Transmission line charateristics measurement?

[touqeerazam]

Transmission line charateristics measurement??

Dear all,

I want to do cable analysis of a 3800m long twisted pair wire to set up a communication system. It is not usual twisted pair wire (telephone), has different impedance, R=194, L=4mH, C=180uF. The things to be analyzed are, characteristic impedance, bandwidth, and SNR. The questions are,

1. I have used Time Domain Reflectometry to get impedance. I used resistive load to get the impedance values, don't know if i needed a complex load for this?

2. I have used signal generator and oscilloscope to get bandwidth at different atteuations. However the attenuations are not linear with frequency, and have a damping effect. i.e certain high frequencies have high attenuation than some lower ones. Dont know why??

3.How can i measure SNR, the equitment I got is Spectrum analyzer. I know Network analyzer could make life easy but don't have it.


I will be very thankful if you guys can shed any light.

Thanks again in advance and enjoy ur time..

T

 

[jsgruszynski]

1. TDR will read off the impedance directly (assuming you are using a modern oscilloscope that has TDR built-in). The downside of TDR: it's most accurate at the instrument side of the cable and least accurate at the far terminated end, but also it's limited to the resolution of an oscilloscope. Most instrument TDRs require multi-GHz oscilloscopes that have only 8-bit ADCs for the vertical.

2. Sig Gen and Oscilloscope is doable but needlessly painful unless you have no choice. The downside: oscilloscopes generally have poor vertical (amplitude) resolution. Best for qualitative measurements; less good for quantitative measurements, especially for this kind of thing.

3. SNR isn't an issue with a passive component like a cable. SNR degradation comes in with active components. For a cable you are at the thermal noise floor right off the bat and what you are measuring with a spectrum analyzer or network analyzer is the noise floor of the instrument, not the cable. Also NWAs only measure accurately close to the characteristic impedance - which if you are already close to 50/75 ohms of the NWA all is good. However, even then, they are less accurate for impedance measurement than LCR meters.

4. Not mentioned but the best way usually: use an LCR meter. Measure the impedance with the cable open and then shorted and then plug into Zo = sqrt(|Zopen| * |Zshort|). LCR meters have the advantage of the best impedance measurement accuracy/resolution of any technique but are restricted to lower frequencies (<3 GHz with the newest Agilent models).

A good impedance measurement reference (for free) is the Agilent Impedance Measurement Handbook:

http://cp.literature.agilent.com/litweb/pdf/5950-3000.pdf

BTW the technique in #4 can be found on page 5-27. You can also measure attenuation and phase constants for the cable.

 

[touqeerazam]

Thanks 4 ur reply :)

You are right the oscilloscopes have poor resolution, I did see this limitation for TDR. Unfortunately the oscilloscope I have don't have TDR built in. I wonder if a resistive load will be ok for TDR?

The other method seems very easy, using LCR for characteristic impedance measurement. I suppose by impednace you mean Resistance? I did try to get resistance values when shorted/open ckt cable. I do get some resistance values for short ckt, however its very big for open ckt, Mega-Ohms. I don't know if you meant taking R or some other values?
Also how can we incorporate frequency dependence using LCR method since LCR meter doesn't send different frequency signals?

I will be thankful if you can put light on it :-)

Cheers
T

 

[Bob S]

A normal twisted pair has a characteristic impedance of ≈70 ohms, and a propagation velocity of ≈0.8 c (speed of light). The R=194 ohms resistance you quote is equivalent to ≈28 Ga. wire. If I use your numbers of L=4 mH and C=180uF, I calculate 1 uH/meter and 47 nF/meter, and imply a characteristic impedance of Z = sqrt(L/C) ≈5 ohms. This is much too low. An optimum impedance for a long line is ≈90 ohms.

I once built a 7000-meter long twisted wire system using a specially built 18 Ga. Z=90-ohm system which worked very well at ≈ 1 MHz bit rate.

Could you give us a detailed description of the twisted pair; wire Ga. number, insulation, and conductor-to-conductor spacing?

Could you also substitute your values of R, L, and C into formulas for lossy transmission lines and determine the inductance or capacitance per unit length, and your propagation constants? What frequency range are you interested in?

Bob S

 

[jsgruszynski]

Thanks 4 ur reply :)

You are right the oscilloscopes have poor resolution, I did see this limitation for TDR. Unfortunately the oscilloscope I have don't have TDR built in. I wonder if a resistive load will be ok for TDR?

The other method seems very easy, using LCR for characteristic impedance measurement. I suppose by impednace you mean Resistance? I did try to get resistance values when shorted/open ckt cable. I do get some resistance values for short ckt, however its very big for open ckt, Mega-Ohms. I don't know if you meant taking R or some other values?

Actually not just resistance but impedance - basically doing Ohm's Law at an AC frequency rather than DC. Resistance is DC Ohm's Law - apply a DC voltage and measure the DC current, divide voltage by current.

Impedance is AC Ohm's Law - apply an AC voltage (keep track of the magnitude and phase) and measure the AC current (keep track of the magnitude and phase relative to the voltage) and divide the complex voltage by the complex current (divide the magnitudes and subtract the phase angles). If you know what phasors are, it's phasor voltage by phasor current.

Also how can we incorporate frequency dependence using LCR method since LCR meter doesn't send different frequency signals?

I will be thankful if you can put light on it :-)

Cheers
T

This is the beauty of the using the LCR meter - you only need to take the measurement at one frequency. This doesn't address the transmission line cut-off or impedance discontinuities due to damage in the cable but if the cable is a "good cable" or perfect transmission line, you only need the two measurements at one frequency and it completely determines the characteristic impedance at all frequencies below the cut-off frequency! Cool, huh!

Since reliable use of a cable requires that it be pretty close to being "perfect" it's not a bad assumption. An LCR meter is a bad substitute for a TDR for finding point flaws in a cable, however.

The Impedance Handbook page shows the math of this - it's because you can approximate a "good" transmission line as a single LC ladder circuit which has just two impedances (hence two measurements required - the open and short temporarily eliminate each leg of ladder at a time). The rest is the consistency of physics. Transmission lines a equivalent to taking that LC ladder and dividing it into arbitrarily smaller LC ladder segments that are exact fraction scaled versions of the total LC.

This assumption does break down when you have a flaw - one or more of those infinitesimal LCs isn't just a fraction of the whole so you get a discontinuity in impedance (and a reflection).

Usually you are operating well below the cut-off frequency. That's the point where you would typically need a network analyzer to find the cut-off anyway.

Hope this helps...

 

[Bob S]

Here in thumbnail are the transmission line equations, taken from

http://en.wikipedia.org/wiki/Transmission_line

For G=0 and R small, the characteristic impedance Z is about sqrt(L/C), where L and C are the inductance and capacitance per unit length. This is usually between 60 and 90 ohms for most twisted pairs.

The attenuation constant (shown in the thumbnail) is Re(γ) = ½ (R/L) sqrt(LC) = ½(R/Z), where R is the resistance per unit length (per meter), and Z is the characteristic impedance given above. Note that Re(γ) has units meters^-1. For a given R, increasing the characteristic impedance Z increases the attenuation length 2Z/R. Using these equations should characterize the twisted pair fairly well.

Bob S

 

[touqeerazam]

Thanks Guys, your replies are very useful :-)

Sorry for a bit of correction, C=1.8uF not C=180uF. It is a 7-core wire but I am using 2-cores as twisted pair. The capacitance is calculated with respect to 6-cores+an armour on these cores. Velocity factor is 0.66, it is #22 AWG, each core is .76mm diameter, insulation 1.47mm, core-core distance is 1.47mm.

@ jsgruszynski : I am using a resistor in series with twisted pair wire to get I=(V2-V1)/R and then Z=V1/I, and looking at amplitude phase changes on oscillscope. It will be a convenient method to do impedance measurement if I can do it right :)

@Bob : The impedance comes to be 46 Ohms from sqrt(L/C) and I got similar value using TDR, 49Ohms. The capacitance and resistance values are given by manufacturer to be 190pF/m, 54.5ohm/Km respectively. However LCR readings are bit different R=194, C=1.8uF, L=4mH. Not sure of the bandwidth and looking to to use -30dB to -40dB frequencies. I guess it would mean 100-200Khz range. I see damped behavior in frequency vs. amplitude plot using signal generator/oscilloscope for attenuation measurements, dnt know why.

I will be thankful if you could post some comments on bandwidth measurement/SNR as well.

Thanks, Have a lovely day every1 :-)

T

 

[Bob S]

I forgot to mention in post #6 that the R in the attenuation equation is actually

R(ω) = R_dc + σ·√ω,

where the first term is the dc wire resistance, and the second term is due to skin effect copper losses, probably important above a few 100 kHz for 22 Ga. wire. R(ω) is in phase quadrature with jωL. The constant σ depends on the wire diameter ond on the copper conductance.

Ordinary 7-core wire is a very poor choice of twisted pair, because the signal will couple to other conductors in the core. Usually, twisted pairs are individually twisted pairs in an aluminum foil shield with a drain wire.

Bob S

 

[touqeerazam]

Thanks for your reply Bob :-)

Yeah 7-core wire is good for mechanical strength nt for transmission, but I am using it for transmission purposes as well.

Ta, T