Modelling tranmission lines with discrete componets

[EternityForest]

So heres a thought:

Transmisson lines are often modeled as a long string of two port circuits each acting like a section of cable. My question is could you build a buch of copies of the elementary transmission line unit in the picture
Telegrapher's equations - Wikipedia, the free encyclopedia
and string them together? and have it be like a simulator in hardware?
would this be accurate fir medium frequency(1-200khz) signals across
100-1000 feet of simulated cable?

I am developing a protocol based on asynchronous serial and a way to simulate cable would be nice to have.

 

[MrAl]

So heres a thought:

Transmisson lines are often modeled as a long string of two port circuits each acting like a section of cable. My question is could you build a buch of copies of the elementary transmission line unit in the picture
Telegrapher's equations - Wikipedia, the free encyclopedia
and string them together? and have it be like a simulator in hardware?
would this be accurate fir medium frequency(1-200khz) signals across
100-1000 feet of simulated cable?

I am developing a protocol based on asynchronous serial and a way to simulate cable would be nice to have.

Hi,

You might be able to get away with that if you could use enough sections. I guess if you had a certain inductance per inch and capacitance per inch and resistance per inch you could do it inch by inch
The equations are developed by letting distances approach zero, but there are simpler equations than those partials shown in Wikipedia, probably on the web somewhere. The simpler equations show how the voltage at the far end behaves due to a signal at the input end. I think that is what you are after. If you cant find these equations i'll try to find them also.

 

[ericgibbs]

hi,
A pdf showing some line loss equations.

 

[JimB]

Some years ago I worked for a company which supplied various bits of equipment which were connected together by long cables (long = 5 to 100km, depending on the project).
To test the various parts of the system in the factory before delivery to site, cable simulators were used.
These simulators were simply sections of R,L and C, wired up to represent the cable.
So if the real cable was 10km long with characteristics of say* 10ohm, 1mH and 0.1uf per km, the simulator would be built with 10 sections each of 10ohm, 1mH and 0.1uf. Or, sometimes with just 5 sections with 20ohm, 2mH and 0.2uf per section.
Always seemed to work OK for what we were trying to do.

* these are just wild guess values for illustration, not typical values for the real cable.


JimB

 

[Diver300]

1000 ft? That's one reel of cable, isn't it? Just buy a reel and do your tests with real cable not simulated cable.

At 200 kHz you could start to get into signal bounce problems if you terminate badly on 1000 ft of cable, but you would be unlucky. I would go with JimB's suggestion, and you certainly don't need to go to more than 5 sections. As long as you use an oscilloscope to make sure that the signals are clean.

If you use a higher frequency signal, then you need more sections. If you put 10 MHz signals into a cable, the wavelength is about 60 ft, so you need at least 1 section for every 30ft of simulated cable, and probably more. At 200 kHz, the wavelength is 3000 ft or so.

You should also limit the rise and fall times of your signals so that you don't introduce high frequency noise. You should also filter the inputs to ignore higher frequency signals.

 

[EternityForest]

Thanks guys. a lot of maxims ICs have built in slew rate limiting capacitors.
I wonder what happens if you put RS485 in a star topology and don't terminate anything and do all the things they say not to?
will it still work if the speed is low enough? but rs485 has a max of 256 devices even with maxims high impedance chips. would leaving off the termination and using speeds to low for reflections allow more devices because there wouldn't be as much load?
if this whole thing is just impossible i suppose 9600 baud is enough.

 

[Diver300]

RS485 should have a pull-up on one line and a pull-down on the other so that the receivers go into a known state when nothing is driving the network.

If you don't have termination, reflections will die out after 3 or 4 bounces at worst. If you can afford the time, that may not matter. The only thing you need to be aware of is that the receivers may trigger on the reflections, so you may get several pulses of each edge that is transmitted. You may need to filter those out, but it depends on the receivers.

Serial receivers vary. Some will ride noisy edges, especially if you fix the baud rate. Others have a very low tolerance for edges arriving at the wrong time. I was fault-finding one system, running at a low baud rate. On the oscilloscope it looked perfect, and a serial port on a PC read exactly what I though was being sent. However, the actual receiver missed characters a lot of the time. The problem was too much variation in the length of the bits, although the variations were only a couple of percent of the bit period. When I tightened up the edge timing, the system worked fine.

 

[EternityForest]

Thank you so much Diver!

I was actually planning to specify a failsafe RS485 chip like the ones maxim makes so that it would be ok to leave off the pullups and downs.

since this protocol is mostly for lighting and interactive installations i think i could probably just run it at low speed to avoid reflection mess. i was also planning to use checksums and confirmation packets for everything important so a little instability is not much of an issue as long as some data actually can get through. If i actually complete this project i intend to OpenSource it totally under GPL or similar. Thanks for the help guys!