The Cyber-Spy.Com Usenet Archive Feeds Directly
From The Open And Publicly Available Newsgroup
This Group And Thousands Of Others Are Available
On Most IS NNTP News Servers On Port 119.
Cyber-Spy.Com Is NOT Responsible For Any Topic,
Opinions Or Content Posted To This Or Any Other
Newsgroup. This Web Archive Of The Newsgroup And
Posts Are For Informational Purposes Only.
Reply-To: "Kevin Aylward"
From: "Kevin Aylward"
References: <3D809F7A.firstname.lastname@example.org> <5Pgg9.email@example.com>
Subject: Re: LT SwitcherCAD Current Source Help
X-Newsreader: Microsoft Outlook Express 6.00.2600.0000
Date: Sat, 14 Sep 2002 07:58:59 +0100
NNTP-Posting-Date: Sat, 14 Sep 2002 07:58:59 BST
"Mike Engelhardt" wrote in message
> Kevin wrote:
> > Mike wrote:
> > > The implementation of the voltage controlled switch
> > > in Berkeley SPICE has a continuous 1st derivative,
> > > though it has a discontinuity in value. The
> > > continuous 1st derivative helps it converge even
> > > when its right at the edge of switching between the
> > > two conductivity values.
> > >
> > > The only improvement is to offer a switch like that
> > > in LTspice which is everywhere continuous in value
> > > and derivative.
> > >
> > Weeeeelllll, you could use a VCIS and build up a voltage controled
> > resister with a B source and have pretty much any transfer function
> > like.
> Precisely. In principle you can get continuous value and
> derivatives with that method. In practice, however, the
> compiler in Berkeley SPICE makes inefficient code and the
> execution is slow.
This may be, but if the circuit is of any decent size and there is only
a few switches, this might not be important. Even so, it would seem that
it might still be faster then using say, a mosfet model.
>Even more serious are the errors in the
> Jacobian computed which can defeat the method for
> getting continuous value and derivatives. Even though these
> issues are fixed in LTspice, you still have to deal with
> Newton iteration. Basically, you get poor performance
> if you use straight Newton iteration without knowledge of
> the non-linear functions you're solving. The integrated
> switch(in LTspice) gives the Newton hints for the next
> iteration which really picks up the performance and
> convergence capability.
I have now reconciled why I stated that spice had a continuous switch
model. It is XSpice that has the continuous switch, which I erroneously
attributed to spice3/2. The description from the xpsice manual is:
Description: The Analog Switch is a resistor that varies either
logarithmically or linearly
between specified values of a controlling input voltage or current. Note
that the input is
not internally limited. Therefore, if the controlling signal exceeds the
specified OFF state
or ON state value, the resistance may become excessively large or
excessively small (in the
case of logarithmic dependence), or may become negative (in the case of
For the experienced user, these excursions may prove valuable for
modeling certain devices,
but in most cases you are advised to add limiting of the controlling
input if the possibility
of excessive control value variation exists.
.model switch3 aswitch(cntl_off=0.0 cntl_on=5.0 r_off=1e6
+ r—on=10.0 log=TRUE)
SuperSpice, a very affordable Mixed-Mode
Windows Simulator with Schematic Capture,
Waveform Display, FFT's and Filter Design.
Go Back To The Cyber-Spy.Com
Usenet Web Archive Index Of
The sci.electronics.design Newsgroup