From: Eddie Suckow (email@example.com)
Date: Mon Feb 12 2001 - 07:57:25 PST
GTL modeling discussion:
A quick note to those interested in XTK GTL simulation. Fairchild Semiconductor has released its Ensigna Backplane Simulator allowing individuals to simulate numerous backplane configurations in a
matter of minutes.
Visit the link below for more information:
In follow-up to Abe Riazi's discussion on calculating backplane impedance, it is important to remember the effect of loading the backplane if it is of parallel configuration. With the insertion of each
card a new capacitance is introduced, thus resulting in a decrease in the effective impedance. As multiple cards are inserted it is not uncommon to see the characteristic impedance of the backplane drop
by 25%. Naturally, this calls for a lower termination resistance Rt to keep the impedance matched. The challenge then becomes finding the optimum Rt for both unloaded and loaded backplane
With VIL and VIH, a 50mV noise margin is extremely tight. I typically like to keep at least 150mV to ensure that a false trigger causing a glitch on the TTL output signal is not seen.
GTLP Applications Engineer
> si-list-digest Saturday, February 10 2001 Volume 01 : Number 415
> In this issue:
> 1.) Re: [SI-LIST] : GTL standard
> "ARiazi" <ARIAZI@prodigy.net>
> Date: Sat, 10 Feb 2001 15:50:19 -0800
> From: "ARiazi" <ARIAZI@prodigy.net>
> Subject: Re: [SI-LIST] : GTL standard
> Suchitha V Wrote:
> > First of all, Min, Typ, Max represent the SLOW, TYPICAL and FAST BUFFER
> > CORNERS respectively.
> > 2% implies that Vref can have a tolerance of +/- 2% on either side of the
> > typical value.
> > For the purpose of simulation, the SLOW, TYPICAL, and FAST buffer models
> > included in the IBIS models
> > which are derived from the SPICE simulations/measurements.
> > SLOW corner is a buffer in its weak performance i.e, slow speed, weak
> > strength, weak voltage.
> > In this case the variation of Vref= 2/3Vtt -2% is considered for deriving
> > the VI data for the IBIS models.
> > FAST corner is a buffer in its fullest performance i.e, fast speed, strong
> > driver strength, strong voltage.
> > In this case the variation of Vref=2/3Vtt +2% is considered for deriving
> > VI data for the IBIS models.
> > TYPICAL corner corner is a buffer with typical performance and therefore
> > typical values are to be
> > considered.
> Hi Suchitha:
> As you stated: "SLOW corner is a buffer in its weak performance i.e, slow
> speed, weak driver strength, weak voltage".
> Additionally, temperature and package impedance also play a role so that:
> Slow Corner Model: slowest (Min) ramp (dV/dt), Min VI curves, Min voltage,
> Max temperature, and Max package impedance.
> Fast Corner Model: fastest ramp, Max VI, Max voltage, Min temperature, and
> Min package impedance.
> Typical Corner Model: All parameters at nominal values.
> In many cases, just a single set (i.e. Min, Typ, Max) of models
> (representing each bus agent) proves sufficient for all signal quality and
> timing margin evaluations. However, situations can arise which demand
> multiple sets of models (of SAME component) for a comprehensive
> investigations of overshoot/undershoot, crosstalk, synchronous timing skew,
> etc. Furthermore, models often evolve through changes and revisions.
> Therefore, care must be taken to obtain the most complete and recent models
> for each simulation.
> The receiver input reference voltage ( Vref ) combined with a specified
> tolerance dictate the input threshold region; i.e. :
> VIL(max) = Vref - 50 mV
> VIH(min) = Vref + 50 mV
> The Input Low Voltage ( VIL ) and Input High Voltage ( VIH ) thresholds in
> turn influence the lower and upper noise margins of the bus.
> Furthermore, Vref in conjunction with a test load are used (by the processor
> manufacturer) to specify the signals' valid timing parameters.
> Consequently, for XTK simulations, the same test load should be utilized
> for TIME_TO_VM calibration of Quad driver models with the
> measurement voltage set at Vref.
> The termination voltage Vtt (equivalent to the processor core supply
> voltage) also varies over a range of values. The smallest value of Vtt
> be used in SS (i.e Slow buffer, Slow environment) and the largest Vtt value
> in FF (i.e Fast buffer, Fast PCB) simulations.
> Vtt, Vref, noise margins, and the models are among critical considerations
> when designing or simulating a high speed bus. Other essential bus elements
> include topology, termination plans, stackup, decoupling techniques,
> bus width, bandwidth, trace impedance, velocity, etc. The final paragraph
> presents comments and examples regarding impedance and propagation delay.
> The formulas for computing bus impedance include numerous parameters
> such as trace width, thickness, height over reference plane, substrate
> dielectric constant, etc. Therefore, to efficiently verify for instance
> 50 +/- 10% Ohms for GTL or 60 +/- 10% Ohms for PCI-X require
> use of a field solver program. The signal propagation delay
> Tpd (inverse of velocity) involves only the substrate dielectric constant
> and simple equations:
> Tpd = 1.017* SQRT(0.475Er + 0.67) nS/ft [For microstrip] and
> Tpd = 1.017* SQRT(Er) nS/ft [For stripline traces]
> For example, assuming FR-4 substrate and Er= 4.2, yields:
> Tpd = 1.66 nS/ft [for microstrip] and
> Tpd = 2.084 nS/ft [for stripline signals]
> When analyzing a high speed bus, it is frequently desirable to ascertain a
> range of Tpd values for the outer and inner traces. This can be easily
> achieved by first assigning appropriate values to Er for the best and the
> worst case corners and then applying the above Tpd equations.
> Best regards,
> Abe Riazi
> **** To unsubscribe from si-list or si-list-digest: send e-mail to
> firstname.lastname@example.org. In the BODY of message put: UNSUBSCRIBE
> si-list or UNSUBSCRIBE si-list-digest, for more help, put HELP.
> si-list archives are accessible at http://www.qsl.net/wb6tpu
> End of si-list-digest V1 #415
> **** To unsubscribe from si-list or si-list-digest: send e-mail to email@example.com. In the BODY of message put: UNSUBSCRIBE si-list or UNSUBSCRIBE si-list-digest, for more help, put HELP.
> si-list archives are accessible at http://www.qsl.net/wb6tpu
**** To unsubscribe from si-list or si-list-digest: send e-mail to
firstname.lastname@example.org. In the BODY of message put: UNSUBSCRIBE
si-list or UNSUBSCRIBE si-list-digest, for more help, put HELP.
si-list archives are accessible at http://www.qsl.net/wb6tpu
This archive was generated by hypermail 2b29 : Tue May 08 2001 - 14:30:50 PDT