Modeling a Twisted pair transmission line

[Brocktune]

Is this the best way to do it?

**broken link removed**

And you can't find the values of C and L unless you look them up in the cable manufacturer's specifications?

 

[Roff]

What simulator are you using? How long is the line? What is the highest frequency (or risetime) of interest?
It appears that you are looking at
"The Twisted-Pair Telephone Transmission Line", so you know that the lumped model needs to be symmetrical.
I looked at Belden 1700R Cat 5e as an example, and came up with Z0=100 ohms, and capacitance=15pF/ft=1.25pF/in. Armed with that and the knowledge that Z0=√(L/C), then L=12.5nH/in (you need L/2 on each side of the transmission line). The loss components are more difficult if not explicitly specified.
I'm not sure I answered even part of your question.

 

[Brocktune]

Actually, you've helped me think about a few other things. And yes, I'm reading that same document

My rise times are pretty fast, but signal frequency is extremely low. I'm receiving ~300ms sqaure waves, over a differential line. Duty cycle of the signal is extremely low. 300ms signal, not repeated for another 10 seconds.

Right now I'm using PSpice, just trying to find out more info about the cables I'm using so I can insert some L's and C's.

 

[Roff]

If you are actually concerned about signal integrity, and you don't have a high data rate and/or narrow pulses, put a big (1k) resistor in series with each side of the line. This will guarantee no ringing at the other end, you won't need to terminate the line, and your risetimes will be slow, but do you care?

 

[red4925]

I just built a line simulator; it modeled 12km of cable. I didn't design the model myself; i'm just an intern and still in school. But if you have a decent impedence analyzer, you could measure the parameters directly from the cable. Or else consult the cable manufacturer and break open an EM text.
You need to make sure your circuit models a length of cable that is less than 1/10 your wavelength (of your operating frequency).

 

[Brocktune]

Unfortunately I do need my levels to rise quite quickly. I'm not concerned with frequency per say, but I do need a minimum rise time of about 20ms or as quickly as possible.

Thanks for all the info guys, I think I'm actually just going to hook up the comm. system, pump some signals through and keep tweaking my balancing/termination stuff until I get my rise time and the least amount of over/undershoot. A little more Edison than Tesla in the technique, but it will work

 

[Nigel Goodwin]

Unfortunately I do need my levels to rise quite quickly. I'm not concerned with frequency per say, but I do need a minimum rise time of about 20ms or as quickly as possible.

Well I wouldn't describe 20mS as a fast rise time - that's actually only a 25Hz sinewave from peak to peak!.

 

[Brocktune]

Unfortunately I do need my levels to rise quite quickly. I'm not concerned with frequency per say, but I do need a minimum rise time of about 20ms or as quickly as possible.

Well I wouldn't describe 20mS as a fast rise time - that's actually only a 25Hz sinewave from peak to peak!.

That is true... 25msec is the longest I can deal with, but the signals are driving inputs of a very high speed digital section of my design, and I need them to rise as fast as possible. They will be generated (at the driver) at less than 100usec rise time, so I'd like to be on par with that at the receiver. Even though that's still pretty slow.

 

[Roff]

Unfortunately I do need my levels to rise quite quickly. I'm not concerned with frequency per say, but I do need a minimum rise time of about 20ms or as quickly as possible.

Well I wouldn't describe 20mS as a fast rise time - that's actually only a 25Hz sinewave from peak to peak!.

That is true... 25msec is the longest I can deal with, but the signals are driving inputs of a very high speed digital section of my design, and I need them to rise as fast as possible. They will be generated (at the driver) at less than 100usec rise time, so I'd like to be on par with that at the receiver. Even though that's still pretty slow.

You can put 1k in series with each side of the cable on the driving end and still get 100us risetimes at the receiving end, even with 1000 feet of cable. Shorter lengths will give you faster risetimes.
For best signal integrity, you would want to terminate the recieving end with a 100 ohm differential termination (which can optionally be center tapped to GND or a voltage), but with such loose risetime requirements, this may put an unnecessary load on your driver.

 

[pebe]

I don't know why you need to model a transmision line. If you correctly terminate the line with a resistive load equal to its characteristic impedance, then you will get out of its end the waveform you put into it. The input of the cable will then look resistive and if you like you can add series resistors as Ron said, but those will only attenuate the signal.

That said, any transmission line has an attenuation that increases with increasing frequency, and that will affect the amplitude of the signal at the end.

If you look up the manufacturer's data on the transmission cable, you will get its characteristic impedance, capacitance and attenuation per unit length and, in the case of RF coax, its velocity factor.