Re: FW: [SI-LIST] : How to identify SSO groups?

Scott McMorrow ([email protected])
Mon, 29 Dec 1997 11:58:36 -0800

To lump or not to lump capacitance is a question based upon
the rise time of the launched waveform on the net. If the distance
between distributed capacitance elements is less than about one
forth the rise/fall time of the waveform, then lumping works as
a first approximation. However, if the distance is larger then the
system acts a transmission line, and the transmission line effects
of driving a line with capacitive mismatches must be analyzed.
In the limit case where all capacitance is lumped at the end of
a fairly long (electrically) transmission line, then the driver launches
its current waveform into a pure transmission line element. Thus, the
characteristic impedance of the line limits the initial sso current, until
the reflection off of the capacitance returns.

Generally, at todays edge rates, all sso capacitive loading scenerios
must be analyzed as a transmission line environment. The current
profile produced can then be used to simulate sso effects in a Spice
simulator. Or a very good swag (approximation) can be achieved
by looking at the current waveform dI/dt and multiplying it by the
ground (or power, when appropriate) inductance of the package and
layout.

regards,

Scott McMorrow
SiQual

Peterson, James F (FL51) wrote:

> Manix makes a good point in regards to Praveen's question on SSO. But
> can you really lump all of that load capacitance together? It seems that
> with a bus that is large enough to contain 500pF, then to be accurate,
> the load has to be distributed, at least somewhat.
> best regards,
> Jim Peterson
> [email protected]
>
> > ----------
> > From: Manix Velu[SMTP:[email protected]]
> > Sent: Monday, December 22, 1997 11:51 PM
> > To: [email protected]
> > Cc: [email protected]
> > Subject: [SI-LIST] : How to identify SSO groups?
> >
> > Hi Alaa F.Alani,
> >
> > It is not even 1ns. Even switching within 10nsec width also may be
> > considered as SSO. What our vendors propose thus is to use a delay of
> > around half-a-clock (say for about 40 to 60 ns) to skew the outputs
> > of around 12mA typ sinking a load capa of 500pF!!!
> >
> > Regards,
> > Manix Velu.
> >
> > > Date: Mon, 22 Dec 1997 09:39:01 GMT
> > > From: [email protected] (alaa alani)
> > > To: [email protected]
> > > Subject: Re: [SI-LIST] : SSO : How to identify SSO groups?
> > > Cc: [email protected]
> >
> > >
> > > Hi Praveen,
> > >
> > > It depends on how long it takes for the current to drop to a certain
> > level (eg 0mA)
> > > but 1nsec seems to be acceptable (i.e. those i/o buffers switching
> > within 1nsec
> > > are considered as simultaneously switching).
> > >
> > > Rgds,
> > >
> > > Dr. Alaa F. Alani
> > > Senior Signal Integrity Eng.
> > > LSI Logic
> > > UK
> > >
> > >
> > >
> > > > From [email protected] Fri Dec 19 23:15:02 1997
> > > > X-From: uucp Fri Dec 19 15:13 PST 1997
> > > > >Received: from mars.Sun.COM (mars.Sun.COM [192.9.22.1]) by
> > mail1.lsil.com with ESMTP id LAA10882
> > > > (8.6.12/IDA-1.6); Fri, 19 Dec 1997 11:31:22 -0800
> > > > Date: Fri, 19 Dec 1997 10:50:28 -0800
> > > > From: [email protected] (Praveen G Shekokar)
> > > > To: [email protected]
> > > > Subject: [SI-LIST] : SSO : How to identify SSO groups?
> > > > Sender: [email protected]
> > > > Content-Length: 1126
> > > >
> > > >
> > > > We are using our foundaries SSO rules to calculate number of
> > VDD/VSS
> > > > pins required for our ASIC. Before we can use the SSO rules we
> > need to
> > > > identify signals that form a SSO group. The conventional technique
> > I know
> > > > of is to group signals of similar functionality into a SSO group
> > e.g.
> > > > Address/Data bus.
> > > >
> > > > In practice we have noticed that not all the signals of such a
> > group have
> > > > identical delays. Further they may not have identical output
> > loads. Variable
> > > > delays implies that not all the outputs switch at the same. E.g.
> > > >
> > > > Let output1 switch at time T and output2 at time T+t1. If t1 is
> > small then the
> > > > switching current for output1, say I1, will overlap with switching
> > current
> > > > for output2, say I2, leading to overall increase in the switching
> > current.
> > > > But if t1 is more than the time required for I1 to reduce to 0
> > then the
> > > > overall switching current is going to be limited to max of I1 and
> > I2.
> > > >
> > > > I want to know whether there are any guide lines for value of t1
> > that one
> > > > can use to decide whether a signal is part of a SSO group.
> > > >
> > > > Value of t1 is also required in designing address/data stepping.
> > > >
> > > > thanks,
> > > >
> > > > - Praveen Shekokar
> > > >
> > > >
> > >
> > >
> >
> > E-Mail: [email protected]
> > Shuttle Technology Ltd.,
> > 32, St. Therese Street,
> > P.B.225,
> > Pondicherry - 605001,
> > India.
> >
> > Office Telephone : 91.413.62301
> > Residence Telephone : 91.413.40535
> > Facsimile Number : 91.413.62095
> > DISA Office Telephone : 91.413.62099 <wait for intercomm tone>
> > <press*> <press208>
> > URL: shuttletech.com
> >