Follow along with the video below to see how to install our site as a web app on your home screen.
Note: This feature may not be available in some browsers.
On Fri, 19 Apr 2019 11:48:50 -0700 (PDT), George Herold
gherold@teachspin.com> wrote:
On Friday, April 19, 2019 at 2:30:06 PM UTC-4, John Larkin wrote:
On Fri, 19 Apr 2019 17:20:14 +0100, Clive Arthur
cliveta@nowaytoday.co.uk> wrote:
On 19/04/2019 14:16, John Larkin wrote:
On Fri, 19 Apr 2019 12:53:48 +0100, Clive Arthur
cliveta@nowaytoday.co.uk> wrote:
I wanted to make a physical device to emulate a long transmission line.
This particular line has lots of C, I know the R and can guestimate the
L. So I built a lumped line using T sections, 10 Rs, 10 Ls and 9 Cs to
ground. So far so standard.
It didn't perform very well, and I think part of the reason was the
impedance being too large - dominated by the first R - so limiting the
power into the line.
So I made another, but this time using 38 Cs and a long helix of
resistance wire wound on a plastic pipe to provide the R and L. It
measures quite close to the other in terms of R, L & C, but performs
much better.
I'm guessing that the reasons for this include the impedance issue, but
maybe also because the L is now one long tapped inductor, ie coupled and
no longer discrete. To my mind, that seems closer to a real line. Is
that a valid assumption?
In addition, simulating (different - we use these a lot) lumped models
using LTspice always shows worse performance than the provided LTRA
model with the same RLC. Is this a similar effect?
Cheers
Simulate a lossless line with just Ls and Cs... no Rs.
A discrete LC line tends to ring on a fast edge. The number of LC
sections grows as the square of Tr/Td, which gets ugly fast.
It needs R, it's far from lossless and carries significant power too.
(In fact most people in this business don't bother with the Ls, but I
want a better emulation.)
So the question is, would a resistive inductor tapped with multiple Cs
be closer to a real line than multiple discrete RLC stages? I can't
test a real line.
Cheers
What's the physics that you'd like to emulate?
If you can't test a real line, the sim will be a guess.
Maybe the trans atlantic cable :^)
George H.
They are all fiber now!
On Friday, April 19, 2019 at 11:45:28 AM UTC-7, Clive Arthur wrote:
would you expect that a resistive inductor tapped with multiple Cs would
make a better model than multiple discrete RLC stages?
NEVER gonna make that work. An inductor with taps is a MUTUAL
inductor, you need decoupled ones. Looks similar on a diagram,
but not at all the same.
Isn't that what a real line is? I mean, continuous, not made of sections?
A discrete LC line tends to ring on a fast edge. The number of LC
sections grows as the square of Tr/Td, which gets ugly fast.
It needs R, it's far from lossless and carries significant power too.
(In fact most people in this business don't bother with the Ls, but I
want a better emulation.)
That corresponds to the classical LCL sub-segment design of aOn 2019-04-19, Clive Arthur <cliveta@nowaytoday.co.uk> wrote:
A discrete LC line tends to ring on a fast edge. The number of LC
sections grows as the square of Tr/Td, which gets ugly fast.
It needs R, it's far from lossless and carries significant power too.
(In fact most people in this business don't bother with the Ls, but I
want a better emulation.)
I think the incuctors should be discrete, not coupled.
On 2019-04-19, Clive Arthur <cliveta@nowaytoday.co.uk> wrote:
A discrete LC line tends to ring on a fast edge. The number of LC
sections grows as the square of Tr/Td, which gets ugly fast.
It needs R, it's far from lossless and carries significant power too.
(In fact most people in this business don't bother with the Ls, but I
want a better emulation.)
I think the incuctors should be discrete, not coupled.
The L was guestimated from the C and an assumed propagation velocity, IOn 21.4.19 07:43, Jasen Betts wrote:
On 2019-04-19, Clive Arthur <cliveta@nowaytoday.co.uk> wrote:
A discrete LC line tends to ring on a fast edge. The number of LC
sections grows as the square of Tr/Td, which gets ugly fast.
It needs R, it's far from lossless and carries significant power too.
(In fact most people in this business don't bother with the Ls, but I
want a better emulation.)
I think the incuctors should be discrete, not coupled.
That's right. The telegraph equations assume separate LC elements,
with constant L/C ratio and infitesimally small elements.
The OP never indicated if the original test had the proper
L/C ratio (being square of the characteristic impedance).
On 21/04/2019 08:17, Tauno Voipio wrote:
On 21.4.19 07:43, Jasen Betts wrote:
On 2019-04-19, Clive Arthur <cliveta@nowaytoday.co.uk> wrote:
A discrete LC line tends to ring on a fast edge. The number of LC
sections grows as the square of Tr/Td, which gets ugly fast.
It needs R, it's far from lossless and carries significant power too.
(In fact most people in this business don't bother with the Ls, but I
want a better emulation.)
I think the incuctors should be discrete, not coupled.
That's right. The telegraph equations assume separate LC elements,
with constant L/C ratio and infitesimally small elements.
The OP never indicated if the original test had the proper
L/C ratio (being square of the characteristic impedance).
The L was guestimated from the C and an assumed propagation velocity, I
can't actually measure it. Then the Z0 is calculated, again, it can't
be measured.
I know the usual method is separate LCR sections, but surely in reality
the L is not discrete sections, neither is the C of course, but there's
nothing I can do about that. In this particular case, the L and R can
be readily made with a long air-cored helix of resistance wire and there
are multiple Cs at taps.
Again, I know this isn't the usual method, and I suppose for most lines
it would be more difficult to do it this way, but why would this be a
worse approximation to reality than separate sections?
Cheers
If it's coax, with conductors and dielectrics, and you know the dimensionsOn 21/04/2019 08:17, Tauno Voipio wrote:
On 21.4.19 07:43, Jasen Betts wrote:
On 2019-04-19, Clive Arthur <cliveta@nowaytoday.co.uk> wrote:
A discrete LC line tends to ring on a fast edge. The number of LC
sections grows as the square of Tr/Td, which gets ugly fast.
It needs R, it's far from lossless and carries significant power too.
(In fact most people in this business don't bother with the Ls, but I
want a better emulation.)
I think the incuctors should be discrete, not coupled.
That's right. The telegraph equations assume separate LC elements,
with constant L/C ratio and infitesimally small elements.
The OP never indicated if the original test had the proper
L/C ratio (being square of the characteristic impedance).
The L was guestimated from the C and an assumed propagation velocity, I
can't actually measure it. Then the Z0 is calculated, again, it can't
be measured.
I know the usual method is separate LCR sections, but surely in reality
the L is not discrete sections, neither is the C of course, but there's
nothing I can do about that. In this particular case, the L and R can
be readily made with a long air-cored helix of resistance wire and there
are multiple Cs at taps.
Again, I know this isn't the usual method, and I suppose for most lines
it would be more difficult to do it this way, but why would this be a
worse approximation to reality than separate sections?
Cheers
--
Clive
On Sunday, April 21, 2019 at 5:18:36 AM UTC-4, Clive Arthur wrote:
On 21/04/2019 08:17, Tauno Voipio wrote:
On 21.4.19 07:43, Jasen Betts wrote:
On 2019-04-19, Clive Arthur <cliveta@nowaytoday.co.uk> wrote:
A discrete LC line tends to ring on a fast edge. The number of LC
sections grows as the square of Tr/Td, which gets ugly fast.
It needs R, it's far from lossless and carries significant power too.
(In fact most people in this business don't bother with the Ls, but I
want a better emulation.)
I think the incuctors should be discrete, not coupled.
That's right. The telegraph equations assume separate LC elements,
with constant L/C ratio and infitesimally small elements.
The OP never indicated if the original test had the proper
L/C ratio (being square of the characteristic impedance).
The L was guestimated from the C and an assumed propagation velocity, I
can't actually measure it. Then the Z0 is calculated, again, it can't
be measured.
I know the usual method is separate LCR sections, but surely in reality
the L is not discrete sections, neither is the C of course, but there's
nothing I can do about that. In this particular case, the L and R can
be readily made with a long air-cored helix of resistance wire and there
are multiple Cs at taps.
Again, I know this isn't the usual method, and I suppose for most lines
it would be more difficult to do it this way, but why would this be a
worse approximation to reality than separate sections?
Cheers
--
Clive
If it's coax, with conductors and dielectrics, and you know the dimensions
then you should be able to guesstimate C and L and R.
And the coupling... which I'd guess to be small, unless there is some
magnetic material involved.
George H.
I wanted to make a physical device to emulate a long transmission line.
This particular line has lots of C, I know the R and can guestimate the
L. So I built a lumped line using T sections, 10 Rs, 10 Ls and 9 Cs to
ground. So far so standard.
It didn't perform very well, and I think part of the reason was the
impedance being too large - dominated by the first R - so limiting the
power into the line.
So I made another, but this time using 38 Cs and a long helix of
resistance wire wound on a plastic pipe to provide the R and L. It
measures quite close to the other in terms of R, L & C, but performs
much better.
I'm guessing that the reasons for this include the impedance issue, but
maybe also because the L is now one long tapped inductor, ie coupled and
no longer discrete. To my mind, that seems closer to a real line. Is
that a valid assumption?
In addition, simulating (different - we use these a lot) lumped models
using LTspice always shows worse performance than the provided LTRA
model with the same RLC. Is this a similar effect?
Cheers
On 22.4.19 02:28, George Herold wrote:
On Sunday, April 21, 2019 at 5:18:36 AM UTC-4, Clive Arthur wrote:
On 21/04/2019 08:17, Tauno Voipio wrote:
On 21.4.19 07:43, Jasen Betts wrote:
On 2019-04-19, Clive Arthur <cliveta@nowaytoday.co.uk> wrote:
A discrete LC line tends to ring on a fast edge. The number of LC
sections grows as the square of Tr/Td, which gets ugly fast.
It needs R, it's far from lossless and carries significant power too.
(In fact most people in this business don't bother with the Ls, but I
want a better emulation.)
I think the incuctors should be discrete, not coupled.
That's right. The telegraph equations assume separate LC elements,
with constant L/C ratio and infitesimally small elements.
The OP never indicated if the original test had the proper
L/C ratio (being square of the characteristic impedance).
The L was guestimated from the C and an assumed propagation velocity, I
can't actually measure it. Then the Z0 is calculated, again, it can't
be measured.
I know the usual method is separate LCR sections, but surely in reality
the L is not discrete sections, neither is the C of course, but there's
nothing I can do about that. In this particular case, the L and R can
be readily made with a long air-cored helix of resistance wire and there
are multiple Cs at taps.
Again, I know this isn't the usual method, and I suppose for most lines
it would be more difficult to do it this way, but why would this be a
worse approximation to reality than separate sections?
Cheers
--
Clive
If it's coax, with conductors and dielectrics, and you know the dimensions
then you should be able to guesstimate C and L and R.
And the coupling... which I'd guess to be small, unless there is some
magnetic material involved.
George H.
Exactly. Due to the inverse logarithmic relation of the diameter ratio,
the impedance cannot be very far from the customary 50 ohms.
On Mon, 22 Apr 2019 13:59:00 +0300, Tauno Voipio
tauno.voipio@notused.fi.invalid> wrote:
On 22.4.19 02:28, George Herold wrote:
On Sunday, April 21, 2019 at 5:18:36 AM UTC-4, Clive Arthur wrote:
On 21/04/2019 08:17, Tauno Voipio wrote:
On 21.4.19 07:43, Jasen Betts wrote:
On 2019-04-19, Clive Arthur <cliveta@nowaytoday.co.uk> wrote:
A discrete LC line tends to ring on a fast edge. The number of LC
sections grows as the square of Tr/Td, which gets ugly fast.
It needs R, it's far from lossless and carries significant power too.
(In fact most people in this business don't bother with the Ls, but I
want a better emulation.)
I think the incuctors should be discrete, not coupled.
That's right. The telegraph equations assume separate LC elements,
with constant L/C ratio and infitesimally small elements.
The OP never indicated if the original test had the proper
L/C ratio (being square of the characteristic impedance).
The L was guestimated from the C and an assumed propagation velocity, I
can't actually measure it. Then the Z0 is calculated, again, it can't
be measured.
I know the usual method is separate LCR sections, but surely in reality
the L is not discrete sections, neither is the C of course, but there's
nothing I can do about that. In this particular case, the L and R can
be readily made with a long air-cored helix of resistance wire and there
are multiple Cs at taps.
Again, I know this isn't the usual method, and I suppose for most lines
it would be more difficult to do it this way, but why would this be a
worse approximation to reality than separate sections?
Cheers
--
Clive
If it's coax, with conductors and dielectrics, and you know the dimensions
then you should be able to guesstimate C and L and R.
The hard part is the R calculation, since after all, the EM field
propagates in the dielectric between the outside of the center
connector inside of the shield. Thus the insulation material losses
are material and frequency dependent. Even when air is used as
dielectric, you have to consider skin dept of the conductor, which is
also material and frequency dependent.
And the coupling... which I'd guess to be small, unless there is some
magnetic material involved.
George H.
Exactly. Due to the inverse logarithmic relation of the diameter ratio,
the impedance cannot be very far from the customary 50 ohms.
Commercial coaxial cables are available to at least 93 ohms.
On 22.4.19 15:12, upsidedown@downunder.com wrote:
On Mon, 22 Apr 2019 13:59:00 +0300, Tauno Voipio
tauno.voipio@notused.fi.invalid> wrote:
On 22.4.19 02:28, George Herold wrote:
On Sunday, April 21, 2019 at 5:18:36 AM UTC-4, Clive Arthur
wrote:
On 21/04/2019 08:17, Tauno Voipio wrote:
On 21.4.19 07:43, Jasen Betts wrote:
On 2019-04-19, Clive Arthur <cliveta@nowaytoday.co.uk
wrote:
A discrete LC line tends to ring on a fast edge. The
number of LC sections grows as the square of Tr/Td, which
gets ugly fast.
It needs R, it's far from lossless and carries significant
power too. (In fact most people in this business don't
bother with the Ls, but I want a better emulation.)
I think the incuctors should be discrete, not coupled.
That's right. The telegraph equations assume separate LC
elements, with constant L/C ratio and infitesimally small
elements.
The OP never indicated if the original test had the proper
L/C ratio (being square of the characteristic impedance).
The L was guestimated from the C and an assumed propagation
velocity, I can't actually measure it. Then the Z0 is
calculated, again, it can't be measured.
I know the usual method is separate LCR sections, but surely
in reality the L is not discrete sections, neither is the C of
course, but there's nothing I can do about that. In this
particular case, the L and R can be readily made with a long
air-cored helix of resistance wire and there are multiple Cs
at taps.
Again, I know this isn't the usual method, and I suppose for
most lines it would be more difficult to do it this way, but
why would this be a worse approximation to reality than
separate sections?
Cheers
--
Clive
If it's coax, with conductors and dielectrics, and you know the
dimensions then you should be able to guesstimate C and L and
R.
The hard part is the R calculation, since after all, the EM
field propagates in the dielectric between the outside of the
center connector inside of the shield. Thus the insulation
material losses are material and frequency dependent. Even when
air is used as dielectric, you have to consider skin dept of the
conductor, which is also material and frequency dependent.
And the coupling... which I'd guess to be small, unless there
is some magnetic material involved.
George H.
Exactly. Due to the inverse logarithmic relation of the diameter
ratio, the impedance cannot be very far from the customary 50
ohms.
Commercial coaxial cables are available to at least 93 ohms.
In my vocabulary, 93 ohms is not very far from 50 ohms, SWR not
even 2:1.
The cable has an extermely thin inner conductor. IBM liked to
use it as network cable.
* Propagation velocity and Zo CAN be measured, maybe not as accuratelyOn 21/04/2019 08:17, Tauno Voipio wrote:
On 21.4.19 07:43, Jasen Betts wrote:
On 2019-04-19, Clive Arthur <cliveta@nowaytoday.co.uk> wrote:
A discrete LC line tends to ring on a fast edge. The number of LC
sections grows as the square of Tr/Td, which gets ugly fast.
It needs R, it's far from lossless and carries significant power too.
(In fact most people in this business don't bother with the Ls, but I
want a better emulation.)
I think the incuctors should be discrete, not coupled.
That's right. The telegraph equations assume separate LC elements,
with constant L/C ratio and infitesimally small elements.
The OP never indicated if the original test had the proper
L/C ratio (being square of the characteristic impedance).
The L was guestimated from the C and an assumed propagation velocity, I
can't actually measure it. Then the Z0 is calculated, again, it can't
be measured.
* Why in the heck use resistive wire, that is counter-productive; addsI know the usual method is separate LCR sections, but surely in reality
the L is not discrete sections, neither is the C of course, but there's
nothing I can do about that. In this particular case, the L and R can
be readily made with a long air-cored helix of resistance wire and there
are multiple Cs at taps.
Again, I know this isn't the usual method, and I suppose for most lines
it would be more difficult to do it this way, but why would this be a
worse approximation to reality than separate sections?
Cheers
Exactly. Due to the inverse logarithmic relation of the diameter ratio,
the impedance cannot be very far from the customary 50 ohms.
Clive Arthur wrote:
I know the usual method is separate LCR sections, but surely in
reality the L is not discrete sections, neither is the C of course,
but there's nothing I can do about that. In this particular case, the
L and R can be readily made with a long air-cored helix of resistance
wire and there are multiple Cs at taps.
* Why in the heck use resistive wire, that is counter-productive; adds
loss, adds noise, and just plain silly.
On 22/04/2019 11:59, Tauno Voipio wrote:
snip
Exactly. Due to the inverse logarithmic relation of the diameter ratio,
the impedance cannot be very far from the customary 50 ohms.
That's out by over 100 times.
Cheers
On 22/04/2019 11:59, Tauno Voipio wrote:
snip
Exactly. Due to the inverse logarithmic relation of the diameter ratio,
the impedance cannot be very far from the customary 50 ohms.
That's out by over 100 times.
Cheers
On 25/04/2019 07:00, Robert Baer wrote:
snipped
Clive Arthur wrote:
I know the usual method is separate LCR sections, but surely in
reality the L is not discrete sections, neither is the C of course,
but there's nothing I can do about that. In this particular case, the
L and R can be readily made with a long air-cored helix of resistance
wire and there are multiple Cs at taps.
* Why in the heck use resistive wire, that is counter-productive; adds
loss, adds noise, and just plain silly.
I'm emulating something with resistive loss, so I figured using
resistance was the way to go. I did consider using dried Aardvark
pelts, but soon dismissed that as plain silly.
Cheers
On 25.4.19 11:04, Clive Arthur wrote:
On 22/04/2019 11:59, Tauno Voipio wrote:
snip
Exactly. Due to the inverse logarithmic relation of the diameter ratio,
the impedance cannot be very far from the customary 50 ohms.
That's out by over 100 times.
Cheers
Please locate the nearest coaxial cable impedance formula,
plug in the dimensions (for starters, you can guess epsilon-r
and mu-r both at 1), and do the calculation.
Even vith a hair as center conductor in an oli barrel, you cannot
get off 100 fold (500 milliohms or 5 kilo-ohms?).