Stepper/Servo/Switch wiring with coax

[htrantx]

I am not quite sure why many using coax for stepper/switch wiring. Would it because noises, EMI issues and/or cross-talked can induce miss-stepping?

I was looking at the CAT 5e spec and it seems to be more than adequate for running wiring to motors and especially switches in a CNC machine. Each CAT 5 cable has 4 pairs (8 wires) of 24AWG , each wire can handle 0.5A DC (note it is DC), so if you need more current, just gang them up.

The NEXT of a CAT 5 is almost 70dB@1MHz and 30db@100MHz and this is way above what we need in term of xtalk.

Personally, I think using coax in a CNC machine wiring is a bad idea, since the coax shield has to be grounded properly. This is where many, including me, have run into trouble, and sometimes might not even know it. Beside, CAT5 cable is dirt cheap comparing to a coax.

What do you think?

ht

[bharbour]

I am not too thrilled with using CAT5 cable for motors or switches. You certainly do not want to use a solid conductor wire for external wiring on a CNC machine. Copper work hardens and breaks under conditions of vibration and motion. Ganging up several conductors is messy in connector pins. A step motor requiring 4 amps on 4 phases is not going to work with a singel CAT5 wire.

A shielded, stranded multiconductor cable works well for me. Coax only gives you two conductors per cable and would be a big pain inside connector shells because you would have multiple cables for each motor. For switch/sensor wiring, some kind of shield will keep noise problems down. Twisted pair is some help, but I prefer the sheilded twisted pair myself. Your milage may vary.

BobH

[htrantx]

Originally Posted by bharbour I am not too thrilled with using CAT5 cable for motors or switches. You certainly do not want to use a solid conductor wire for external wiring on a CNC machine. Copper work hardens and breaks under conditions of vibration and motion. Ganging up several conductors is messy in connector pins. A step motor requiring 4 amps on 4 phases is not going to work with a singel CAT5 wire. A shielded, stranded multiconductor cable works well for me. Coax only gives you two conductors per cable and would be a big pain inside connector shells because you would have multiple cables for each motor. For switch/sensor wiring, some kind of shield will keep noise problems down. Twisted pair is some help, but I prefer the sheilded twisted pair myself. Your milage may vary. BobH

Good points on copper wire work hardened, definitely not for high motion/vibration wiring.

I think a multiple CAT5 wires set should be ok for motors; basically the wires work as a multi-stranded-twisted cable. Unlike most coax, CAT5 has 100 ohm impedance and that should soften the rise/fall time.

Great point on the coax connectors. This is a huge problem as the mismatching impedance generates standing waves. So what coax is good for when noises are concentrated outside the coax? This is just like using coax connecting noise source to antennas (motor's pig tails) for transmitting. Would'nt you get stronger signals?

So contra to the believe, by using coax cables for motors, we have inadvertently made good RF noise generators... Am i wrong here?

Coaxing switches and grounded the shields to the system common ground is a bad idea, since the receivers might not referred to the system ground. I'll say forget the coax, just a low pass at the receive side would do it.

ht

[Torchhead]

Gawww! the 100 ohm impedance is at 10Mhz! The "softening" of a waveform is a product of the impedance of the conductor at a given freq. In a wire it's more shaped by the capacitance than the inductance or DC resistance. You also have to match the line impedance with the same effective DC resistance on each end to get any benefit. Driving several amps of PWM current down a wire to a motor (low impedence) will do little to change the waveform characteristics. COAX is a cable with an outer shield and single inner conductor with a specific dialectric giving it a known capacitance and inductance per linear foot. Multi conductor cable with an outer shield is not coax. Since the shield is better at keeping noise out than in, it works better in low current, low level signals than the high currents of motor PWM. There are higher freq components in a square wave but not much in the higher RF ranges.

UTP is most effective with a balanced line (differential signaling). Ethernet has a transformer on each end and no system referenced ground between the two ends. The twists tend to cancel out outside noise because it is "common mode" (same nose on both conductors) but that is not as effective on an unbalanced line.

The prevailing problem in CNC electronics is conducted noise not RFI. You have lots more noise problems from "dirty" common ground connections in a power system than what gets transmitted through the air.

Most systems will work fine with the motor wires unshielded and simple low pass filtering on the inputs (opto's help by not passing the noise from the grounds). If you just don't hook up the shield on either end, it than has little effect one way or the other except in the capacitance of the wire per foot.

The wavelength of the signals in the power parts of the CNC machine are in the kilohertz. Antenna theory and standing waves (based on the wavelength and harmonics of a wave) at the kilohertz range end up in hundreds of feet! Figure the wavelength of a 25 KHZ wave. Its 3.9343e+4 feet! That is 39,343 ft. A wire that would have standing waves at even 10 times the primary frequency would be have to be 3,934 ft long or a harmonic division (1/4 wavelength being a common value) The normal lengths of wiring in a machine is in the sub 30ft range.

The type of wire you use is not nearly as important than it's ability to carry the current and how you route it. Keep high current conductors from running close to and in parallel with logic level wires.

The bending and wear characteristics of a solid wire with it's own insulation jacket is NOT the same as a multiwire cable all in one jacket. It's more flexible to have several smaller wires in parallel rather than one bigger solid one but a solid wire of one conductor with a jacket has a given bend and fatigue characteristic that does not change if you group a bunch of them together.

TOM Caudle
www.CandCNC.com

[der_fisherman]

I like the way this blog is going, as my personal interest is to learn exactly what cable types accept flexing best while still carrying the current needed....limit switch cabling etc etc.

How should one route the cables? How does one get hold of those "Snake cable guides" or should one just build ones own? (I have seen a simple method of doing this).

Any good tips with regard to this area of the build would be much appreciated, especially photos/drawings.

Regards
Der_Fisherman

[htrantx]

Originally Posted by Torchhead UTP is most effective with a balanced line (differential signaling). Ethernet has a transformer on each end and no system referenced ground between the two ends. The twists tend to cancel out outside noise because it is "common mode" (same nose on both conductors) but that is not as effective on an unbalanced line. TOM Caudle

Excellent point! UTP has little advantage for single ended signals.

Originally Posted by Torchhead Most systems will work fine with the motor wires unshielded and simple low pass filtering on the inputs (opto's help by not passing the noise from the grounds). If you just don't hook up the shield on either end, it than has little effect one way or the other except in the capacitance of the wire per foot.

It is not a good idea to "float the shield", I would ground one end.

Originally Posted by Torchhead The type of wire you use is not nearly as important than it's ability to carry the current and how you route it. Keep high current conductors from running close to and in parallel with logic level wires. TOM Caudle

Finally, someone knows what is going on. Shielding blocks RF, but do little to stop electromagnetically induced noises. The only practical way is to physically separate them. Them "shielded wire" cables help nothing but headache.

ht

[Torchhead]

We have to deal with extreme cases of EMI and RFI with the plasma kits we sell. You can ground things until you turn blue and if it's done wrong you can't get rid of the conducted noise. Sometimes doing things that seem logical produces strange results. Example: you would think running wires down a metal conduit would be a good thing but it forms a Faraday shield and keeps the EMI inside and even concentrates it, so crosstalk between conductors in the same conduit gets worse. Grounding done wrong can add to the problem instead of solving it.

Industrial EMI and dealing with high noise environments from nearby machinery is the subject of books.

Wire is wire. You can put 100 amps across a 20 ga piece of wire if you can keep the series resistance down. Immerse it in liquid Helium and it will be 99% efficient. Wire and magnetics and motors are all things that take gross overloads and live. You can put 4 A of PWM down a small wire and the DC resistance will drop the peak voltage slightly and the wire will heat based on I^2 * R times the duty cycle. Unlike semiconductors that will fail within micro seconds with an overload or overvoltage (faster than a fuse) wire will just get warm, then hot , then change color and finally melt at it's weakest point. In a lot of circumstances the biggest loss of power is at the connectors. Bad crimps or solder joints can have a lot more resistance than the connecting wire.

There are things to worry about in CNC but wire is one of those things that live under the 20% rule. Get in the ballpark and you are alright.

The more conductors a wire has in one jacket the more flexible. The alloy of copper used and if it's been annealed after it's drawn will change the flexibility. Does it matter? Maybe, if you want the machine to last for 20 years instead of 10. The larger the arc of the wire at bending points, the less stress. Strain relief at connectors will keep the fatigue points spread across the bundle rather than the weakest points. There is always a trade off in engineering of cost VS failure rates. Typically electronics last longer if kept cool. Is it worth spending money on huge coolers and extra fans to increase MTBF from 15 years to 30? MTBF says that a certain percentage of components will fail across the given time based on the temperature they are run at (and for how long). Most will fail in the center part of the curve. Some will fail no matter how cool you keep them and others will last no matter the conditions (within the max specs of the device). You try to design with cost in one hand and reliability in the other.

Which brings us back to wire. You can pay .10 per foot or 3.00 per foot for wire and you may never see a difference. It may be more flexible, it may be lower resistance per foot, it may have a jacket rated to 600V instead of 300V. If it makes you sleep better at night knowing you used premium high dollar cables then there is value in that. If it's because someone (that is not paying for it) told you that you needed it then the value is harder to quantify.


TOM Caudle
www.CandCNC.com
Totally Modular CNC Electronics