Why star ground????

I think half the problem with noise issues is how people route wires in the box and on the machine, the other half is because they try to do things on the cheap and use components like cheap £5 Bob's using the parallel port with long and cheap cables.

I've lost count of how many control boxes and machines I've wired/built over the years (I've done 6 in the last 2 wks) and I very very rarely suffer from noise issues on stepper based machines, servo based can be a little more touchy but still very minor and rare. I don't fit any filters or ferrite rings on anything, I can place VFDs inside and outside control boxes and this is mostly using cheap Chinese VFDs like Huanyang, Fuling, etc. Now, this isn't because I'm a master electrician or Wizzard or anything it's because I use good components and good cable management along with good wiring practice.

I don't understand why people try to cut corners by using cheap £5 Bob's and the parallel port these days when a decent Ethernet motion controller like AXBB-E which includes the BOB costs approx £200 with the license file. Now, I know some are going to say £200 is a lot of money compared to £5.!! . . . But this is a false economy.

The first time the machine throws a bitch fit because of noise while cutting all your savings can be blown as it breaks the cutter, wrecks the material not to mention time wasted, esp if doing this for a living. Even if you are doing this as a hobby it doesn't take long to recover the costs of doing things correctly. Not to mention the stress and frustration it will save.
Many people spend more money trying to chase the noise away than doing it right the first time.! Then throw in the performance increase and reliability and it just doesn't make sense not to do it.

If I were building from scratch I'd definately be going for the AXBB-E and UCCNC software combination rather than my current parallel port and cheap BOB combo. I have spent a lot of time fault-finding on that in the past and am not impressed with my failure to see the real cause before. But as that's what I have and I'm used to LinuxCNC and I've had it all working correctly in the past so I'm going to stick with it for now, though I have recently checked what the delivery time to Australia would be. The local supplier's markup equates to an arm and a leg so I would not use them. If this hobby turns into a money spinner in my retirement then an upgrade to a more reliable system would be a sensible investment.

Having seen the amount of noise present on the earthed case of my controller and everything inside it when the VFD is running I'd want a filter on the mains to protect everything else nearby. For $35AUD it's an easy fix (fingers crossed) though they want $40 for a simple aluminium box to put it in!! I've gone for a 10A two stage filter very similar in design to the one Muzzer linked to. RS are sending it from the UK so many readers on here could have one sooner than me if required. It's this one... https://au.rs-online.com/web/p/power...lters/1863517/

As I mentioned before there is no filtering on the earth cable and interference at high frequencies can escape via this route. There is no such thing as an earth at radio frequencies. Muzzer mentioned clamp-on ferrites as good for RF supression and one option for a filter I plan to test out is simply three ferrite rings with the live, neutral and earth wires into the VFD threaded through one each. This would prevent the spurious tripping of earth leakage breakers which can occur with conventional mains filters.

Kit

I'll side with Muzzer on this one. As far as EMC is concerned, thar be dragons,

Quote:

---

Originally Posted by Doddy

I'll side with Muzzer on this one. As far as EMC is concerned, thar be dragons,

---

Thar be dragons indeed! One of the problems with protecting against electrical interference is that you're dealing with a wide spectrum of frequencies in different circumstances from mains frequency hum to very fast spikes and surious oscillations with frequency components into the mulit-megahertz. There is no single catch-all solution that is best for all ocassions. Only grounding screens at one end is an example. This advice originates from analogue audio and video practice where ground loops lead to mains hum which is both audible and visible even at low amplitude. But as Muzzer pointed out, grounding at both ends has worked for him in situations involving VFD installations where an important aim is to keep fast pulses out of digital circuits.

Quote:

---

Originally Posted by Kitwn

Only grounding screens at one end is an example. This advice originates from analogue audio and video practice where ground loops lead to mains hum which is both audible and visible even at low amplitude. But as Muzzer pointed out, grounding at both ends has worked for him in situations involving VFD installations where an important aim is to keep fast pulses out of digital circuits.

---

In seriousness - I've read and tend to agree with the general philosophy - ground both ends on an emissive line (steppers, spindle) - you really don't care with those and it helps avoid these becoming antennas of high speed, high current signalling whenever the shielding ends/fails. And use star-connected single-ended for susceptible lines (limit switches, estops, etc) where you are trying to avoid any nasties getting INTO the line. I consider it a rule of thumb that's likely subject to any amount of abuse, but it generally works for me. YMMV.

Quote:

---

Originally Posted by Kitwn

There is no such thing as an earth at radio frequencies.
Kit

---

Well I was taught that this is not quite true, but only if it is a plane (which could be the walls of an enclosure as long as they're conductive and conductively bonded across any joints etc.) - as soon as you give it length >> width (i.e. a piece of wire) it has inductance and everything goes to rat shit, and smelly piles of it.

Quote:

---

Originally Posted by Voicecoil

Well I was taught that this is not quite true, but only if it is a plane (which could be the walls of an enclosure as long as they're conductive and conductively bonded across any joints etc.) - as soon as you give it length >> width (i.e. a piece of wire) it has inductance and everything goes to rat shit, and smelly piles of it.

---

Absolutely true, but you have to take great care.

When I was a shiny new trainee at the BBC Transmitter Dept. back in the early 80s there were two demos that brought home the point I was wanting to make...

In the antenna fields of Daventry shortwave radio station the open wire feeders were supported on insulators hung about 3m up from tubular steel posts about 200mm in diameter which were firmly concreted into the ground. If you held a coin and ran it down the post under a live feeder you could clearly see sparks between the coin and the post.

At the Droitwich station there was a large neon bulb connected between the outer ground of a medium wave feeder where it left the building and the station earth strap which ran all round the inside of the building. It glowed a lovely neon pink and flickered in time to the music.

Happy days!

Couple of things I have looked for when choosing a BOB (or motion controller with integrated BOB) are differential signalling (BOB to stepper driver), and 24V signalling for external signals, not 5V as often seen. Add in attention to wiring runs (keep power/signal separate) and star earthing and you are starting from a good place. I have stepper/spindle/limit switch wiring all running in the same cable chains - although the "power" cables are separated from "signal" by the spindle cooling pipes - and have been lucky enough not to suffer any noise problems. But you do have to work at it to try to head off the problems from the outset - I really hated the idea of having to track down random noise issues. I have also run a ground continuity wire through the cable chains to bond moving bits together (not relying on connections via bearings/ballscrews). I suspect that the proximity switches wired in series are happier with 24V - when I tested them before installation in a series configuration their minimum voltage was around 10V so I suspect even a 12V supply would make them more noise-susceptible. Industrial systems presumably use 24V for a reason!

CANbus lines only wobble up and down by a volt - so although higher signal levels can appear to make the job easier, they aren't necessary. And of course, they can become the cause of noise problems themselves.

In SMPS and VFD design (my background), you may need to be switching hundreds of amps and hundreds of volts simultaneously, with transition times in the 10s of nanoseconds and switching frequencies of 100kHz or more. In the middle of all this you may need to implement accurate analogue voltage conditioning and measurement circuits and operate microcontrollers, whilst also meeting stringent EMC standards. Some of the noise problems we come across in cabinet wiring are relatively simple in comparison but can still make life pretty difficult.

As pointed out, there is often no such thing as a hard ground, due to real world inductances. Even a ground plane within a circuit board has inductance - I've seen many examples of noise problems where the designer has assumed that a large ground plane / copper fill under all the circuits will provide some form of magical ground. However, induced voltages are easy to generate between points in a ground plane and can cause unpredictable behaviour, particularly in ICs if you take the internal substrate even a few hundred mV above ground. The solution is to bring the 0V connections for the various analogue, digital, power etc nodes together at a star point - they are often essential at circuit board level, too.