(IGNORE PREVIOUS EMAIL - HIT SEND BEFORE INSERTING THE FILE)
IBIS Folks:
BIRD49.3 is issued to reflect changes in BIRD48.4 associated with introducing
submodels:
- The Title is changed from Add Model Dynamic Clamps to Add Submodel Dynamic
Clamps
- The text is rewritten to reflect the changes in BIRD48.4.
- The Clocked Mode was controversial. We decided not to include it in
BIRD49.3 at the June 19, 1998 IBIS Summit meeting since we could work
around its problems using the Triggered mode. The primary concern is that
the real specification of the clocked mode requires IBIS to handle timing
information.
Thus the following modes still exist depending upon whether there exist
optional [Submodel Spec] and the [GND Pulse Table] and [POWER Pulse Table]
entries:
[Submodel Spec] Pulse Table
Yes Yes Triggered
Yes or No No Static
- A complete example is provided.
- Some confusing, historical discussion in the ANALYSIS PATH/DATA THAT LED TO
SPECIFICATION that related to the superceded BIRD45.2 was deleted.
A large portion of the BIRD49.3 text has been rewritten, so changes are not
noted.
Bob Ross
Interconnectix/Mentor Graphics
******************************************************************************
******************************************************************************
BIRD ID#: 49.3
ISSUE TITLE: Add Submodel Dynamic Clamps
REQUESTER: Neven Orhanovic, Bob Ross, Mentor G., Arpad Muranyi, Intel
DATE SUBMITTED: 4/2/98, 5/1/98, 5/21/98, 6/19/98
DATE ACCEPTED BY IBIS OPEN FORUM: Pending
******************************************************************************
******************************************************************************
STATEMENT OF THE ISSUE:
A novel type of termination technique is used in today's integrated circuits.
The termination consists of a pair of built in dynamic clamps whose V-I curves
change with time. The clamp is switched "on" when needed and switched
"off" otherwise (to conserve power). When the clamp is switched "on" its V-I
curve provides more clamping than a regular static clamp and when it is
turned "off" it behaves like a normal clamp.
The "on" switching of the dynamic clamps can be triggered by an input signal
crossing a triggering threshold or by some external clock. The "off" switching
can be triggered by a built in timer or an external clock.
******************************************************************************
STATEMENT OF THE RESOLVED SPECIFICATIONS:
The dynamic clamp functionality is added in the new Section 6a discussion:
| Dynamic Clamp:
|
| When the Submodel_type subparameter under the [Submodel] keyword is set to
| Dynamic_clamp, the submodel describes the dynamic clamp functionality.
|
| The [GND Pulse Table] and [POWER Pulse Table] keywords are defined. An
| example for a complete dynamic clamp model is provided.
|
|=============================================================================
| Keyword: [GND Pulse Table], [POWER Pulse Table]
| Required: No
| Description: Used to specify the offset voltage versus time of [GND Clamp]
| and [POWER Clamp] tables within submodels.
| Usage Rules: Each [GND Pulse Table] and [POWER Pulse Table] keyword
| introduces a table of time versus vs. points that describe
| the shape of an offset voltage from the [GND Clamp Reference]
| voltage (or default ground) or the [POWER Clamp Reference]
| voltage (or default [Voltage Range] voltage). Note, these
| voltage values are inherited from the top-level model.
|
| The table itself consists of one column of time points, then
| three columns of voltage points in the standard typ, min, and
| max format. The four entries must be placed on a single line
| and must be separated by at least one white space or tab
| character. All four columns are required. However, data is
| only required in the typical column. If minimum or maximum
| data is not available, use the reserved word "NA". Time
| values must increase as one parses down the table. The
| waveform table can contain of maximum of 100 data points.
|
| Each table must contain at least two entries. Thus, numerical
| values are required for the first and last entries of any
| column containing numerical data.
|
| The voltage entries in both the [Gnd Pulse Table] and [POWER
| Pulse Table] tables are directly measured offsets. At each
| instance, the [Gnd Pulse Table] voltage is ADDED to the [GND
| Clamp] table voltages to provide the shifted table voltages.
| At each instance, the [POWER Pulse Table] voltage is
| SUBTRACTED (because of polarity conventions) from the [POWER
| Clamp] table voltages to provide the shifted table voltages.
|
| Only one [GND Pulse Table] and one [POWER Pulse Table] are
| allowed per model.
|
| The [GND Pulse Table] and [POWER Pulse Table] interact with
| [Submodel Spec] subparameters V_trigger_f and V_trigger_r.
| Several modes of operation exist based on whether a pulse
| table and its corresponding trigger subparameter are given.
| These modes are classified as triggered, clocked, and static.
|
| Triggered Mode:
|
| For triggered mode a pulse table must exist and include the
| entire waveform; i.e., the first entry (or entries) in a
| voltage column must be equal to the last entry.
|
| Also, a corresponding [Submodel Spec] V_trigger_* subparameter
| must exist. The triggered interaction is described:
|
| The V_trigger_f subparameter under [Submodel Spec] is used
| to detect when the falling edge waveform at the die passes
| the trigger voltage. At that time the [Gnd Pulse Table]
| operation starts. Similarily, the V_trigger_r subparameter is
| used to detect when the rising edge waveform at the die passes
| the trigger voltage. At that time [POWER Pulse Table]
| operation starts. The [GND Pulse Table] dependency is shown
| below:
|
|
| Waveform at Die
|
| o o o o
| o
| o
| o -------
| |o ^
| | o | V_trigger_f
| | o v time
| | o o-------------------->
| |
| |
| | [GND Pulse Table]
| |
| | o o o o
| | o o
| | o o
| | o o
| | o o
| | o o time
| o o o o o o o o -------->
|
| ^
| |_ [GND Pulse Table] operation starts at this time
|
| The V_trigger_r and [POWER Pulse Table] operate in a similar manner. When
| the V_trigger_r voltage value is reached on the rising edge, the [POWER
| Pulse Table] is started. Normally the offset voltage entries in the [POWER
| Pulse Table] are negative.
|
| Static Mode:
|
| When the [GND Pulse Table] keyword does not exist, but the added model
| [GND Clamp] table does exist, the added model [GND Clamp] is used directly.
| Similarly, when the [POWER Pulse Table] keyword does not exist, but the
| added model [POWER Clamp] table does exist, the added model [POWER Clamp]
| is used directly.
|
| This mode provides additional fixed clamping to an I/O_* buffer or a
| 3-state buffer when it is used as a driver.
|------------------------------------------------------------------------------
|
| Example of Dynamic_clamp Model with both dynamic GND and POWER clamps:
|
[Submodel] Dynamic_Clamp_1
Submodel_type Dynamic_clamp
|
[Submodel Spec]
| Subparameter typ min max
|
V_trigger_f 1.4 1.2 1.6 | Falling edge trigger
V_trigger_r 3.6 2.9 4.3 | Rising edge trigger
|
| typ min max
| [Voltage Range] 5.0 4.5 5.5
| Note, the actual voltage range and reference voltages are inherited from
| the top-level model.
|
[GND Pulse Table] | GND Clamp offset table
| Time V(typ) V(min) V(max)
|
| 0 0 0 0
| 1e-9 0 0 0
| 2e-9 0.9 0.8 1.0
| 10e-9 0.9 0.8 1.0
| 11e-9 0 0 0
|
[GND Clamp] | Table to be offset
|
| Voltage I(typ) I(min) I(max)
|
-5.000 -3.300e+01 -3.000e+01 -3.500e+01
-4.000 -2.300e+01 -2.200e+01 -2.400e+01
-3.000 -1.300e+01 -1.200e+01 -1.400e+01
-2.000 -3.000e+00 -2.300e+00 -3.700e+00
-1.900 -2.100e+00 -1.500e+00 -2.800e+00
-1.800 -1.300e+00 -8.600e-01 -1.900e+00
-1.700 -6.800e-01 -4.000e-01 -1.100e+00
-1.600 -2.800e-01 -1.800e-01 -5.100e-01
-1.500 -1.200e-01 -9.800e-02 -1.800e-01
-1.400 -7.500e-02 -7.100e-02 -8.300e-02
-1.300 -5.750e-02 -5.700e-02 -5.900e-02
-1.200 -4.600e-02 -4.650e-02 -4.550e-02
-1.100 -3.550e-02 -3.700e-02 -3.450e-02
-1.000 -2.650e-02 -2.850e-02 -2.500e-02
-0.900 -1.850e-02 -2.100e-02 -1.650e-02
-0.800 -1.200e-02 -1.400e-02 -9.750e-03
-0.700 -6.700e-03 -8.800e-03 -4.700e-03
-0.600 -3.000e-03 -4.650e-03 -1.600e-03
-0.500 -9.450e-04 -1.950e-03 -3.650e-04
-0.400 -5.700e-05 -2.700e-04 -5.550e-06
-0.300 -1.200e-06 -1.200e-05 -5.500e-08
-0.200 -3.000e-08 -5.000e-07 0.000e+00
-0.100 0.000e+00 0.000e+00 0.000e+00
0.000 0.000e+00 0.000e+00 0.000e+00
5.000 0.000e+00 0.000e+00 0.000e+00
|
[POWER Pulse Table] | POWER Clamp offset table |
| Time V(typ) V(min) V(max)
|
| 0 0 0 0
| 1e-9 0 0 0
| 2e-9 -0.9 -1.0 -0.8
| 10e-9 -0.9 -1.0 -0.8
| 11e-9 0 0 0
|
[POWER Clamp] | Table to be offset
|
| Voltage I(typ) I(min) I(max)
|
-5.000 1.150e+01 1.100e+01 1.150e+01
-4.000 7.800e+00 7.500e+00 8.150e+00
-3.000 4.350e+00 4.100e+00 4.700e+00
-2.000 1.100e+00 8.750e-01 1.300e+00
-1.900 8.000e-01 6.050e-01 1.000e+00
-1.800 5.300e-01 3.700e-01 7.250e-01
-1.700 2.900e-01 1.800e-01 4.500e-01
-1.600 1.200e-01 6.850e-02 2.200e-01
-1.500 3.650e-02 2.400e-02 6.900e-02
-1.400 1.200e-02 1.100e-02 1.600e-02
-1.300 6.300e-03 6.650e-03 6.100e-03
-1.200 4.200e-03 4.750e-03 3.650e-03
-1.100 2.900e-03 3.500e-03 2.350e-03
-1.000 1.900e-03 2.450e-03 1.400e-03
-0.900 1.150e-03 1.600e-03 7.100e-04
-0.800 5.500e-04 9.150e-04 2.600e-04
-0.700 1.200e-04 4.400e-04 5.600e-05
-0.600 5.400e-05 1.550e-04 1.200e-05
-0.500 1.350e-05 5.400e-05 1.300e-06
-0.400 8.650e-07 7.450e-06 4.950e-08
-0.300 6.250e-08 7.550e-07 0.000e+00
-0.200 0.000e+00 8.400e-08 0.000e+00
-0.100 0.000e+00 0.000e-08 0.000e+00
0.000 0.000e+00 0.000e+00 0.000e+00
|
|==============================================================================
******************************************************************************
ANALYSIS PATH/DATA THAT LED TO SPECIFICATION:
*** Comments Below are in terms of terminology that has been changing ****
The proposal is designed to retain as much of the existing IBIS clamp syntax
as possible. The dynamic clamp V-I curve tables follow all of the
conventions of the existing V-I tables.
The overall modeling approach is to decompose the dynamic clamp into its
static and dynamic portions. The static part can be modeled by a regular
clamp. The dynamic part is connected to the same rail voltage as its static
part and modeled by a V-I curve that is shifted along the voltage axis by
the offset voltage pulse.
BIRD49 replaces the BIRD45.1 proposal of some new keywords shown below, but
preserves the intended functionality. It also reponses to the comment
that some of the V_trigger subparameters should be expressed in a
typ-min-max format. The [Model Spec] keyword structure is proposed to
do this. The replaced structure is below.
*** Note Refer to Rejected BIRD45.1 for some Original Discussion on the
Dynamic Clamp Functionality. The Discussion here is deleted since
the older terminology and details introduced confusion ***
BIRD49.1 adds the rules for when the time is negative and when the V_trigger_l
and V_trigger_h thresholds are not used and can be omitted. The intent is
to support having the dynamic clamps controlled by an external control so
that they can be preset before the signal arrives at the device. A method
was chosen that is based on the activation pulse that the simulator uses
to initiate the driver transition. Most simulators do not have an independent
pulse control to preset some other parameter.
This method can be user configurable to deal with simulator differences in
time relationships for driver activation on the net and possible phase
differences of this component on a net. The time values may have to be
readjusted for a particular part. However, the approach taken here should
work for most practical cases.
BIRD49.2 simplifies extends and simplifies the rules. Several modes of
operation are defined based on the existance of non-existency of the
[Add Model Spec] trigger subparameters (or keyword itself) and also for
when the pulse table itself is missing. The dependency on an artificial
negative time for one of the modes of operation in BIRD49.1 was not
appealing.
*** BIRD49.3 discussion with the revised Submodel terminology of BIRD48.4 ****
BIRD49.3 is modified to use the submodel terminology and keywords of BIRD48.4.
Also, the clocked mode is deleted because its operation interacted to much
with timing simulator operation - outside the realm of IBIS. It was not
clear that we could do describe the intended effects correctly (or even
by approximation) independently. This could be a candidate for further
research. We could adapt the triggered mode, even on a component-by-
component basis to approximate the effect, as a work around.
******************************************************************************
ANY OTHER BACKGROUND INFORMATION:
Based on a conversation with Arpad Muranyi on 11/14/97. Modified per a
discussion with Bob Ross, Chris Reid and Arpad Muranyi on March 11, 1998.
******************************************************************************
Received on Fri Jun 19 18:36:34 1998
This archive was generated by hypermail 2.1.8 : Fri Jun 03 2011 - 09:52:29 PDT