Measuring Characteristic Impedance

[shedeed]

characteristic impedance+twisted pair+formula

I have a CAT5 cable am using it to send and receice between two transcivers through RS485.
the cable's characteristc impedance of the cable is not 120 OHM, so it doesn't match to the loads.
Now i want to maych the loads, i want to have a technique to measure the impedance of the cable.

 

[jallem]

19200 baud termination impedance

there is several techniques for impedance measurement.
one that is useful in RF is using a TDR. A TDR can be
built using only a fast pulse generator and an oscope.
the pulse is send trough the cable and is reflected if
there is a mistmatching. the amount of mismatching
is an indication of the impedance against the reference,
in this case the pulse generator output impedance.

Added after 21 minutes:

The impedance of RS485 is 120 OHM, I took the following
statement from **broken link removed**

Characteristic Impedance of Twisted-Pair Wire

Depending on the geometry of the cable and the materials used in the insulation, twisted-pair wire will have a "characteristic impedance" associated with it that is usually specified by its manufacturer. The RS-485 specification recommends, but does not specifically dictate, that this characteristic impedance be 120 ohms. Recommending this impedance is necessary to calculate worst-case loading and common-mode voltage ranges given in the RS-485 specification. The specification probably does not dictate this impedance in the interest of flexibility. If for some reason 120-ohm cable cannot be used, it is recommended that the worst-case loading (the number of transmitters and receivers that can be used) and worst-case common-mode voltage ranges be recalculated to make sure the system under design will work. Publication TSB89 has a section specifically devoted to such calculations.

 

[xxargs]

phase shift cat5 cable

there is several techniques for impedance measurement.
one that is useful in RF is using a TDR. A TDR can be
built using only a fast pulse generator and an oscope.
the pulse is send trough the cable and is reflected if
there is a mistmatching. the amount of mismatching
is an indication of the impedance against the reference,
in this case the pulse generator output impedance.

Added after 21 minutes:

The impedance of RS485 is 120 OHM, I took the following
statement from h**p://www.maxim-ic.com/appnotes.cfm/appnote_number/763

Characteristic Impedance of Twisted-Pair Wire

Depending on the geometry of the cable and the materials used in the insulation, twisted-pair wire will have a "characteristic impedance" associated with it that is usually specified by its manufacturer. The RS-485 specification recommends, but does not specifically dictate, that this characteristic impedance be 120 ohms. Recommending this impedance is necessary to calculate worst-case loading and common-mode voltage ranges given in the RS-485 specification. The specification probably does not dictate this impedance in the interest of flexibility. If for some reason 120-ohm cable cannot be used, it is recommended that the worst-case loading (the number of transmitters and receivers that can be used) and worst-case common-mode voltage ranges be recalculated to make sure the system under design will work. Publication TSB89 has a section specifically devoted to such calculations.

good papers...

But remeber 100 - 120 Ohm real impedance is only valid for frequecy higher than 200 KHz and if you communicate for example 19200 Baud on CAT-5, your equipment see different and complex impedances and possible need complex capacitive terminator and generator impedances to absorb reflexes on long cable.

(is very well know problem in telephone world...)


formula for impedanse on cable is: [1]


Z=sqrt((r+jwl) / (g + jwc))

r = resistance
g = conductance
w = 2* PI*frequency
l = inductance
c = capacitance
j = imaginay number j^2 = -1

and typical value for CAT-5 cable (Z=100 Ohm on 1 MHz) is:

c = 50 nF/km
r = 188 Ohm/km
l = 0.50 mH/km
g = 1E-7 S/km (frequency depend - incrase slowly on higher frequency depend of skin effect on copper and dielectric loss on insulator)



formula [1] is easy to handle if you have complex capable calculators (try free42 - is simulator for hp42s and very easy handle with complex numbers if you know handle HP RPN and stack operation.)


simplified form for low frequency ( DC ~= 20000 Hz):

Z = sqrt( r / (jwc))

and result always complex capacitive impedanses value and phase angle -45 degree (current going 45 degree before voltage)

Impedances going from 1327 ohm on 300 Hz, 733 Ohm on 1000 Hz to 419 Ohm on 3400 Hz on CAT-5 cable, value changes dramatic depend of frequence and always -45 degree phase shift... (if using ekv. [1], phase shift going from 44.9 degree on 300 Hz to 40 degree on 3400 Hz...)

simplified form for high frequency (>> 200 KHz):

Z = sqrt(l/c)

and impedances is reell and for CAT-5 around 100 - 120 Ohm depend of manufactory.

---

frequency between 20 Khz and 200 KHz is tricky part depend impedances going from complex to reell impedances and not so easy to make terminator with same character as cable for reflex free termination (on old telegraph era using very, very complex terminator to handle telegraph keying from DC or part of Hertz to 30 Hertz on cable to suppress reflexes ... telephone frequency range as 300 -3400 Hz is much easy to handle compare to telegraph, but still need complex terminators - 2 Mbit on CAT 5 is very easy to handle and only need resistiv terminator...)).


if you communicate on 1-2 Mbit Baud burst in datapackage, you not have truble if using 100-120 Ohm terminate - but if using 19200 Baud burst on long cable, you have possibly trouble without complex terminate - if sending 11000000001100000000... (10 bit per byte) on 19200 Baud, is strongest working frequency energy for this pattern is around 2, 4, 6, 8... KHz and definitly in audio band and possibly need (very) complex terminate and sendning generator impedances!!.


---

Use always wastable train sequence and sync before importent data on every data package-burst

- if you sending data package 1 time/sek - you lowest working frequency in system
is 1 Hz and need in 'principe' complex terminate, depend of 'DC' step-function makings from starting of sending datapackage-burst have frequency from 1 Hz and up, and all frequency together in step-funktion can (but not always) ruin first bits on package if cable is not proper terminated from just 1 Hz and up whole way to highest usable frequency...

but if you can tolerate waste few bits in beginning on every package burst, you only need make correct termination for data:s bandwidth and working frequency. ie if using 2 Mbit, need only make valid termination from 200 kHz and up and reell impedances as 100 - 120 Ohm on CAT-5...

if using 2 Mbit-burst and usable data bandwith not going lower than 200 KHz ,you can also use transformer to break up galvanic path between units if you have problem on ground currents and other saftey reason etc. - but you must build data protocoll to ignore depends of DC-levels and low frequency noise from RS485-unit and only listen for correct training (possible even with wrong polarisation) and sync sequence before reading data from RS485-circurits on you equipment - and make this fool-proof!!!.


---

Is not only physical solution of cable decise if you build easy and robust network or weak and sensivity network, selected right format of datapackage, baudrate, syncing and crambling (ie data bandwidth and frequency) is also importent if system only need 'easy' terminators or need complex dito and high know skills people to building _working_ network!!!

Remeber IDE-bus have only maximal capacity of 166 MB/s with special expensive 0.3 m short high density ribbon cable, designed with no respect to transmission theory, compare to SCSI have twice bandwidth on only 8 data pair and working on couple of meter of cheap cable, depend of respect of transmission theory and make usable compromisses on harware and protocol for it, and using terminators...
- exampe of 'weak' solution (IDE) to 'robust 'solution (SCSI).


- ie. don't use low baudrate data protocol as RS485 and 19200 - 115200 baud on long cable and/or complex network if possible... if not want and _can_ sell very much fault search and trouble shooting bills for strange problems to customer...


(i hope this text is understudable - i not native english and i know written bad english... But I try to write this text! )


/Xxargs