Forums

transformer K in LT Spice

Started by Unknown January 18, 2020

I recall seeing, somewhere, that LT Spice may handle a transformer
with K=1 different from one where K<1. I have a system that is stable
if K=0.999 but not if K=1, and there's no obvious reason why.




-- 

John Larkin         Highland Technology, Inc

The cork popped merrily, and Lord Peter rose to his feet. 
"Bunter", he said, "I give you a toast. The triumph of Instinct over Reason"



On 1/18/2020 7:16 PM, jlarkin@highlandsniptechnology.com wrote:
> > > I recall seeing, somewhere, that LT Spice may handle a transformer > with K=1 different from one where K<1. I have a system that is stable > if K=0.999 but not if K=1, and there's no obvious reason why. >
Please post your LTSpice list.
On Sunday, January 19, 2020 at 12:17:09 PM UTC+11, jla...@highlandsniptechnology.com wrote:
> I recall seeing, somewhere, that LT Spice may handle a transformer > with K=1 different from one where K<1. I have a system that is stable > if K=0.999 but not if K=1, and there's no obvious reason why.
One possible explanation is that K=1 is physically impossible, while K=0.999 is attainable. The problem with finite difference numerical integration is that rounding errors can accumulate, if there's no mechanism to attenuate them. https://www.amazon.com/Computational-Techniques-Chemical-Engineers-International/dp/1483124711 The book was written in 1966, and Spice 1 wasn't written until 1973, so the application in electronics didn't make it into the title. -- Bill Sloman, Sydney
Am 19.01.20 um 02:16 schrieb jlarkin@highlandsniptechnology.com:
> > > I recall seeing, somewhere, that LT Spice may handle a transformer > with K=1 different from one where K<1. I have a system that is stable > if K=0.999 but not if K=1, and there's no obvious reason why.
I think one uses as many 9s as one expects decades of bandwidth (crude rule of thumb). Then your signal simply might simply be out of transformer BW. Check the inductances? regards, Gerhard
Potentiometers work the same way, at 0.0 or 1.0 one or the other resistor is 
exactly zero and the network is physically different.  IIRC SPICE optimizes 
out supernodes like these.  Easy fix: don't short nodes together, leave 
1mohm or uohm or 1nH or whatever between them.

A nonideal transformer can be expressed as an ideal transformer with series 
and parallel inductances, and an ideal transformer is a pair of dependent 
ideal sources.  Same thing, but with an arbitrary ratio between nodes.

As noted, K=1 is physically impossible anyway, so if you need K=1 for the 
sim, you've already failed.  Find the smallest feasible K and solve 
backwards to a real transformer design.

Tim

-- 
Seven Transistor Labs, LLC
Electrical Engineering Consultation and Design
Website: https://www.seventransistorlabs.com/

<jlarkin@highlandsniptechnology.com> wrote in message 
news:3ab72flgvur93b7s8hd8ka1c2agf96b7kf@4ax.com...
> > > I recall seeing, somewhere, that LT Spice may handle a transformer > with K=1 different from one where K<1. I have a system that is stable > if K=0.999 but not if K=1, and there's no obvious reason why. > > > > > -- > > John Larkin Highland Technology, Inc > > The cork popped merrily, and Lord Peter rose to his feet. > "Bunter", he said, "I give you a toast. The triumph of Instinct over > Reason" > > >
On Sat, 18 Jan 2020 21:42:50 -0600, "Tim Williams"
<tiwill@seventransistorlabs.com> wrote:

>Potentiometers work the same way, at 0.0 or 1.0 one or the other resistor is >exactly zero and the network is physically different. IIRC SPICE optimizes >out supernodes like these. Easy fix: don't short nodes together, leave >1mohm or uohm or 1nH or whatever between them. > >A nonideal transformer can be expressed as an ideal transformer with series >and parallel inductances, and an ideal transformer is a pair of dependent >ideal sources. Same thing, but with an arbitrary ratio between nodes. > >As noted, K=1 is physically impossible anyway, so if you need K=1 for the >sim, you've already failed. Find the smallest feasible K and solve >backwards to a real transformer design. > >Tim
Failed? I'm experimenting with a control loop and noticed the difference, which has no immediately obvious explanation. Learning is not failing. I recall hearing about the K=1 vs K<1 thing somewhere. Thought someone might know something. -- John Larkin Highland Technology, Inc The cork popped merrily, and Lord Peter rose to his feet. "Bunter", he said, "I give you a toast. The triumph of Instinct over Reason"
On Sunday, January 19, 2020 at 4:02:02 PM UTC+11, jla...@highlandsniptechnology.com wrote:
> On Sat, 18 Jan 2020 21:42:50 -0600, "Tim Williams" > <tiwill@seventransistorlabs.com> wrote: > > >Potentiometers work the same way, at 0.0 or 1.0 one or the other resistor is > >exactly zero and the network is physically different. IIRC SPICE optimizes > >out supernodes like these. Easy fix: don't short nodes together, leave > >1mohm or uohm or 1nH or whatever between them. > > > >A nonideal transformer can be expressed as an ideal transformer with series > >and parallel inductances, and an ideal transformer is a pair of dependent > >ideal sources. Same thing, but with an arbitrary ratio between nodes.
"Ideal transformers" aren't a good approach. Spice implements the transformer equation, which is a lot safer.
> >As noted, K=1 is physically impossible anyway, so if you need K=1 for the > >sim, you've already failed. Find the smallest feasible K and solve > >backwards to a real transformer design. > > Failed? I'm experimenting with a control loop and noticed the > difference, which has no immediately obvious explanation. > > Learning is not failing.
Except that you do seem to do a lot of failing to learn.
> I recall hearing about the K=1 vs K<1 thing somewhere. Thought someone > might know something.
Spice has a tendency to get thrown by ideal - as in non-dissipative - components. You seem to need something like a milliohm or so of series resistance in inductors and capacitors. K<1 might be the same kind of requirement. -- Bill Sloman, Sydney