Need Spice Model for twisted pair
Bruce Nepple
evaluate crosstalk effects when 1 wire of the pair is not terminated, but
both are driven, and some other bizarre conditions.
The only thing I'm sure about is that pspice tlossy isn't the answer.
bruce
remove extra stuff to send email
Todd K. Moyer
This gets very complicated; I know because I'm doing it for my work. I think
HSpice (by Avant!) has come out or will come out very soon with a new
transmission line circuit element, "W", that would handle your problem. I
believe they are even working on a geometrical solver which would let you
enter dimensional information to build the model. You could investigate that
if you have tens of kilo-bucks and a Sun workstation.
If your transmission line is not extremely long (by "long" I mean, say, more
than a few wavelengths at the highest frequency of interest), you could
build up a transmission line model with R, L, and C lumped elements. Each
"lump" would be a resistor and inductor in series for each wire, and a
capacitor in shunt, as shown below (please excuse artwork). The more of
these lumps you cascade, the better the approximation to a real, distributed
transmission line. To model coupling between pairs, then, you'd have to put
little capacitors from one transmission line to the other and couple the
inductors. Of course it would be a big project to set the values of those
capacitors and the coupling coefficients for the inductors.
---/\/\----^^^^---|
_|_
___
|
---/\/\----^^^^---|
Now to get really depressing, this model won't properly account for skin
effect unless you can make the resistors and inductors frequency dependent.
I don't think there is an easy way to do that in PSpice, but maybe with the
behaviorial modeling. If I'm not mistaken, the "tlossy" element doesn't even
account for that; it just, essentially, allows you to have the resistors in
the above lumped model, while the lossless line doesn't.
If you're still not discouraged, let me know, and I can give some guidance
on how to come up with values of the lumped elements in the model. But I'd
say this could easily blow up into a several month project (it has for me);
you might be better off getting some hunks of wire and making measurements.
I was truly surprised at how difficult it is to get an accurate model of a
long twisted pair line. Efficient transient simulation of such a line is an
area of active research. If you're so inclined, search IEEE publications for
an author named Dmitri Kusnetsov. For example, "Optimal Transient Simulation
of Transmission Lines", Kusnetsov et al, IEEE Trans on Circuits and
Systems-I, February 1996. Kusnetsov's work is implemented in the HSpice "W"
element.
Todd
Bruce Nepple wrote in message <8y7r2.197$E55....@news7.ispnews.com>...
John Woodgate
newsgroup posts.
<78j32l$se$1...@remarQ.com>, Todd K. Moyer <todd_...@yahoo.com> wrote:
>To model coupling between pairs, then, you'd have to put
>little capacitors from one transmission line to the other and couple the
>inductors. Of course it would be a big project to set the values of those
>capacitors and the coupling coefficients for the inductors.
Not really. You can measure the capacitance and inductance of say a
metre length of the cable and divide into say six discrete sections. To
determine the coupling, form the cable into a loop and measure the
inductance of one conductor (about 1.6 uH) with the other conductor
first open and then short-circuited. You can calculate the coupling from
these values. Use as low a frequency as possible to minimise any
capacitance effects.
--
Regards, John Woodgate, Phone +44 (0)1268 747839 Fax +44 (0)1268 777124.
OOO - Own Opinions Only. ERROR! OUT OF CORNFLAKES. Please check cereal port
configuration.
Todd K. Moyer
pair of wires, but I meant to say that the coupling between 2 different
pairs was difficult to model. I would think that depends on the "lay" of the
pairs within the jacket, and may be significantly different from cable to
cable.
Todd
John Woodgate wrote in message ...
Bruce Nepple
expected no problem finding the information I needed, especially in this day
of 'everyone has spice".
I'm looking at 30ft of twisted pair, and 5-10ns rise times. A rule of thumb
I read somewhere was a 10:1 ratio of rise time to element delay to get
decent results.
Capacitance per foot between pairs and to the shield is specified by the
cable manufacturers and you can calculate inductance from that and
impedance. Do those values work for you?
I have no idea of the magnitude of the inductive coupling. Have you done
any experiments with that?
Regarding skin effect, I recently downloaded a "paper" from a consulting
company that addressed that. It didn't contain much real information, but
it pointed out a methodology. (NESA is the company name). I don't have a
soft copy or i'd email it to you. It should be easy enough to find.
I'm interested in whatever you are willing to share.
I have a friend with HSpice, so I check that out also.
bruce
Todd K. Moyer wrote in message <78j32l$se$1...@remarQ.com>...
>Bruce,
>
>This gets very complicated; I know because I'm doing it for my work. I
think
>HSpice (by Avant!) has come out or will come out very soon with a new
>transmission line circuit element, "W", that would handle your problem. I
>believe they are even working on a geometrical solver which would let you
>enter dimensional information to build the model. You could investigate
that
>if you have tens of kilo-bucks and a Sun workstation.
>
>If your transmission line is not extremely long (by "long" I mean, say,
more
>than a few wavelengths at the highest frequency of interest), you could
>build up a transmission line model with R, L, and C lumped elements. Each
>"lump" would be a resistor and inductor in series for each wire, and a
>capacitor in shunt, as shown below (please excuse artwork). The more of
>these lumps you cascade, the better the approximation to a real,
distributed
>transmission line. To model coupling between pairs, then, you'd have to put
>little capacitors from one transmission line to the other and couple the
>inductors. Of course it would be a big project to set the values of those
>capacitors and the coupling coefficients for the inductors.
>
-Bodnar,B.L.
Bruce Nepple <bru...@imagenation.extra.com> wrote:
>I'm amazed that there isn't a paper somewhere that really goes into this. I
>expected no problem finding the information I needed, especially in this day
>of 'everyone has spice".
>
>I'm looking at 30ft of twisted pair, and 5-10ns rise times. A rule of thumb
>I read somewhere was a 10:1 ratio of rise time to element delay to get
>decent results.
>
>Capacitance per foot between pairs and to the shield is specified by the
>cable manufacturers and you can calculate inductance from that and
>impedance. Do those values work for you?
>
>I have no idea of the magnitude of the inductive coupling. Have you done
>any experiments with that?
>
>Regarding skin effect, I recently downloaded a "paper" from a consulting
>company that addressed that. It didn't contain much real information, but
>it pointed out a methodology. (NESA is the company name). I don't have a
>soft copy or i'd email it to you. It should be easy enough to find.
>
>I'm interested in whatever you are willing to share.
>
[interesting text deleted for brevity]
I don't know if this will be of any help...
On a hunch, I pulled out my old edition of TRANSMISSION SYSTEMS FOR
COMMUNICATIONS, revised 4th edition, Bell Telephone Laboratories (1971). The
following is data which may be of interest:
Typical insulation is polyethylene. Common wire sizes are 19, 22, 24, and 26
gauge. The wires are usually made of high purity copper.
Quoting from the book:
"The primary constants of twisted pair cables are subject to manufacturing
deviations, and change with the physical environment such as temperature,
moisture, and mechanical stress. The inductance, L, is of the order 1 mH/mile
for low frequencies and the capacitance, C, has two standard values of 0.066
and 0.083 uF per mile although lower capacitance cables are under development.
Of the primary constants, only C is relatively independent of frequency; L
decreases to about 70 percent of its initial value as frequency increases from
50 kHZ to 1 MHz and is stable beyond; G is very small for PIC (polyethylene
insulated cables) and roughly proportional to frequency for pulp insulation;
and R, approximately constant over the voiceband, is proportional to the
square root of frequency at higher frequencies where skin effect and proximity
effect dominate."
I do not deal in this area, and so I certainly am not an "expert" on this
topic. The R, C, G, L are the usual parameters used in transmission line work
(which I haven't done in at least 15 years).
Best wishes,
Bohdan Bodnar
bbo...@lucent.com
[blanks added to keep silly posting software happy]
Mike Palmer
Two hot areas for chips right now are high speed ethernet and DSL. Both of
these send high speed data over twisted pair wires. If, by some chance,
you're using phone wire as your twisted pair, the specs for high speed
ethernet and DSL can provide some information on the specific wire you're
using. If not, then they might still potentially provide you with some
valuable information.
In both these cases, some kind of adaptive equalization has to be used to
compensate for (among other things) variations in the cable.
-- Mike --
Bruce Nepple wrote in message <8y7r2.197$E55....@news7.ispnews.com>...
>Any ideas where I can get help modeling twisted pair cable. I want to
>evaluate crosstalk effects when 1 wire of the pair is not terminated, but
>both are driven, and some other bizarre conditions.
>
>The only thing I'm sure about is that pspice tlossy isn't the answer.
>
John Woodgate
newsgroup posts.
<78ksg7$jmp$1...@remarQ.com>, Todd K. Moyer <todd_...@yahoo.com> wrote:
>You're right, John, it is not too difficult to model the parameters of the
>pair of wires, but I meant to say that the coupling between 2 different
>pairs was difficult to model. I would think that depends on the "lay" of the
>pairs within the jacket, and may be significantly different from cable to
>cable.
Oh, yes, you have to measure the cable you are actually going to use, or
pay for 'star-quad' which is carefully constructed to meet low coupling
specifications between pairs.
Allan Herriman
As other posters pointed out, a lumped approximation
consisting of N RLC sections will work quite well up to a certain
cutoff frequency.
If r, l and c are the component values in each section (not the same
as the per km values) then the input Z of the line will go to zero at
the same frequency that the gain hits the "brickwall".
This frequency will be about 2 * (1/ (2 pi sqrt(lc)))
The simulation is perhaps usable up to half this frequency.
(An N section simulation is actually a 2N pole low pass filter.
Think about it!)
For a given R, L, C (the per km values), the number of sections needed
is proportional to the product of the attenuation (in dB) and the
bandwidth required.
For details of how to model skin effect and dielectric loss, please
use Dejanews to search for my comp.dsp post of 1997/08/28 with subject
"simulation of ANSI test loops"
http://x10.dejanews.com/getdoc.xp?AN=268722752&search=thread&threaded=1&CONTEXT=917402827.1451163837&HIT_CONTEXT=917402591.1522466820&HIT_NUM=36&hitnum=1
Values of the parameters for various wire gauges with various coatings
at various frequencies can be obtained from any of the
telecommuncations standards that use these models for testing modems,
etc.
See:
ETSI ETR-152 (HDSL standard)
ETSI ETR-080 (ISDN U interface standard)
ANSI T1.601 (ISDN U interface standard)
etc.
These are all available on the web (for a price...)
The ANSI standards seem to be based on Bellcore research of US cable
plant parameters.
good luck,
Allan.
martin griffith
<bru...@imagenation.extra.com> wrote:
>Any ideas where I can get help modeling twisted pair cable. I want to
>evaluate crosstalk effects when 1 wire of the pair is not terminated, but
>both are driven, and some other bizarre conditions.
>
>The only thing I'm sure about is that pspice tlossy isn't the answer.
>
>bruce
>
>remove extra stuff to send email
>
>
Give Belden cable a call, http://www.belden.com/products/tpvutp85.htm
they are pushing 350MHz cat 5 stuff quite hard.There are a few tech
docs on their sites that may help. but probably no spice stuff
martin
delete SPAM in my address
"Whatever you do will be insignificant, but it is very important that
you do it." --Mahatma Gandhi
James Meyer
>
>How the hell do these people get their stuff through the emissions
>regulations?
>
>>Give Belden cable a call, http://www.belden.com/products/tpvutp85.htm
>>they are pushing 350MHz cat 5 stuff quite hard.There are a few tech
>>docs on their sites that may help. but probably no spice stuff
How the hell would a spice model improve the chances of
meeting emmissions regulations?
Jim
John Woodgate
newsgroup posts.
<36b20f72....@news.netcomuk.co.uk>, Peter <z1...@nospam24.com>
wrote:
>How the hell do these people get their stuff through the emissions
>regulations?
>
>>Give Belden cable a call, http://www.belden.com/products/tpvutp85.htm
>>they are pushing 350MHz cat 5 stuff quite hard.There are a few tech
>>docs on their sites that may help. but probably no spice stuff
>
Well, 350 MHz is probably a very significant frequency, for Europe
anyway. Most current emission standards do not control emissions above 1
GHz, and the thrd harmonic of 350 MHz is ...
It won't last, of course. Extension of the limits to higher frequencies
has been under study for some time.
martin griffith
>
>How the hell do these people get their stuff through the emissions
>regulations?
>
snip
>--
>Peter.
>
>Return address is invalid to help stop junk mail.
>E-mail replies to zX...@digiYserve.com but remove the X and the Y.
>Please do NOT copy usenet posts to email - it is NOT necessary.
It's differential, its twisted, its unsheilded, its square waves, and
maybe everything cancels out, as it does in theory
( fat chance....)
martin
OK, how do you want the quote, Cheap or Inexpensive?
cfo
I have very limited electronics background but can follow "very" simple
scematics/instructions and would like to get one of those "dawn simulating"
lights that turn on very slowly over the period of about 1/2 and hour but I
don't have $200.00 to spend on the ones I've seen advertized.
Does anyone have a simple design I could impliment?, or know of an
inexpensive device available somewhere I could use?
I was thinking of connecting something to an alarm clock and a lamp to get
the whole thing started, leaving just the "slow on" circuit to be completed.
Thanks for any ideas