Re: BIRD #61 -- Characterization of Receivers

nikolai@avanticorp.com wrote:
>
> Hi All,
>
> Can't anyone clarify this new spec. about receivers for me
> and what the intention is?

> Qestions:
>
> *********
> * 1 *
> *********
>
> 1. What is a receiver? Is it
> a) A part of output buffer which receives controlloing
> signal (in other words, input of output type buffer).
> In this case it can be extended to enable node
> of 3-state and I/O buffers.
> b) That part of Input buffer, which is looking into
> a transmission line and receives signals from
> some output type buffer through the transmission line.

(B)

> I would expect the unswer should be b). However, the spec.
> says:
>
> ===begin===
>
> 1) The following new keyword is defined, and is placed in the
> specification just below the [Rising Waveform] and [Falling Waveform]
> keyword descriptions
>
> === end ===
>
> But Input buffer has no [Rising Waveform] or [Falling Waveform].
> I/O buffer does have, but Input buffer does not have and
> we have no place to place [Receiver Delay] keyword.

Keep in mind that IBIS specifications are relatively free-form. Within
a given [Model] there are few constraints on where you place a given
keyword; the paragraph you quoted is purely editorial and describes the
location where the new text will appear in the revised IBIS standard
document.

> *********
> * 2 *
> *********
>
> 2) Examples are confusing. The should give receiver delay.
> What we have in example 1:
>
> ===begin===
>
> [Receiver Delay]
> Start_point = 0.8v
> Slope = 1v/1ns
> End_point TABLE
> |variable typ min max
> 1.4 INF 7.0ns INF
> 1.5 INF 5.0ns INF
> 1.51 INF 3.0ns 10.0ns
> 1.53 7.0ns 0.0ns 7.0ns
> 1.55 2.0ns 0.0ns 1.0ns
> 1.6 0.0ns 0.0ns 0.0ns
> 1.7 0.0ns 0.0ns 0.0ns
> 2.0 0.0ns 0.0ns 0.0ns
> 2.1 0.1ns 0.1ns 0.1ns
> 2.5 -0.2ns -0.1ns -0.5ns

NB: SP got caught by the usual gotcha and has max=slow instead
of max=fast

> === end ===
>
> delay is about 1ns, ok

No, delay isn't specified at all here. What this table says is that if
your input waveform rises at 1v/ns from 0.8v to 2.0v (the databook
test condition) then there is no change in the input delay from the data
book timing (DUH!) OTOH, if the input rises from 0.8v to 1.55v at 1v/ns
then the input delay under slow conditions is unchanged, the delay under
typical conditions is 2.0ns longer than databook, and the delay under
fast conditions is 1.0ns slower than databook. (OK, so these are mixed up.
See above.)

> example 2:
>
> ===begin===
>
> [Receiver Delay]
> Start_point = 0.8v
> End_point = 2.0v
> Slope TABLE
> |variable typ min max
> 0.05/1ns -1.0 -3.5 -1.0
> 0.25/1ns -0.75 -2.0 -0.05
> 0.50/1ns -0.01 -0.5 -0.0
> 0.60/1ns 0.0 0.25 0.0
> 0.75/1ns 0.0 0.0 0.0
> 1.00/1ns 0.0 0.0 0.0
> 2.00/1ns +0.5 +0.2 +1.0
>
> === end ===
>
> is it a delay about 1s, or data are in Volts,
> or Nano is missing, or what is it?

Again, these are relative rather than absolute delays.
And you're correct that the dimensions in the table entries
should be nanoseconds.

> *********
> * 3 *
> *********
>
> 3) Some delays are negative. This is very very bad.
> Should be always positive. Just from stand point of
> causality. You apply a signal and then 1ns BEFORE that event
> your receiver tells you the signal has arrived.
> Simulators have to add some 'base' delay.
> Simulator_A will add 1ns, Simulator_B will add 10ns, etc.
> Why not IBIS spec add this 'base' delay and all have
> less headahe?

The delays are adjustments to databook timing, so negative numbers are
not a problem. As for 'base', IBIS is not intended to contain an
entire part data sheet. If we were to try THAT, we'd need to specify
the setup time to arbitrary signals (eg setup to end of RESET) and so
forth, which is far, far, farther into the functional space than we
can go. Not gonna happen.

Instead, this BIRD just applies signal-integrity adjustments to the
timing already described in some other format (eg databook). For
instance, DDR SDRAM devices are characterized with input waveforms
swinging from Vil(ac) to Vih(ac) at 1v/ns (or the reverse). In real
applications, though, some inputs could actually slew faster and this
might result in a hold-time violation. The proposed BIRD allows system
designers to refine their timing analyses based on input wave SHAPE as
well as threshold crossing.

> *********
> * 4 *
> *********
>
> 4) Input buffer has Vinl, Vinh. If [Receiver Delay]-s
> are specified, then Vinl, Vinh should be discarded. Right?
> If ONLY 1 [Receiver Delay] is given, say for rising ramp,
> then we still use Vinl, Vinh, right ?
> Vinl, Vinh MUST be always given and BE CONSISTENT with
> [Receiver Delay] if the last is given. Right?

Vinh and Vinl are holdovers from TTL data sheets. They have some minor
utility as extreme-worst-case noise margin specs; they are quite
useless for timing analysis. Vmeas is much more relevant, and in
these tables have no effect on its use. In fact, they depend on it.

We would hope that based on the information in these tables simulator
companies would refine their noise analysis models to recognize that
a transient that (e.g.) rang back to 200mV above a Vinl of 800mV for
250ps in reality has no chance of causing a logic violation.

> *********
> * end *
> *********
>
> ----- Begin Included Message -----
>
> Buffer Issue Resolution Document (BIRD)
>
> BIRD ID#: 61
> ISSUE TITLE: Enhanced Characterization of Receivers
> REQUESTOR: D.C Sessions (Philips), Richard Mellitz, Stephen Peters,
> Arpad Muranyi (Intel Corp)
> DATE SUBMITTED: August 9, 1999
> DATE ACCEPTED BY IBIS OPEN FORUM:
>
> *******************************************************************************
> *******************************************************************************
>
> STATEMENT OF THE ISSUE: The current specification allows the user to
> describe the static Vinh and Vinl thresholds of a receiver. However, these
> parameters specify only the DC input thresholds at which the output of
> a receiver switches state. They do not describe a digital logic receiver's
> dynamic performance. Dynamic performance includes such items as how a
> device's setup or hold time varies with input overdrive, or how a receiver
> behaves when an input waveform rings back into the area between Vinh and
> Vinl. In addition, the current specification does not support simulation
> methodologies that rely on specifying the total delay from the input of an
> output buffer to the output of the receiver. This bird offers a way for the
> user to specify a receiver's propagation delay and dynamic characteristics in
> enough detail so that a simulation tool can model a receiver's response to
> any arbitrary waveform.
>
> *******************************************************************************
>
> STATEMENT OF THE RESOLVED SPECIFICATIONS:
>
> 1) The following new keyword is defined, and is placed in the
> specification just below the [Rising Waveform] and [Falling Waveform]
> keyword descriptions
>
> |=============================================================================
> | Keyword: [Receiver Delay]
> | Required: No
> | Sub-Params: Start_point, End_point, Slope
> | Description: This keyword specifies how the propagation delay of an
> | input receiver varies as a function of input overdrive or
> | input waveform slope.
> |
> | Usage Rules: The [Receiver Delay] keyword is followed immediately by the
> | three subparameter names and their arguments. The Start_point
> | and End_point subparameters define the starting and ending
> | voltage points of a linear ramp while the Slope parameter
> | specifies the slope of that ramp. Slope is given in units of
> | volts per second (V/S). All three subparameters are required.
> |
> | Subparameter Usage:
> | The intent of the subparameters is to specify a set of linear
> | ramps that vary only in starting point, ending point, or
> | slope. Therefore, when assigning values to the subparameters,
> | two of the three subparameters must be assigned fixed values,
> | while the third subparameter uses as its argument the reserved
> | word TABLE. The use of the word TABLE indicates that this
> | subparameter is the independent variable in the associated
> | receiver delay table. The subparameter that uses the TABLE
> | argument must appear after the other two subparameters and
> | before the receiver delay table.
> |
> | Numerical arguments are separated from their associated
> | subparameter by an equals sign (=); white space around the
> | equals sign is optional. The TABLE argument is separate
> | from its associated subparameter by white space.
> |
> | Receiver Delay Table:
> | Following the subparameters is the receiver delay table itself.
> | This table consists of four columns. The first column lists
> | either the starting voltage, ending voltage or slope of the
> | linear ramp (i.e. the independent variable) as determined by
> | the subparameter with the TABLE argument. The second thru
> | fourth columns list the change in the receiver's propagation
> | delay. This "change in delay" is relative to the receiver's
> | delay when it is stimulated using the same overdrive and edge
> | rate conditions used to specify the device's setup and hold
> | times. The delay columns are arranged in the standard typ,
> | min, max format. All four entries must be placed on the same
> | line and must be separated by at least one white space. All
> | four columns are required. However, data is required only for
> | the typical column. If minimum or maximum data is not available
> | use the reserved word NA. Each receiver delay table is limited
> | to 100 rows and only one receiver delay table is allowed per
> | [Receiver Delay] keyword. However, there are no restrictions
> | on the number of [Receiver Delay] keywords per [Model] keyword.
> | Note that the [Receiver Delay] keyword is not allowed unless
> | the Model_type of the [Model] is one of the Input or I/O
> | models types.
> |
> | Use of INF as a Receiver Delay:
> | When building a receiver delay table the user may specify an
> | input condition that does not result in the receiver's output
> | changing state. In that case, the receiver delay is considered
> | infinite, and the reserved word INF is used in the delay
> | column. See the examples below.
> |
> | Other Notes:
> | It is expected that the user will provide enough [Receiver
> | Delay] keywords (i.e. characterization data) to allow a CAE
> | tool to build a model of the input receiver. It is expected
> | that at least four [Receiver Delay] keywords will be required;
> | two low to high going ramps and two high to low going ramps.
> | However, the exact number of ramps and there content is up
> | to the user and recommendations provided by the various
> | simulation vendors.
> |
> | An example table showing how receiver delay varies vs. overdrive.
> | Note the use of the reserved word INF.
> |
> [Receiver Delay]
> Start_point = 0.8v
> Slope = 1v/1ns
> End_point TABLE
> |variable typ min max
> 1.4 INF 7.0ns INF
> 1.5 INF 5.0ns INF
> 1.51 INF 3.0ns 10.0ns
> 1.53 7.0ns 0.0ns 7.0ns
> 1.55 2.0ns 0.0ns 1.0ns
> 1.6 0.0ns 0.0ns 0.0ns
> 1.7 0.0ns 0.0ns 0.0ns
> 2.0 0.0ns 0.0ns 0.0ns
> 2.1 0.1ns 0.1ns 0.1ns
> 2.5 -0.2ns -0.1ns -0.5ns
> |
> | A second table that characterizes receiver delay vs. the slope of the
> | input waveform.
> |
> [Receiver Delay]
> Start_point = 0.8v
> End_point = 2.0v
> Slope TABLE
> |variable typ min max
> 0.05/1ns -1.0 -3.5 -1.0
> 0.25/1ns -0.75 -2.0 -0.05
> 0.50/1ns -0.01 -0.5 -0.0
> 0.60/1ns 0.0 0.25 0.0
> 0.75/1ns 0.0 0.0 0.0
> 1.00/1ns 0.0 0.0 0.0
> 2.00/1ns +0.5 +0.2 +1.0
> |
> |
> |2) Item 2 under General Syntax Rules and Guidelines is modified to include
> |the following reserved words
>
> INF - used when a data value is so large it is considered infinite
> TABLE - used to indicate that a subparameter's value(s) are part of a
> table.
> *******************************************************************************
>
> ANALYSIS PATH/DATA THAT LED TO SPECIFICATION: This bird originated from
> discussions held at the January '99 IBIS summit regarding the lack of a
> way to accurately and realistically specify the input thresholds and
> dynamic performance of a digital logic receiver. At the mid year '99 IBIS
> summit D.C. Sessions, Stephen Peters, Richard Mellitz and Arpad Muranyi
> agreed to meet to hammer out a proposal. This bird is one of the results.
> The technical basis of this bird derives from simulation work done by
> Richard Mellitz.
>
> *******************************************************************************
>
> ANY OTHER BACKGROUND INFORMATION:
>
> *******************************************************************************
>
> ----- End Included Message -----

-- 
D. C. Sessions
dc.sessions@vlsi.com