BIRD34 - Stored Charge Effects

To IBIS Members:

One of my action items was to recast EGG9 into BIRD34 to deal with the
topic of handling Stored Charge effects in Diodes.

Because of stored charge, a clamping diode will not release instanteously
from the clamped mode as predicted by IBIS models. There will be a delay
while the diode is turning off. This creates a glitch which itself can
propogate up and down a line. This glitch is significant enough that IBIS
needs to include a way to model it.

The major contributor to the effect is the TT parameter which produces
a non-linear transit time capacitance when a diode is conducting. The
proposal here is to add effective parameters defined by [TTgnd] and [TTpower]
keywords to the [Model] keyword description.

Bob Ross,
Interconnectix, Inc.

******************************************************************************
******************************************************************************

BIRD ID#: 34
ISSUE TITLE: Stored Charge Effects
REQUESTER: Bob Ross, Interconnectix, Inc.
DATE SUBMITTED: 3/5/96
DATE ACCEPTED BY IBIS OPEN FORUM: Pending

******************************************************************************
******************************************************************************

STATEMENT OF THE ISSUE:

The effect of transit time capacitances is not currently handled in the
IBIS format, yet its effects are important for certain simulations.

******************************************************************************

STATEMENT OF THE RESOLVED SPECIFICATIONS:

Two new keywords [TTgnd] and [TTpower] provide effective transit time
parameters for modeling the ground clamp and the power clamp transit time
capacitances independently. Approximation equations are included in the
description. The additional keywords are described after the [Model]
keyword as follows.

|==============================================================================
| Keywords: [TTgnd], [TTpower]
| Required: No
| Description: The data for these keywords enters the transist time parameters
| to estimate the transit time capacitances for the [Gnd Clamp]
| and [Power Clamp] tables.
| Usage Rules: For each of these keywords, the three columns hold the transit
| values corresponding to the typical, minimum and maximum
| [Gnd Clamp] or [Power Clamp] tables, respectively. The
| entries for TT(typ), TT(min), and TT(max) must be placed on
| a single line and must be separated by at least one white
| space or tab or tab character. All three columns are
| required under these keywords. However, data is required
| only in the typical column. If minimum and/or maximum values
| are not available, the reserved word "NA" must be used
| indicating the TT(typ) value by default.
| Other Notes: The transit time capacitance is added to C_comp. It is
| in a Spice reference model as Ct = TT * d(Id)/d(Vd) where
| d(Id)/d(Vd) defines the DC conductance at the incremental
| DC operating point of the diode, and TT is the transit time.
| This expression does not include any series resistance
| component. Assume that the internal diode current (Id) -
| voltage (Vd) relationship is Id = Is * (exp(q(Vd)/kT) - 1)
| where Is is the saturation current, q is electron charge, k is
| Boltzmann's constant, and T is temperature in degrees Kelvin.
| Then d(Id)/d(Vd) is approximately (q/kT) * Id when the diode
| is conducting, and zero otherwise. This yields the simplifi-
| cation Ct = TT * (q/kT) * Id. The Id is found from the
| [Gnd Clamp] and [Power Clamp] operating points, and the cor-
| responding TTgnd or TTpower is used to calculate the Ct value.
| If the [Temperature] keyword is not defined, then use the
| default "typ" temperature for all Ct calculations.
|
| The effective TT parameter values are intended to APPROXIMATE
| the effects. They may be different from the values found in
| the Spice diode equations. Refer to the NOTES ON DATA
| DERIVATION METHOD for extracting the effective values.
|------------------------------------------------------------------------------
| variable TT(typ) TT(min) TT(max)
[TTgnd] 10n 12n 9n
[TTpower] 12n NA NA
|
|==============================================================================

Add Section (4) to NOTES ON DATA DERIVATION METHOD:

| 4) Transit Time Extractions:
| The transit time parameter is indirectly derived to be the value that
| produces the same effect as that extracted by the reference measurement
| or reference simulation.
|
| The test circuit consists of the following:
| a) A pulse source (10 ohms, 1 ns at full duration ramp) or equivalent
| and transitioning over the full expected input transition, such as
| between 0 V and Vcc for CMOS technology, and between 0 V and 3.3 V
| for TTL and Low voltage technologies,
| b) A 50 ohm, 1 ns long trace or transmission line,
| c) A 500 ohm termination to the ground clamp reference voltage for TTgnd
| extraction and to the power clamp reference voltage for TTpower
| extraction (to provide a convenient, minimum loading 450 ohm -
| 50 ohm divider for high-speed sampling equipment observation
| of the device under test), and
| d) The device under test.
|
| The TTgnd extraction will be done only if a [Gnd Clamp] table exists.
| A high to low transition that produces a positive "glitch", perhaps
| several nanoseconds later indicates a stored charge in the ground
| clamp circuit. The test circuit is simulated using the complete
| IBIS model with C_comp and the Ct model defined under the [TTgnd] and
| [TTpower] keywords. An effective TTgnd value that produces a "glitch"
| with the same delay is extracted.
|
| Similarily, the TTpower extraction will be done only if a [Power Clamp]
| table exists. A low to high transition that produces a negative
| "glitch", perhaps several nanoseconds later indicates a stored charge
| in the power clamp circuit. An effective TTpower value that produces a
| glitch with the same delay is extracted.
|
| It is preferred to do the extractions with the package parameters
| removed. However, if the extraction is done from measurements, then
| the package model should be included in the IBIS based simulation.

******************************************************************************

ANALYSIS PATH/DATA THAT LED TO SPECIFICATION:

Refer to Spice diode reference information concerning the complete equations.
The emission coefficient (N) is assumed to be 1. For measured values or
cases where N is not 1, use the effective TT values. So the TT values
may not be the exact values used in a Spice model.

The Spice diode and transistor models differs slightly from the IBIS Clamp
models. The definition and position of the capacitances is different.
Furthermore, the IBIS table combines the diode and resistor, but can support
more complex non-linear characteristics.
                          
                   Intrinsic
             RS Diode +-----+
                      |\ | ----> Id | | ----> Id
       o---/\/\/\--+--| >|--+----o +-----|Clamp|-----+
                   | |/ | | | | | |
                   | | o----| +-----+ |----o
                   | | | | | | | |
                   +---| |--+ +-------| |-------+
                       | | | |

                     Cj + Ct C_comp + Ct

Fixed values of C_comp serve to estimate voltage dependent non-linear
junction capacitance plus other metalization effects. Note, while shown
here across the diode, the C_comp and Ct term would be lumped as capacitances
between the terminal and ground to provide an equivalent AC circuit. Methods
to connect capacitance contributions to each of the voltage rails are not
available in the IBIS model.

Note, this example illustrates one diode. It can be connected to the
ground reference voltage (usually ground) or to the power reference voltage
(usually Vcc) serving as either a [Gnd Clamp] or a [Power Clamp]. In
practice, both clamps can exist and C_comp would contain the effects of
both diodes and metalization.

The Ct value is a function of absolute temperature. As an approximation,
the default typical temperature is sufficient if [Temperature] is not
specified. It is recommended to include the [Temperature] keyword when
TT is specified to remove ambiguity regarding minimum and maximum transit
time calculations for CMOS versus Bipolar technologies.

An underlying assumption is that the equations will be applied only to
the Clamping data, not the combined data that includes [Pulldown] and
[Pullup] data. So when this detail is needed for Output models, the
clamping data needs to be derived, if not provided.

Because of these differences and possible missing details, the simplified
equation and approximation approach is justified to capture the dominant
behavioral effects. So the TT values may be the effective values that are
consistent with the IBIS model and data.

There are several formatting choices to describe transit time:

(1) A [Model] subparameter similar to Vinl and Vinh, e.g.,

TTgnd = 12n
TTpower = 10n

This was not chosen because minimum and maximum process Spice models may have
different values.

(2) A [Model] subparameter similar to C_comp, e.g.,

TTgnd 10n 12n 9n
TTpower 12n NA NA

This was not chosen because the rules for minimum and maximum columns for
C_comp are based on magnitude, whereas the rules for TT are based on process
extremes. This may be confusing if they are similarly formatted. However,
this is the second choice since TTgnd, TTpower and C_comp all relate to
capacitances.

(3) A [Gnd Clamp] and [Power Clamp] subparameter similar to C_comp, e.g

[Gnd Clamp]
TT 10n 12n 9n

This was not chosen because it can cause unnecessary [Gnd Clamp] and [Power
Clamp] table complexity and it can make the TT parameters hard to locate.

(4) A [Model] keyword modifier similar to [Temperature] and other keywords
which are a function of typical, minimum and maximum process and measurement
conditions. e.g.,

[TTgnd] 10n 12n 9n
[TTpower] 12n NA NA

Since these keywords are related to the [Gnd Clamp] and [Power Clamp]
tables and to the conditions under which they were derived, this format was
chosen.

BIRD34 also includes in its description a proposed extraction method. The
method is based on using the typical input signal transitions. It is also
designed so that it can be done under actual measurement conditions since the
"real" effect may not be accurately modeled in Spice.

An alternative which does not closely reflect actual conditions would be
to measure the delay forward biased (perhaps by one volt) clamp diode takes
to turn off when the bias voltage is removed or reverse biased. This approach
has not been fully simulated. However, it is expected to be very sensitive
to the selected bias voltages and not necessarily mimic actual operating
conditions.

******************************************************************************

ANY OTHER BACKGROUND INFORMATION:

BIRD34 expands EGG9 and adds more detail on modifying IBIS. It also expands
on the mathematical basis as requested by Kellee Crisafulli. It also adds
some comments based on EGG9 response from Stephen Peters and others on the
positioning of C_comp in the ANALYSIS PATH section.

******************************************************************************
Received on Mon Mar 4 10:45:05 1996