Transmission line modelling
Allan Herriman
<pir...@elric.freeserve.co.uk> wrote:
>Hi all,
>I'd like to simulate a two node network using a lossy transmission
>line in Pspice.
>I've got C, and R calculated L knowing Z0 but what do I use for G?
>How can I calculate it?
A reasonable first approximation would be to set G to 0.
G models the loss in the dielectric. Many (but not all) transmission
lines have a much higher copper loss than dielectric loss, so you can
safely ignore the dielectric loss as it is insignificant.
At DC (0 Hz) G models the DC leakage in the line.
What sort of line is being used? At what frequency? It might be
possible to get the exact figures for G if you give us more
information about your application.
A word of warning about PSpice lossy transmission lines - they don't
model the frequency dependence of the G,L,R,C parameters. In real
lines both R and G increase with frequency.
Regards,
Allan.
Tom Bruhns
parameters for the high-frequency end of that band and you shouldn't be
TOO far off. A good approximation of R per 100 feet for twisted-pair
line is (0.2/d)*sqrt(f) where d is the conductor diameter in inches and
f is the frequency in megahertz. It would be the unusual line that had
G significantly different from 0 up to 100MHz at least...or at least
over its intended operating frequency. Polyethylene and Teflon
insulated lines generally don't see effects from G till above 500MHz, at
least for low-impedance lines.
Cheers,
Tom
Chilean Jem wrote:
>
> > A word of warning about PSpice lossy transmission lines - they don't
> >
> > model the frequency dependence of the G,L,R,C parameters. In real
> > lines both R and G increase with frequency.
>
> Well that's the problem.... Any ideas on how I could model a lossy
> transmission line a bit closer to reality? Or am I just trying the
> impossible?
>
James Meyer
wrote:
><!doctype html public "-//w3c//dtd html 4.0 transitional//en">
><html>
>
>Well that's the problem.... Any ideas on how I could model a lossy
><br>transmission line a bit closer to reality? Or am I just trying the
>impossible?
><br> </html>
If you are using the full, not demo, version of PSpice, then you could
always make a lumped model.
Simulate an inch of the twisted pair with inductors, capacitors,
resistors, and include the magnetic coupling between the wires. Then connect
enough of the inch long segments in series to make a line as long as you need.
It's easy to do since the computer is doing all the work for you. The
simulation will be slow for several thousand elements, but the answer will be
very close to reality.
Jim
Tom Bruhns
Spice built-in model, and how it will handle the line's
frequency-dependent R (and potentially G). Can you put in a
frequency-dependent resistance in the lumped model? Does the inductance
per length not already properly model the magnetic coupling between the
two wires of the pair?
Cheers,
Tom
James Meyer
>Help! I'm not understanding how a lumped model will be better than the
>Spice built-in model, and how it will handle the line's
>frequency-dependent R (and potentially G). Can you put in a
>frequency-dependent resistance in the lumped model? Does the inductance
>per length not already properly model the magnetic coupling between the
>two wires of the pair?
The frequency dependant characteristics are handled in the lumped model
the same way nature handles them in the real world. That's the beauty of the
method. By adding extra elements, even the skin effect can be modeled.
Inductors and the magnetic coupling between them are two different
things and PSpice allows you to specify them independently.
The built-in model of a lossy transmission line is a drastic
simplification that requires terms like frequency dependant resistances and the
like in order to work. The lumped model is simpler in concept and models the
transmission line more like nature does. But the lumped model requires many
more elements and slows the computation down accordingly.
Jim
Robert
Allan Herriman wrote:
> On Tue, 14 Dec 1999 16:46:12 -0800, Tom Bruhns
> <tom_b...@agilent.com> wrote:
>
> >Help! I'm not understanding how a lumped model will be better than the
> >Spice built-in model, and how it will handle the line's
> >frequency-dependent R (and potentially G). Can you put in a
> >frequency-dependent resistance in the lumped model? Does the inductance
> >per length not already properly model the magnetic coupling between the
> >two wires of the pair?
>
> Hi Tom,
>
> Please see:
> http://www.deja.com/=dnc/[ST_rn=ps]/getdoc.xp?AN=518266031
>
> Regards,
> Allan.
Roy McCammon
>
> If your signals are effectively in a relatively narrow band, put in
> parameters for the high-frequency end of that band and you shouldn't be
> TOO far off. A good approximation of R per 100 feet for twisted-pair
> line is (0.2/d)*sqrt(f) where d is the conductor diameter in inches and
> f is the frequency in megahertz. It would be the unusual line that had
> G significantly different from 0 up to 100MHz at least...or at least
> over its intended operating frequency. Polyethylene and Teflon
> insulated lines generally don't see effects from G till above 500MHz, at
> least for low-impedance lines.
Some data on 22 gauge telephone cable.
I so the correct quantity , then the effect of letting
G=0 and then the effect of letting R = 0.
Impedance
| ---- per mile ----| G=0 R=0
F L R G C real imag real imag real imag
KHz uH ohm uS nF ohm ohm ohm ohm ohm ohm
1 798 167 1 82.0 409. -396 409 -396 99 0
10 792 169 30 81.7 149. -111 148 -111 98 0
100 773 209 603 81.0 100. -20 100 -21 98 1
1000 654 545 8900 80.8 90. -5 90 -6 90 1
10000 593 1725 78400 80.8 86. -1 86 -2 86 1
Velocity (K Miles/sec)
| ---- per mile ----|
F L R G C G=0 R=0
KHz uH ohm uS nF
1 798 167 1 82.0 29.9 29.9 123.6
10 792 169 30 81.7 82.6 82.5 124.3
100 773 209 603 81.0 123.8 123.7 126.4
1000 654 545 8900 80.8 137.3 137.3 137.6
10000 593 1725 78400 80.8 144.4 144.4 144.5
Attenuation (dB / mile )
| ---- per mile ----|
F L R G C G=0 R=0
KHz uH ohm uS nF
1 798 167 1 82.0 -1.8 -1.8 .0
10 792 169 30 81.7 -5.0 -5.0 .0
100 773 209 603 81.0 -9.4 -9.1 -.3
1000 654 545 8900 80.8 -29.7 -26.3 -3.5
10000 593 1725 78400 80.8 -116.6 -87.4 -29.2
So it looks like neglecting R is always a mistake
and G can be neglected up to 100KHz, but its starting
to make an effect at 1MHz and is significant at 10MHz.
Of course it does depend on how far you want to go.
Roy McCammon
> Well that's the problem.... Any ideas on how I could model a lossy
> impossible?
Here is what I do, using C of Fortran. Maybe you
can do it with excel of mathcad.
Look up the attenuation and velocity at several frequencies.
Convert velocity to phase shift (radians) per unit length by
phase shift per unit length = 2*pi*f/V
Put the attenuation (dB per unit length) and
phase shift into a table vs logarithm of frequency.
Spline fit those tables for interpolation.
Take one cycle of 5MHz and spread it
over 4096 samples. (2048 at +1, 2048 at -1).
FFT that.
Apply interpolated attenuation and phase shift
bin by bin to the FFT results. Be sure you
handle the bins above the Nyquist frequency properly.
Inverse FFT that.
Viola, a time domain wave form.
Allan Herriman
<tom_b...@agilent.com> wrote:
>Help! I'm not understanding how a lumped model will be better than the
>Spice built-in model, and how it will handle the line's
>frequency-dependent R (and potentially G). Can you put in a
>frequency-dependent resistance in the lumped model? Does the inductance
>per length not already properly model the magnetic coupling between the
>two wires of the pair?
Hi Tom,
Regards,
Allan.
>Cheers,
>Tom
>
>James Meyer wrote:
>>
>> On Tue, 14 Dec 1999 17:44:46 +0000, Chilean Jem <pir...@elric.freeserve.co.uk>
>> wrote:
>>
>> ><!doctype html public "-//w3c//dtd html 4.0 transitional//en">
>> ><html>
>> >
>>
>> >Well that's the problem.... Any ideas on how I could model a lossy
>> ><br>transmission line a bit closer to reality? Or am I just trying the
>> >impossible?
Allan Herriman
wrote:
>This is a bad URL.
I just checked it (again), and it still works ok.
The problem might be that your newsreader thinks that the ']' is the
end of the URL, so the wrong thing happens when you click on it. If
this is the case, you'll have to manually paste the URL into your
browser URL window.
This is a commonly reported problem with Deja.com URLs.
>Allan Herriman wrote:
[snip]
>> Please see:
>> http://www.deja.com/=dnc/[ST_rn=ps]/getdoc.xp?AN=518266031
[snip]
Allan.
William E. Sabin
Tom Bruhns wrote:
> Help! I'm not understanding how a lumped model will be better than the
> Spice built-in model, and how it will handle the line's
> frequency-dependent R (and potentially G).
Any analysis or simulation program for transmission lines must utilize the
mathematics of transmission lines. Personal computers have no problem at all with
these equations. Why anyone would want to do it any other way is hard to understand.
Bill W0IYH
> Can you put in a
> frequency-dependent resistance in the lumped model? Does the inductance
> per length not already properly model the magnetic coupling between the
> two wires of the pair?
>
Allan Herriman
<sab...@mwci.net> wrote:
>
>
>Tom Bruhns wrote:
>
>> Help! I'm not understanding how a lumped model will be better than the
>> Spice built-in model, and how it will handle the line's
>> frequency-dependent R (and potentially G).
>
>Any analysis or simulation program for transmission lines must utilize the
>mathematics of transmission lines. Personal computers have no problem at all with
>these equations.
I agree with this in general.
Be careful when applying the mathematics of transmission lines. All
the equations I've seen assume a narrow bandwidth (i.e. the parameters
are not functions of frequency).
> Why anyone would want to do it any other way is hard to understand.
The built-in PSpice transmission line doesn't model the frequency
dependence of the parameters. Other methods (such as the lumped one
we've been discussing) does not have this limitation.
For a large class of transmission line problems the frequency
dependence of R is quite important.
Regards,
Allan.
James W. Swonger
James Meyer <notj...@earthlink.net> wrote:
>On Tue, 14 Dec 1999 17:44:46 +0000, Chilean Jem <pir...@elric.freeserve.co.uk>
>wrote:
>
>always make a lumped model.
>
> Simulate an inch of the twisted pair with inductors, capacitors,
>resistors, and include the magnetic coupling between the wires. Then connect
>enough of the inch long segments in series to make a line as long as you need.
>
> It's easy to do since the computer is doing all the work for you. The
>simulation will be slow for several thousand elements, but the answer will be
>very close to reality.
ringing and a lot of simulator thrashing. The little tanks all want
to be solved, and the smaller they are, the lower the timestep has
to go. If your simulator stores every timestep solution point, this
also turns into huge raw files. It really annoyed me to have 1GHz
low level modulation on my output waveforms, and it annoyed the
receiver circuit too.
If you do go this route, be sure your segment time constant (1/sqrt(LC))
is well less than (like <1/10) the edge transition times you are going to
apply, and less than the small signal response time of whatever you're
driving.
--
##########################################################################
#Irresponsible rantings of the author alone. Any resemblance to persons #
#living or dead then yer bummin. May cause drowsiness. Alcohol may inten-#
#sify this effect. Pay no attention to the man behind the curtain. Billy!#
Malcolm Reeves
<notj...@earthlink.net> wrote:
> The built-in model of a lossy transmission line is a drastic
>simplification that requires terms like frequency dependant resistances and the
>like in order to work. The lumped model is simpler in concept and models the
>transmission line more like nature does. But the lumped model requires many
>more elements and slows the computation down accordingly.
The built-in transmission line model is better than the lumped model
as it doesn't have the lumped model discontinuity artifacts. Those
are what force you to use more lumps. If you have fast edges you will
need a lot of lumps :-!. The built in model is pure mathematics you
don't need to do anything other than specify RLCG and length. I have
found it a very good match for a lumped model. The easiest solution
if you are in doubt is try both. Pspice includes a lumped model as
well.
But I fail to understand the other threads saying that lumped models
handle component variation with frequency. The basic RLCG model does
not and is no different to the built in model in that respect. Of
course it is much easier to add addition components around the RLCG to
include the effects of skin effect etc, but not trivial.
A quick solution to this is to split the simulation into frequency
bands and change the RLGC parameters for each. If you combine these
files into one file you will get a single curve in probe. Basically
multiple curves where one stops where the next starts. It is quite
easy to automate this with a bit of programming and it saves the work
of curve fitting to some practical measurements.
Pspice does have a problem with short lines on AC analysis. If you
ramp the length you get a discontinuity in the graph. Solution is to
decrease RELTOL (See my web site). Also if you use coupled
transmission lines (NB not 2 wires in one line) be careful of the
results. If the values of the lines are very different then the order
of the lines in the spice will affect the results i.e. results are no
good for coupled lines of very different parameters.
....malcolm
--
Malcolm Reeves BSc CEng MIEE MIRSE, Full Circuit Ltd, Chippenham, UK
(mre...@fullcircuit.com or mre...@iee.org). Design Service for
Analogue/Digital H/W & S/W Railway Signalling and Power electronics.
For more details plus freeware, Win95/98 DUN and Pspice tips, see:
William E. Sabin
Allan Herriman wrote:
> On Wed, 15 Dec 1999 05:02:02 -0600, "William E. Sabin"
> <sab...@mwci.net> wrote:
>
> >Tom Bruhns wrote:
> >
> >> Help! I'm not understanding how a lumped model will be better than the
> >> Spice built-in model, and how it will handle the line's
> >> frequency-dependent R (and potentially G).
> >
> >Any analysis or simulation program for transmission lines must utilize the
> >mathematics of transmission lines. Personal computers have no problem at all with
> >these equations.
>
> I agree with this in general.
>
> Be careful when applying the mathematics of transmission lines. All
> the equations I've seen assume a narrow bandwidth (i.e. the parameters
> are not functions of frequency).
There is no reason why that should be so. My software provides corrections for
frequency variations.
Bill W0IYH
John Larkin
<sab...@mwci.net> wrote:
|Any analysis or simulation program for transmission lines must utilize the
|mathematics of transmission lines. Personal computers have no problem at all with
|these equations. Why anyone would want to do it any other way is hard to understand.
|
|Bill W0IYH
Bill,
sometimes lumped approximations are useful, especially for weird cases
like tapered lines. The math is nasty, though. For a 10:1
delay:risetime ratio, you need about 35 LC sections. For 100:1, you
need just over 1000.
Incidentally, my simulator (CircuitMaker) treats transmission lines as
if they have ideal transformers inside; you can have different DC
potentials on opposite ends of the 'coax braid'. Is this universal in
Spice? I can see how NOT doing it this way would be tricky.
John
==
A little nonsense now and then,
is cherished by the wisest men.
- Willy Wonka
William E. Sabin
John Larkin wrote:
> On Wed, 15 Dec 1999 05:02:02 -0600, "William E. Sabin"
> <sab...@mwci.net> wrote:
>
> |Any analysis or simulation program for transmission lines must utilize the
> |mathematics of transmission lines. Personal computers have no problem at all with
> |these equations. Why anyone would want to do it any other way is hard to understand.
> |
> |Bill W0IYH
>
> Bill,
>
> sometimes lumped approximations are useful, especially for weird cases
> like tapered lines.
A tapered line can be approximated by a sequence of short lines having different values
of Z0. Text books such as Kraus "Electromagnetics" and others describe the procedure.
The resultant line can be used over a limited frequency band, depending on the number of
segments.
Bill W0IYH
Allan Herriman
<sab...@mwci.net> wrote:
>
>
>Allan Herriman wrote:
>
>> On Wed, 15 Dec 1999 05:02:02 -0600, "William E. Sabin"
>> <sab...@mwci.net> wrote:
>>
>> >Tom Bruhns wrote:
>> >
>> >> Help! I'm not understanding how a lumped model will be better than the
>> >> Spice built-in model, and how it will handle the line's
>> >> frequency-dependent R (and potentially G).
>> >
>> >Any analysis or simulation program for transmission lines must utilize the
>> >mathematics of transmission lines. Personal computers have no problem at all with
>> >these equations.
>>
>> I agree with this in general.
>>
>> Be careful when applying the mathematics of transmission lines. All
>> the equations I've seen assume a narrow bandwidth (i.e. the parameters
>> are not functions of frequency).
>
>There is no reason why that should be so. My software provides corrections for
>frequency variations.
Can you post details?
Allan.
Allan Herriman
<sab...@mwci.net> wrote:
>
>
>Tom Bruhns wrote:
>
>> Help! I'm not understanding how a lumped model will be better than the
>> Spice built-in model, and how it will handle the line's
>> frequency-dependent R (and potentially G).
>
>Any analysis or simulation program for transmission lines must utilize the
>mathematics of transmission lines. Personal computers have no problem at all with
>these equations.
>Why anyone would want to do it any other way is hard to understand.
Another reason to want to simulate a lumped approximation to a
transmission line is to help design a hardware line simulator.
It's a lot easier to work with a box full of RLC parts than to carry
around a huge cable drum on a truck.
We have quite a few of these in the lab here.
(There's a terminology problem here. "Simulation" could refer to what
spice (etc) is doing, but it could also refer to the actions of a
(non-simulated) lumped RLC circuit that mimics the characteristics of
a transmission line. The original poster used the former meaning, so
I'll shut up now.)
Regards,
Allan.
>
>> Can you put in a
>> frequency-dependent resistance in the lumped model? Does the inductance
>> per length not already properly model the magnetic coupling between the
>> two wires of the pair?
>>
>> Cheers,
>> Tom
>>
>> James Meyer wrote:
>> >
>> > On Tue, 14 Dec 1999 17:44:46 +0000, Chilean Jem <pir...@elric.freeserve.co.uk>
>> > wrote:
>> >
>> > ><!doctype html public "-//w3c//dtd html 4.0 transitional//en">
>> > ><html>
>> > >
>> >
>> > >Well that's the problem.... Any ideas on how I could model a lossy
>> > ><br>transmission line a bit closer to reality? Or am I just trying the
>> > >impossible?
>> > ><br> </html>
>> >
>> > If you are using the full, not demo, version of PSpice, then you could
>> > always make a lumped model.
>> >
>> > Simulate an inch of the twisted pair with inductors, capacitors,
>> > resistors, and include the magnetic coupling between the wires. Then connect
>> > enough of the inch long segments in series to make a line as long as you need.
>> >
>> > It's easy to do since the computer is doing all the work for you. The
>> > simulation will be slow for several thousand elements, but the answer will be
>> > very close to reality.
>> >
>> > Jim
>
>
>
William E. Sabin
Allan Herriman wrote:
> Another reason to want to simulate a lumped approximation to a
> transmission line is to help design a hardware line simulator.
I agree with that objective.
Bill W0IYH
William E. Sabin
Allan Herriman wrote:
> On Fri, 17 Dec 1999 13:21:45 -0600, "William E. Sabin"
> <sab...@mwci.net> wrote:
>
> >
> >
> >Allan Herriman wrote:
> >
> >> On Wed, 15 Dec 1999 05:02:02 -0600, "William E. Sabin"
> >> <sab...@mwci.net> wrote:
> >>
> >> >Tom Bruhns wrote:
> >> >
> >> >> Help! I'm not understanding how a lumped model will be better than the
> >> >> Spice built-in model, and how it will handle the line's
> >> >> frequency-dependent R (and potentially G).
> >> >
> >> >Any analysis or simulation program for transmission lines must utilize the
> >> >mathematics of transmission lines. Personal computers have no problem at all with
> >> >these equations.
> >>
> >> I agree with this in general.
> >>
> >> Be careful when applying the mathematics of transmission lines. All
> >> the equations I've seen assume a narrow bandwidth (i.e. the parameters
> >> are not functions of frequency).
> >
> >There is no reason why that should be so. My software provides corrections for
> >frequency variations.
>
> Can you post details?
I use the ARRL Radio Designer program. The line loss is specified at some frequency and is
modified according to the square root of frequency, which is a pretty good approximation
for most transmission lines that are using the usual TEM mode.
Bill W0IYH