Grounding and electrical noise around a VFD
The drive switches several hundred volts several thousand times a second. If your USB drops out, limit switches false-trigger or a probe reads nonsense when the spindle starts, this is where the answer is — and it is almost never the software.
Symptoms that mean noise
- USB disconnects, or the controller stops responding, only while the spindle is running
- Limit or homing inputs trigger with nothing touching them
- Touch probe or tool setter readings that change with spindle speed
- A spindle speed that wanders when commanded to hold steady
- Radio or audio interference that appears at particular speeds
The last one is the diagnostic shortcut: if the fault tracks spindle speed rather than spindle load, it is electrical. Mechanical problems track load.
Do these in this order
The order matters because each step is cheaper than the next, and each one removes a cause that would otherwise mask the following ones.
| # | Action | Why it works |
|---|---|---|
| 1 | Bond the spindle frame and the VFD chassis to earth with a short, thick conductor | Gives the switching current a low-impedance path home instead of through your signal wiring |
| 2 | Use screened motor cable and clamp the screen 360° at the drive end | The screen carries the coupled current; a pigtail does not work at these frequencies |
| 3 | Route signal cables away from the motor cable; cross at right angles if they must meet | Coupling falls off steeply with separation |
| 4 | Separate the supplies: controller on its own supply, not off the drive's input | Removes the shared-impedance path entirely |
| 5 | Use screened signal cable, earthed at one end | Keeps the screen from becoming a ground loop |
| 6 | Lower the drive's carrier frequency | Less energy at high frequency to couple out. Costs audible noise |
| 7 | Fit a line reactor or input EMI filter | Keeps the drive's switching noise off your supply |
| 8 | Ferrites on the signal cables | Cheap last resort; works, but treat it as confirmation of a routing problem, not a fix for it |
One earth point, or two?
You will find both practices advocated, and both are defensible depending on the length of the run:
- Screen earthed at the drive end only is the usual recommendation for short runs inside one machine. It cannot form a ground loop, because there is only one connection.
- Screen bonded at both ends, with a full 360° clamp at each, is better at high frequency and is standard practice on long runs — but only where you are confident there is no low-frequency potential difference between the two ends to drive a loop current.
What is never right is a screen connected by a long flying lead, or not connected at all. A screen left floating does nothing; a screen earthed through a pigtail has enough inductance at switching frequencies to be nearly as useless.
The speed reference is the sensitive part
A 0–10 V speed command is a high-impedance analog input, which makes it the easiest thing in the machine to corrupt. Two rules cover most cases:
- Run the analog reference with its own return conductor back to the controller. Do not use the chassis or the spindle frame as the return path — that is how a speed command acquires a few volts of noise riding on it.
- Keep the analog pair twisted and, if you screen it, earth the screen at one end only.
If the speed display on the drive is steady while the controller's commanded value is not, the problem is on the controller's output; if the drive's display itself wanders, it is the wiring.
Carrier frequency: the trade you are making
Raising the carrier frequency moves the switching noise out of the audible range, so the spindle gets quieter. It also increases the drive's switching losses, which means a hotter drive and slightly less output available, and it increases the leakage current into earth — which can make an RCD nuisance trip. Lowering it does the reverse.
There is no correct value, so set it by outcome: low enough that the drive stays cool and the machine behaves, high enough that the spindle noise is tolerable. If changing it changes your noise symptoms, you have confirmed the diagnosis.
When it is not noise
Before spending an afternoon on ferrites, rule out the two mechanical causes that look identical:
- A tool pulling out of the collet under load presents as erratic behaviour and audible change.
- A spindle with failing bearings draws more current and generates broadband electrical noise, because the drive is fighting a mechanical load that varies within a single revolution.
Both are cheaper to check than to misdiagnose.
Parts
Filtering and suppression
Questions
Why does everything run fine until I start a cut?
Load current rises, and the noise rises with it. The part that misbehaves is usually the 0–10 V speed reference or the limit-switch wiring, not the spindle.
Should the motor cable screen be connected at both ends?
Yes, for a VFD output cable — bonded at both ends with a full 360° connection, because the high-frequency return current needs the screen as its path. Bonding one end only helps at audio frequency and makes drive noise worse.
Will lowering the carrier frequency fix the interference?
It usually reduces it, at the cost of more audible motor whine and slightly rougher current. It is a trade, not a fix — see the separation and routing section above.
Sources and method
Three kinds of statement appear on this page, and each is labelled where it occurs:
- Derived — the failure mechanisms are described as mechanisms; the cures are practice, and where practice is not a standard the page says so.
- Rule of thumb — workshop practice, not a standard. Given with the range it is meant to hold over, and it loses to your manual wherever the two disagree.
- Maker-specific — parameter numbers, connector pin numbers, minimum brake resistance, maximum cable length, bearing speed limits. Not published here. They move between firmware revisions of the same model, and a wrong one is worse than none.
Standards and documents
- IEC 61000-6-2 — EMC immunity, industrial environments
- IEC 61000-6-4 — EMC emission, industrial environments
- IEC 60204-1 — electrical equipment of machines
- IEC 60364 — low-voltage installations: protective conductors and earthing
- The manual for your own drive and spindle — the only valid source for maker-specific values