Given a 100 ohm (differential Z) shielded twisted pair, what's a good Spice model? I can assume the shield is grounded on at least one end.
Two separate 50 ohm lines, each with grounded low sides, works as long as you assume that the signals are perfectly symmetric/balanced. But if the pair is used with unbalanced signals, the wire-wire coupling needs to be included.
A good model would get the measurable odd and even mode impedances right. Actually, for a real cable, I'd trust the data sheet for odd mode, and use TDR to measure even mode.
So would a good model be three lines, two grounded and one differential, from end to end? I guess so.
In real life, if the shield is not grounded on both ends, the shield has an impedance to the universe, and that depends on its routing and proximity to other things. Maybe I should bail on that assume the shield is grounded on both ends.
--
John Larkin Highland Technology, Inc
picosecond timing precision measurement
jlarkin att highlandtechnology dott com
http://www.highlandtechnology.com
My end has a grounded shield and, in effect, one wire of the pair grounded. The far end has a photodiode across the pair, and I'm not sure what the customer will do about the shield there. There may also be a vacuum feedthrough roughly halfway down the cable, where the shield may be grounded again. The cold wire is actually my photodiode bias, and the other wire goes to a 50 ohm resistor to ground in my box.
It all looks terribly mismatched to me, so I want to spice it to show them what might happen. They want 100 MHz bandwidth and I'm confident they can't get it. The PD has hundreds of pF shunt capacitance, and it's at the end of the 100 ohm pair, and Hobbsonian bootstraps look scary with all that transmission line in the way. "Transmission line oscillator."
50 ohms is probably not the ideal termination resistor, either.
I guess I'll sim some 3-transmission-line things until I get them to match measured odd/even mode impedances of the cable.
Grrrrr.
--
John Larkin Highland Technology, Inc
picosecond timing precision measurement
jlarkin att highlandtechnology dott com
http://www.highlandtechnology.com
How much photocurrent have they got? A transformer plus a decent 50-ohm amp can put 100 ohms at about 100K across the PD. If they have at least
170 uA they'll be in the shot noise limit.
Cheers
Phil Hobbs
--
Dr Philip C D Hobbs
Principal Consultant
ElectroOptical Innovations LLC
Optics, Electro-optics, Photonics, Analog Electronics
160 North State Road #203
Briarcliff Manor NY 10510
hobbs at electrooptical dot net
http://electrooptical.net
10s of milliamps! I'm not too concerned about shot noise.
I was considering a balun on my end, to keep the signals in the twisted-pair cable balanced. RF baluns seem like good ideas, until you see them in the time domain. RF jocks think that a couple of dB is flat.
This is a new project, so we are still ramping up towards the point of maximum confusion.
--
John Larkin Highland Technology, Inc
picosecond timing precision measurement
jlarkin att highlandtechnology dott com
http://www.highlandtechnology.com
Well, the behaviour of the photodiode will probably be pretty non-ideal at that level. You have to trade off the gross nonlinearity at low bias versus gross power dissipation in the photodiode at high bias.
Cheers
Phil Hobbs
--
Dr Philip C D Hobbs
Principal Consultant
ElectroOptical Innovations LLC
Optics, Electro-optics, Photonics, Analog Electronics
160 North State Road #203
Briarcliff Manor NY 10510
hobbs at electrooptical dot net
http://electrooptical.net
The light pulses are short, small explosions actually, low duty cycle, but I still expect some nonlinearity at 20 mA or so. Of course, that doesn't keep the customer from demanding absurd THD vs frequency in my amplifier.
With a 10 volt bias supply, 20 mA into 50 ohms makes 1 volt of signal, which immediately drops the diode voltage to 9. That has got to make nonlinearities. So a Hobbsonian bootstrap would not only increase speed, it could really help linearity.
--
John Larkin Highland Technology, Inc
lunatic fringe electronics
Twisted pair should not be modeled as a pair of independent wires that happen to be parallel, because they are magnetically coupled with one another. Any current in one wire will induce a voltage on the other wire, defeating independence.
A better appoach is to model them as a really long and thin air-core 1:1 transformer. A common-mode choke wired backwards.
RF Baluns often consist of such a long and thin air-core transformer (now deemed a transmission line) wrapped around a toroidal core. The effect is to force the wire currents to be exactly equal and opposite.
In the metrology world, a Coaxial AC Bridge is a Whetstone built with coaxial cables wound around ferrite toroids, to force shield and center conuctors to be equal and opposite, thus dramatically suppressing sesitivity to varying stray magnet fields (like AC hum). These bridges usually run at a kilohertz or so.
Me too! Don't make both splices in the same spot, and try to keep the spliced wires close to their opposites, and beyond that I dunno...
--
Tim Wescott
Control systems, embedded software and circuit design
I'm looking for work! See my website if you're interested
http://www.wescottdesign.com
I assume that having an amplifier at the diode end is a no-go. It would solve problems, though.
--
Tim Wescott
Control systems, embedded software and circuit design
I'm looking for work! See my website if you're interested
http://www.wescottdesign.com
If the shield is a good one, and if it extends to the load, then I'm not sure that the impedance to the universe matters as long as the wires are kept balanced at each end.
--
Tim Wescott
Control systems, embedded software and circuit design
I'm looking for work! See my website if you're interested
http://www.wescottdesign.com
Balun + compensation? I haven't seen a balun's time response, so I don't know how silly that idea may be.
I'm reminded of a Tektronix current probe technology (well, probably everyone's), where they'd use a Hall effect device to measure flux at DC, and then a compensated current transformer for high frequencies, with a network to cut over from one to the other somewhere in between.
I can see how it might be possible to do something like that to drive a balanced line -- something op-ampish that's good up to 1MHz or so, with the balun there to clean things up above that.
--
Tim Wescott
Control systems, embedded software and circuit design
I'm looking for work! See my website if you're interested
http://www.wescottdesign.com
On Fri, 17 Feb 2017 13:49:30 -0800, John Larkin wrote:
This looks like a working LT Spice model for a twisted pair. There are an infinite combination of line impedances that would net 100 ohms differential. Those depend on how close the pair couples versus how hard each wire couples to the shield, one limiting case being no pair coupling, namely two separate coaxes with no wire-wire coupling.
The other case is a twisted pair in free space, which still has common-mode coupling to the universe. The common-mode is a sort of goofy Goubau line, which might be ballpark 300 ohms (?) for the paralleled pair.
I assumed equal sharing, but I'll have to TDR some actual cables to tune the model.
Version 4 SHEET 1 1368 680 WIRE -960 -288 -1008 -288 WIRE -896 -288 -960 -288 WIRE -480 -288 -816 -288 WIRE -336 -288 -480 -288 WIRE -256 -288 -336 -288 WIRE -112 -288 -256 -288 WIRE 128 -288 -16 -288 WIRE 176 -288 128 -288 WIRE 208 -288 176 -288 WIRE 336 -288 288 -288 WIRE 384 -288 336 -288 WIRE 512 -288 384 -288 WIRE -112 -256 -176 -256 WIRE 32 -256 -16 -256 WIRE -480 -224 -480 -288 WIRE 32 -224 32 -256 WIRE 384 -208 384 -288 WIRE -896 -128 -896 -288 WIRE -816 -112 -816 -288 WIRE -816 -112 -848 -112 WIRE -256 -96 -256 -288 WIRE -112 -96 -256 -96 WIRE 128 -96 128 -288 WIRE 128 -96 -16 -96 WIRE -640 -80 -704 -80 WIRE -480 -80 -480 -144 WIRE -480 -80 -560 -80 WIRE -416 -80 -480 -80 WIRE -384 -80 -416 -80 WIRE 240 -80 192 -80 WIRE 384 -80 384 -128 WIRE 384 -80 320 -80 WIRE 448 -80 384 -80 WIRE -816 -64 -848 -64 WIRE -112 -64 -256 -64 WIRE 128 -64 -16 -64 WIRE 192 -48 192 -80 WIRE -480 -32 -480 -80 WIRE 384 -16 384 -80 WIRE -896 0 -896 -48 WIRE -704 0 -704 -80 WIRE 448 0 448 -80 WIRE 448 0 432 0 WIRE 512 48 512 -288 WIRE 512 48 432 48 WIRE -816 160 -816 -64 WIRE -480 160 -480 48 WIRE -480 160 -816 160 WIRE -352 160 -480 160 WIRE -256 160 -256 -64 WIRE -256 160 -352 160 WIRE -112 160 -256 160 WIRE 128 160 128 -64 WIRE 128 160 -16 160 WIRE 176 160 128 160 WIRE 208 160 176 160 WIRE 336 160 288 160 WIRE 384 160 384 64 WIRE 384 160 336 160 WIRE -176 192 -176 -256 WIRE -112 192 -176 192 WIRE 32 192 -16 192 WIRE 32 224 32 192 WIRE -400 256 -480 256 WIRE -256 256 -320 256 WIRE -176 256 -176 192 WIRE -176 256 -256 256 WIRE -480 304 -480 256 FLAG 32 -224 0 FLAG 32 224 0 FLAG 192 -48 0 FLAG -480 304 0 FLAG -704 0 0 FLAG -896 0 0 FLAG -960 -288 DIFFOUT FLAG -336 -288 P FLAG -352 160 Q FLAG 336 -288 G+ FLAG 336 160 G- FLAG -416 -80 RCM FLAG 176 -288 X FLAG 176 160 Y FLAG -256 256 K SYMBOL tline -64 -80 R0 WINDOW 0 -20 -47 Left 2 WINDOW 3 -74 52 Left 2 SYMATTR InstName Tcm SYMATTR Value Td=5n Z0=200 SYMBOL tline -64 -272 R0 WINDOW 0 -12 -97 Left 2 WINDOW 3 -82 -61 Left 2 SYMATTR InstName Tx SYMATTR Value Td=5n Z0=100 SYMBOL tline -64 176 R0 WINDOW 0 -18 -98 Left 2 WINDOW 3 -81 -54 Left 2 SYMATTR InstName Ty SYMATTR Value Td=5n Z0=100 SYMBOL res 304 -304 R90 WINDOW 0 -48 54 VBottom 2 WINDOW 3 -39 53 VTop 2 SYMATTR InstName R2 SYMATTR Value 50 SYMBOL res 304 144 R90 WINDOW 0 73 52 VBottom 2 WINDOW 3 84 51 VTop 2 SYMATTR InstName R3 SYMATTR Value 50 SYMBOL voltage 384 -224 R0 WINDOW 0 188 25 Left 2 WINDOW 3 158 76 Left 2 WINDOW 123 0 0 Left 2 WINDOW 39 170 71 Left 2 SYMATTR InstName V1 SYMATTR Value PULSE(0 1 1n 100p 100p 20n) SYMBOL e 384 -32 M0 WINDOW 0 -215 31 Left 2 WINDOW 3 -211 72 Left 2 SYMATTR InstName E1 SYMATTR Value -1 SYMBOL res 336 -96 R90 WINDOW 0 -46 58 VBottom 2 WINDOW 3 -40 56 VTop 2 SYMATTR InstName R4 SYMATTR Value 1m SYMBOL res -496 -240 R0 WINDOW 0 66 34 Left 2 WINDOW 3 64 73 Left 2 SYMATTR InstName R1 SYMATTR Value 1K SYMBOL res -496 -48 R0 WINDOW 0 67 54 Left 2 WINDOW 3 69 91 Left 2 SYMATTR InstName R5 SYMATTR Value 1K SYMBOL res -544 -96 R90 WINDOW 0 70 54 VBottom 2 WINDOW 3 79 54 VTop 2 SYMATTR InstName R6 SYMATTR Value 1m SYMBOL res -304 240 R90 WINDOW 0 70 54 VBottom 2 WINDOW 3 79 54 VTop 2 SYMATTR InstName R7 SYMATTR Value 1m SYMBOL e -896 -144 M0 WINDOW 0 72 38 Left 2 WINDOW 3 78 75 Left 2 SYMATTR InstName E2 SYMATTR Value 1 TEXT 608 256 Left 2 !.tran 100n TEXT 576 208 Left 2 ;JL Feb 18 2017 TEXT 504 160 Left 2 ;100 Ohm Twisted Pair Model
--
John Larkin Highland Technology, Inc
lunatic fringe electronics
A transformer is the LF asymptotic equivalent. If that's all you're working with, it's a good equivalent to use.
(I would like to avoid using "better" to say "more suitable", so that "better" only means a more accurate model. A simpler model may be more /suitable/ for an application.)
That's one kind. The HAMs usually call that a current balun.
Note that such an arrangement can never, ever achieve balance: the winding has a series impedance, determined by the ferrite core, and whatever resonances the winding has. At best, the impedance can peak to infinity, but we also know that's not possible without a power source. So there will always be residual current, and imbalance.
That can still be an advantage, because it gets arbitrarily close to balance, even when the load is imperfectly balanced. Which also happens often with HAM antennas!
The voltage balun does produce perfect balance, but must be made carefully. The best is a Guanella transmission line transformer, which has matched delays to both ports (taking the CT of the bal side as two ports + and -), and a double delay between those ports.
It can be further improved by stacking more transformers: a CMC (aka current balun), and a DMC (a single CT winding, made by taking a transmission line CMC and connecting one end to the other, so they add in series). This reduces the delay between +/- ports to one TLT length, minimizing LF-equivalent leakage inductance, and isolating unbal and bal ports (and either allowing the bal side to float a bit WRT ground, or nailing it to ground as hard as the CT and ground return path allows).
Or if you like, the DMC's CT provides an unbal injection point, handy for coupling to mixers.
I have two examples here:
formatting link
The input coupling (top) is a delay line (same length as the other one), and an inverter transformer (same as the DMC, CT grounded -- TL comes in with signal and ground, goes out with ground and signal. What could be simpler or more obvious? :) ). The TLs are 100 ohms each (and terminated), in parallel at the BNC, so the input is 50 ohms.
The two pot cores handle merging the balanced signals. One is a 25 ohm DMC (CT goes to +V, of course). The other is a Guanella TLT balun, matching it to 50 ohms (which is a step up, in this case).
The resulting bandwidth is limited by the output transistors, with no visible impact from the transformers. (If I had used the same lengths of windings, in a scramble wound transformer not designed with TLT theory, the impact would be siggnificant.)
On Feb 18, 2017, Tim Williams wrote (in article ):
I agree with averything you say about baluns and all, and thanks for the circuit example, but the original question was how best to coerce Spice into making a reasonable job of modelling a shielded twisted pair.
I?ve seen an electrical engineer model a shielded twisted pair as two independent wires over which one can send two unrelated signals without significant crosstalk. Digital guy, no idea of analog and transmission lines.
The other problem was that these signals had to experience the exact same delay, even though one was at baseband and the other was at 10 MHz or 80 MHz. The solution was to modulate the MHz carrier with the baseband signal, and send the result over coax.
More or less. Usually a bit higher than that, but for the same reasons, it's not a big deal: the mismatch is modest, and the length is shorter than the amplifier's bandwidth, so it's pretty good.
Sure:
formatting link
(The blue-star "parallel" windings are two strands of wire, out of a twisted triple; this gives about 50 ohms differential.)
ElectronDepot website is not affiliated with any of the manufacturers or service providers discussed here.
All logos and trade names are the property of their respective owners.