here. 8b/10b will work if you
> > > > > have a DC block in your system. But for SSO transients you've got
> > > > > to get current supplied during the switching event - not just
> > > > > out the current demand over 10 bits.
> > > >
> > > > Parallel 8b/10b. Explained in another message.
> > >
> > > OK, I just read in more detail your explanation of parallel 8b/10b.
> > > First off, what is your worst case switch imbalance?
> > There are 254 codes with 5 high and 5 low. Add one each with 6/4 and
> > the worst case imbalanced transition goes from 6 high/4 low to 4 high
> > and 6 low (or vice versa). Either way the unbalanced currents go out
> > not only over the supplies but over the other signal lines (so the net
> > supply inductance is pretty low.)
> > > Secondly, you
> > > seem to be more concerned about balancing current going into
> > > the off-chip (on-board) termination supply.
> > That's just one consideration. If the off-chip termination supply doesn't
> > have to handle large unbalanced currents it takes quite a bit of cost out
> > of the system. It also does a lot towards reducing EMI.
> > > I would think that
> > > noise on the termination supply is much less of a concern than
> > > noise on the chip power or ground supplies (SSO).
> > For performance, yes. For EMI, probably. For cost, maybe.
> > Fortunately all three are in the same direction so the comparison
> > is somewhat academic.
> > > > > Can the confusion be that you have a single rail output driver, such
> > > > > as GTL, in mind whereas many of us are thinking more of a two-rail
> > > > > output driver such as LVTTL or full-swing CMOS?
> > > >
> > > > Wash your mouth!
> > > >
> > > > Actually, I *have* done some really fast current-mode stuff that had
> > > > to be open-drain because we couldn't manage the crossover distortion
> > > > across process. I'd rather everyone forgot it, though, and GTL is
> > > > one sin I've never committed. Push-pull is *much* more fun. (e.g.,
> > > > SSTL, HSTL, GLVDS, etc.)
> > >
> > > Not having designed with SSTL and HSTL, I'm guessing that both
> > > these I/O are terminated with 50 ohms to some VTT supply.
> > Yuppers. You can get the details at
> > http://www.jedec.org/download/freestd/jesd8-xx/
> > > Their
> > > HIGH and LOW state currents must be high enough so that when
> > > one pin switches from HIGH to LOW and another switches from
> > > LOW to HIGH, then dI/dt current into both ground and power
> > > supplies is approximately zero?
> > Two and oh.
> > > If so, then parallel 8b/10b
> > > can be attractive if the worst case SSO event is still fairly
> > > well balanced.
> > Indeed.
> > > We are looking into the future and realize that full swing
> > > CMOS won't cut it. My thoughts were that GTL was a good option.
> > > So your comments on GTL are timely. Four questions:
> > >
> > > 1) Do you envision push-pull to be more attractive as you
> > > can split dI/dt noise between power and ground rails (whereas
> > > with GTL it all gets dumped into ground).
> > That's one reason.
> > > 2) Can you explain what you meant by "crossover distortion?"
> > A push-pull rising edge has to coordinate the turnoff of the
> > pulldown with the turnon of the pullup. In the analog camp
> > the unavoidable noice accompanying the transition from sinking
> > to sourcing current is called "crossover distortion" and for
> > want of a better term I use it to describe the jitter that
> > can be introduced in digital signaling under the same conditions.
> > > 3) Why were you so negative on GTL?
> > GTL is slow, noisy, and burns lots of power. If that's what
> > you need, you're in luck.
> > Slow because an open-drain signaling system is extremely sensitive
> > to line impedance mismatch. Noisy because the currents are unbalanced
> > and and tend to form large loops. Power-hungry because the open-drain
> > signaling requires that common-mode potential be maintained by current
> > in the terminators.
> > > 4) What is GLVDS?
> > A (currently proprietary) low-swing signaling method that Ericsson uses.
> > Stands for Ground-referenced Low-Voltage Differential Signaling.
> > --
> > D. C. Sessions
> > email@example.com
> > **** To unsubscribe from si-list: send e-mail to
> firstname.lastname@example.org. In the BODY of message put: UNSUBSCRIBE
> si-list, for more help, put HELP. si-list archives are accessible at
> http://www.qsl.net/wb6tpu/si-list ****
-- D. C. Sessions email@example.com
**** To unsubscribe from si-list: send e-mail to firstname.lastname@example.org. In the BODY of message put: UNSUBSCRIBE si-list, for more help, put HELP. si-list archives are accessible at http://www.qsl.net/wb6tpu/si-list ****