[Bertho] sent in a great tutorial on terminating transmission lines. If you’ve ever tried to send a high frequency signal a long way down a wire, you know the problems that can crop up due to electronic strangeness. Luckily [Bertho]’s tutorial explains just about everything, from where and when to terminate a cable and why signals get screwed up in long wires.
[Bertho] begins his lesson by taking two oscilloscopes and 20 m of CAT5 cable with the twisted pairs wired in series to make an 80 meter long transmission line. A ~100kHz square wave was sent down the cable after being displayed on the first oscilloscope, and picked up on the other end by the second oscilloscope. It’s a great way to show the changes in a signal over a long cable run, and how small changes in the circuit (just adding a simple resistor) can affect the signal coming out of a cable.
It’s a great post that demystifies the strange electrical gremlins that pop up when you’re running a length of wire. Great job, [Bertho].
Great information. Check also on his site the excellent tutorial on the use of decoupling capacitors, it will save you a lot of time, and the page about the Instructables site TOS which basically explains why you should avoid at any cost submitting your projects to them. Most of us already knew, but many newbies don’t.
Resonance need not be a problem, consider the works of Tesla. I’m now wondering what it would take to make a linear Tesla “coil” if the wire were long enough you wouldn’t need a primery. If it could be made to oscillate at it’s resonant frequency and used as a receiver it could make for an interesting device to capture solar wind.
Just a thought. Any who this made for an interesting read. Reminds me of the old computer delay line memory.
(quartz crystal memory from old tv’s and vcr’s made for a somewhat elegant solution compaired to some of the other methods) I’m guessing their resonant frequencies might be a tad lower :D
Hmm… Next to make a Tesla coil that opperates in the visible spectrum… Or maybe gamma to pump a maser…
-jokes. A project for another afternoon (:
I quite enjoy posts that lead one off onto a tangent (:
are all those inverters really necessary? cant you just use one npn-pnp pair?
No, they are not required at all. The whole idea was to use few components for demonstration.
One of the points to demonstrate was: what happens when you put your logic output on a cable (f.ex. a CPU output). This is roughly equivalent to one inverter.
However, to show the cable capacity, I just needed a lot of drive current. What better than to use what I already have. Sure, a separate mosfet push-pull drive can be made, but the inverters have them already at the output stage.
Note that a “simple” pnp/npn pair as a push-pull drive is a bad choice. The turn-on and turn-off times are different and then the totempole will short in the transition. That is why you use specialized drivers that are matched not to short in transition.
or nchannel phannel fets
Yes, or even two n-channel MOSFETS with the appropriate drive circuitry. However you need to take some care in driving the transistors to reduce or eliminate shoot-through because both are on at the same time and you have a brief short from the power rail to ground.
You don’t always need a lot of power to drive a transmission line. If you terminate at the insertion point, then you have an implicit current limit.
This will work as long as you can recover the signal properly. For short lines, lets say less than 10..20m or so, you will normally see no problems with termination at insertion. When the line gets really long, then you feel the pressure of the capacitance and need to drive harder, which means that you need to terminate at the end.
It all depends on your specific setup.
Note that this is all about single-ended digital signals in on-off style. If you transmit using analogue techniques (and/or balanced), and then recover the signal to form your signal again, you may press a lot more bandwidth through a cable. However, it makes demands and requires extra hardware.
I’m surprised that the last part of the article was about balanced transmission lines – especially with his list of example balanced transmission lines.
I only mention it as Ethernet is a balanced transmission line.
Also, ethernet should be terminated by differential and common mode terminations, typically 100R and 50R. This is achieved by using a 50R resistor for each wire in the pair, then a 25R resistor between the node where the 2 50R’s meet and ground (in a Y configuration)
Balanced lines are only mentioned to note that there are different approaches. Primarily to show that the grounding problem can be solved differently.
Of course, you are right, Ethernet is a balanced line. It is also has galvanic separation (the Ethernet magnetics). All mentioned in the article do not have galvanic separation by default.
The termination in Ethernet is interestingly on the “other” side of the magnetics and in a special Y configuration. The magnetics are transparent to the line and the termination is at both ends. The Ethernet signal on the wire is a very analogue signal with very specific characteristics and much more complex than simple logic signals.
Thanks Bertho! It is always nice to brush up, and see this stuff applied. Great Document, easy to read and understand. If only teachers in EE taught this way.
Keep these types of articles coming Hackaday!