0
\$\begingroup\$

Okay, I have a motion control application that I want to wire up. The motor drives run from single-ended step and direction signals. I am designing a circuit board, and was thinking of using 8-pin RJ45 on my board, then routing the signals out to the motor drivers using Cat5 cable.

My question is: if I run the step signal through pin 1 of the Cat5 cable, and leave the other wire of that particular twisted pair unused and unconnected, are there any potential negative repercussions to that?

What if I ran the step signal through one wire of a pair and the direction signal through the other wire of the same pair? Is there any reason why that would be a bad idea?

\$\endgroup\$
  • \$\begingroup\$ It's all about distance and speed. Short distances and slow (non critical timing) signals will be fine. \$\endgroup\$ – Andy aka Jul 29 '15 at 7:09
  • \$\begingroup\$ What is supplying the ground reference? If you want to do overkill, route 2 pairs with ground and then use the other two pairs each pair connected together as a single conductor, to give you step and dir. If you hope to drive 4 step and 4 dir signals on one cable with no ground you will likely have problems unless you have a ground in common some other way. \$\endgroup\$ – KalleMP Jul 29 '15 at 21:31
0
\$\begingroup\$

As Andy says in the comment, for low speed signals over short distances you can use almost any cable for any purpose. Where your step pulses are likely at most 10kHz signals they are many orders of magnitude slower than what we would consider the tipping point where it could become problematic.

Even 100kHz or 1MHz for the distance between a control board and a motor board will not see any problems.

If you run the "low frequency" step signal through a wire of a pair and put the direction signal through the other wire, as long as the direction signal is actively driven to Vcc and GND it should not notice too much of it. Again, it's a small distance and at low frequency over small distances twisted pair is just another bunch of wires that are close to each other.

EDIT: Little more added while modifying the schematic below

If you want to use "low frequency" and get a little reflection immunity with only a single component, you could add a resistor at the transmitter side of the pulse signal, so that the reflections get "tied down" at the transmission side.

But this requires the output resistance of your driver to be known as compared to the signal on the other wire to work properly.

To avoid a lot of look-up or testing you can try 75 Ohm if the other wire is a hard ground or 50 Ohm if it is also driven by a modern digital chip's output. Many logic chips have internally something between 15 and 40 Ohm output impedance, but that's very general and may well be wrong in specific situations.

I averaged it into about 25 Ohm, in which case if you use ground on the other wire and as such just one output and then add a 75 Ohm resistor that adds up to 100 Ohm on the TX side. If you use two outputs (direction signal on the other wire) that will be two pins of 25 Ohm, so you only need 50 Ohm to add up to 100 Ohm.

To note: This is not a very neat trick and it involves a lot of guess work (since the PCB also has certain impedances and the whole trick with the resistors will then mismatch those, etc, etc), so you should avoid it if you can and use proper Differential drivers and receivers when single ended start to be so quick that you start seeing problems.

End of Addition


Some sideways thinking about higher speeds (Also Edited):


If you ever do need higher speed with single ended pulses, you can quite easily modify the system to at least reduce the problems over the cable. It's far from perfect, but it will be a little better than nothing, if for some mysterious reason a true differential driver isn't possible:

schematic

simulate this circuit – Schematic created using CircuitLab

\$\endgroup\$
  • 1
    \$\begingroup\$ Excuse my ignorance, but what that circuit actually work? You'll never have a -ve differential voltage at the end so the comparator would always output a '1' regardless of the input. You would have to bias the second input to Vcc/2 to get a signal out of the comparator. For high-frequency a better approach would be a single ended to LVDS converter to drive the cable (very cheap and in SOT23-5 packages so small), there are then options for going the other way (LVDS to single ended) at the other end. Also Cat5 cable is 100 Ohm Zo. \$\endgroup\$ – Tom Carpenter Jul 29 '15 at 14:03
  • \$\begingroup\$ @TomCarpenter Ah, you're right, most comparators (though some can be tweaked for a >+100mV or so thresh) would get into indeterminate behaviour. In my hurry to get back to work I modified the isolated-ground decoupled situation for common ground wrongly. If you'd have two power domains and decouple with caps (since it's a HF signal) that would work fine, possibly even with a simple transistor on the receiver end. Re the Zo, right again. I'm working with/debugging RS422 this month, so 110 comes floating up instantly, but wrongly. So you see, hurry is good for naught. \$\endgroup\$ – Asmyldof Jul 29 '15 at 14:38
  • \$\begingroup\$ Thanks @Asmyldof. My signals are up to 75kHz in frequency. I don't know if that changes anything or not. And what's the comparator for in your diagram? \$\endgroup\$ – JayGee Jul 29 '15 at 15:12
  • \$\begingroup\$ @JayGee as my original (and now modified) answer tried to imply: for 10's of cm, or even a meter or two, even 1MHz is unlikely to give you any real and noticeable problems, so 75kHz is definitely going to be fine. If my top-of-head estimation isn't very far off, 75kHz could make it tens of meters through a normal twisted pair before you get problems that a 3.3V or 5V digital system will notice. \$\endgroup\$ – Asmyldof Jul 29 '15 at 15:16

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.