Installing and commissioning a variable frequency drive (VFD)

What a VFD does and why you install one

A variable frequency drive, a VFD, controls the speed of an AC induction motor by varying the frequency and the voltage it sends to the windings. The motor's speed follows the frequency, so drop the frequency from 60 Hz toward 30 Hz and the motor turns at roughly half speed. The drive raises and lowers voltage along with frequency to keep the magnetic flux right, which is why people call the basic mode volts per hertz.

Three reasons put a drive on a motor, and the first one pays for it. On a pump or a fan, the affinity laws mean power follows the cube of speed. Slow a fan to 50 percent speed and it draws close to 12 percent of full power, because flow tracks speed, pressure tracks speed squared, and shaft power tracks speed cubed. Throttling a damper or a valve to get the same flow just burns the energy as heat across the restriction. The drive saves it instead.

The second reason is the soft start. A motor started across the line pulls six to eight times its running current for the first moment and slams the driven equipment with full torque. A drive ramps frequency up from zero, so the inrush is gone and the belts, couplings, and pipe see a controlled start. The third reason is process control: holding a tank level, a duct pressure, a line speed, by trimming motor speed to a feedback signal. The motor sizing rules behind the load still apply, and they are covered in the motor circuit conductor sizing guide.

How a VFD works inside the box

A VFD does its job in three stages, and knowing them tells you where the problems live. The front end is a rectifier, a bank of diodes that turns the incoming AC into DC. That DC charges a bank of capacitors that form the DC bus, the reservoir that holds a steady DC voltage. On a 480 V drive the bus sits around 650 to 680 V DC.

The output stage is the inverter: a set of IGBT transistors that switch the DC bus on and off thousands of times a second to build an AC waveform of whatever frequency the drive wants. It does not make a smooth sine wave. It makes a string of full-voltage pulses, wide and narrow, that average out to a sine wave the motor can use. That is pulse width modulation, PWM.

The rate the IGBTs switch is the carrier frequency, set in the parameters and usually adjustable from about 2 kHz to 16 kHz. Raise it and the motor runs quieter and the current waveform gets smoother. Raise it and you also make more heat in the drive, push more common-mode current into the motor bearings, and shorten how far the drive can sit from the motor. The carrier frequency is a trade, not a free quality knob, and most of the install problems further down this guide trace back to those fast PWM edges hitting a cable and a motor that were not ready for them.

How do you size a VFD?

Size the drive to the motor's full-load current and the load type, not to horsepower alone. The drive is a current device. Its output transistors care how many amps flow through them, so the rated output amps of the drive have to meet or beat the motor full-load amps you intend to run, with the load type setting how much overload headroom you get on top of that.

Load type splits into two families. Variable torque is the pump and fan world, where torque falls off as speed drops, so the drive rarely needs more than its rated current. Manufacturers also call this normal duty, and a normal-duty rating commonly allows about 110 percent current for 60 seconds. Constant torque is conveyors, compressors, extruders, positive-displacement pumps, anything that needs full torque at low speed. That family is heavy duty, and a heavy-duty rating commonly allows about 150 percent current for 60 seconds because the motor draws full current even when it is barely turning.

Here is the part that bites people: the same drive carries two horsepower ratings. A unit rated 20 hp variable torque might be only 15 hp constant torque, because the silicon is the same but the duty it can survive is not. Pick the drive off the nameplate amps and the duty class, then confirm the horsepower rating matches the way the load actually runs. Put a variable-torque drive on a constant-torque load and it trips on overload or fails. The manufacturer's normal-duty and heavy-duty current tables govern the selection.

The input side: line reactor, fusing, and SCCR

The diode front end of a drive does not draw a clean sine wave from the line. It pulls current in sharp gulps at the peaks of the voltage wave, and that distorted current is harmonic current flowing back into your service. A bare drive can run a current total harmonic distortion north of 80 percent, which heats transformers and neutrals and disturbs other equipment on the same bus. The drive is a harmonic source, and on a building full of drives it adds up.

The standard first defense is impedance ahead of the drive: an AC line reactor on the input, or a DC link choke inside the drive across the bus. A 3 percent reactor commonly pulls the current distortion down from that 80 percent range toward 30 to 40 percent, and a DC choke does a little better on the dominant 5th and 7th harmonics. The reactor does double duty. It also buffers the drive from line transients and notching, so it protects the diodes and the bus caps from the disturbances a utility or a nearby load throws at them. On a stiff service or a long-lead drive, treat the input reactor as standard, not optional. For the facility-wide harmonic budget against a target like IEEE 519, the harmonic mitigation guidance covers the larger study.

Two more input items the inspector will look for. The branch overcurrent device has to match what makes the drive's short-circuit current rating, its SCCR, valid. A drive's SCCR is not a fixed number on the label alone. It depends on the protective device paired with it, and the same drive can list a higher SCCR with the right Class J or high-speed fuses than with a breaker, because fuses clear faster and let through less energy. Size the conductor feeding the drive at 125 percent of the drive's rated input current, commonly cited at NEC 430.122 for power conversion equipment, and confirm the article and section against the adopted code edition.

Why do long motor leads damage the motor?

Long motor leads damage the motor because of the reflected wave, also called transmission line effect, and it is the failure that catches crews who treat VFD output cable like ordinary motor leads. The drive's PWM pulses have very fast rising edges, several kilovolts per microsecond. The cable and the motor have different impedances, so when a pulse reaches the motor terminals it does not just stop. Part of it reflects back toward the drive like an echo down a pipe.

On a long enough cable the reflected pulse arrives back at the motor while the next pulse is still rising, and the two add. The voltage at the motor terminals can reach nearly twice the DC bus voltage. On a 480 V drive with a bus near 680 V, terminal peaks can climb toward 1400 V, and measured numbers near 1000 V on cable runs as short as 25 ft are common. That overvoltage pounds the first few turns of the motor winding insulation every switching cycle. The motor does not fail at startup. It fails months later, an insulation breakdown that gets blamed on a bad motor when it was the install.

Two things fix it, used together or apart depending on the run. A dV/dt filter or an output reactor at the drive slows the rising edge so the peaks never build, and it belongs right at the drive output, not partway down the run. The other half is the motor itself, an inverter-duty motor built to take the spikes, covered in the next sections. Long, hard runs get both.

How far can a VFD be from the motor?

How far a VFD can sit from its motor depends on the drive, the carrier frequency, and the motor's insulation, so the only number that truly governs is the one in the manufacturer's manual. As a working range, many low-voltage drives allow roughly 50 to 100 ft of motor cable with no output filter before the reflected wave gets serious. Past that, the peaks at the motor climb into damaging territory and you add a filter.

A practical breakpoint a lot of manufacturers use: for a motor that is not rated to NEMA MG-1 Part 31, a dV/dt filter is commonly called for once the run reaches about 50 ft. With a true inverter-duty motor the unfiltered distance stretches, but it is still bounded. For very long runs, into the high hundreds or past 1000 ft, a stronger reflected-wave or output filter, or a terminator at the motor, is the usual answer, and you accept some voltage drop across the filter.

Raising the carrier frequency shortens every one of these distances, because faster switching means more edges per second hitting the cable. If a run is marginal, dropping the carrier frequency a step or two often buys back the distance without a filter. Measure the actual routed cable length, not the plan distance, and check it against the drive manual before you commit the conductor. This is the cheapest place to get it right and the most expensive place to get it wrong.

Why do VFDs cause bearing damage?

VFDs cause bearing damage through common-mode voltage, and it is the single most common way a VFD quietly kills a motor that was otherwise fine. The PWM switching does not just make the line-to-line voltage the motor needs. It also makes a common-mode voltage, a voltage that the whole motor sees with respect to ground, oscillating at the carrier frequency. That voltage couples capacitively across the air gap onto the rotor and builds a voltage on the shaft.

When the shaft voltage gets high enough to break through the thin oil film in the bearing, it discharges through the bearing in a tiny spark, electrical discharge machining, EDM. One spark does nothing. Millions of them over weeks pit the races, then carve the evenly spaced washboard pattern called fluting, and the bearing roars and fails at a fraction of its normal life. The tell is a bearing that fails early on a motor that never had a bearing problem before the drive went in.

The fix is to give that current a path that is not the bearing. A shaft grounding ring, a ring of conductive microfibers that rides the shaft and shorts it to the frame, sends the common-mode current safely back to ground. On larger motors the common pairing is a grounding ring at the drive end and an insulated or ceramic bearing at the opposite end, so a ring alone does not just push the current through the far bearing. Many inverter-duty motors come with a ring fitted. If yours does not, fit one. This is not a refinement. It is the difference between a motor that lasts and one that does not.

The inverter-duty motor and low-speed cooling

A motor fed by a VFD lives in a harsher world than one across the line, and a standard motor often is not built for it. The reflected-wave spikes hammer the winding insulation, and the distorted current adds heat the motor was not rated to shed. An inverter-duty motor, built to NEMA MG-1 Part 31, has an insulation system designed to take it.

Part 31 sets the bar a definite-purpose inverter motor has to clear. For motors rated up to 600 V, the insulation has to withstand repeated peaks on the order of 1600 V with a rise time around 0.1 microsecond, which is exactly the kind of spike the reflected wave delivers at the terminals. Standard 230 V or 460 V general-purpose motors do not carry that rating, which is why pairing one with a drive on a long lead is asking for an early winding failure. Confirm the Part 31 marking and the manufacturer's inverter rating, and verify the current edition rather than quoting a clause from memory.

Cooling is the other half. Most motors are cooled by a fan on their own shaft, so when the drive slows the motor the fan slows with it and moves less air. A constant-torque load running for hours at low speed can cook a motor that is making full current but barely turning its own cooling fan. The answer is a motor rated for the speed range, a separately powered auxiliary blower that runs full speed regardless of motor speed, or a programmed minimum speed that keeps enough airflow. Size this to how the load actually runs, not to the nameplate it sat on across the line.

Output cable, grounding, and EMI

The output of a drive is a noise source, and the cable carries that noise everywhere it runs. Use VFD-rated cable: a symmetrical three-conductor cable with a continuous shield and a proper ground, sized so the ground can carry the high-frequency return current the PWM creates. Ordinary building wire in a shared conduit will radiate into every signal pair it parallels and turn a clean control system into a haunted one.

Bond the cable shield at both ends. This is the place electricians trained on signal cable get nervous, because on a low-level analog signal you ground a shield at one end to avoid a ground loop. VFD output is the opposite case. The shield has to carry high-frequency current back to the drive, so it is bonded at the drive and at the motor, with a full 360-degree gland or clamp, not a drain wire twisted to a lug. A pigtail ground undoes most of the benefit.

Keep the output cable away from control and signal wiring. Cross signal runs at 90 degrees rather than running them alongside, hold separation where they share a tray, and keep the drive's own grounding short and direct back to the system ground. The motor frame, the drive, and the panel share one low-impedance ground. Most of the strange intermittent control faults on a drive system are this: noise coupling that good cable, proper shield bonding, and clean grounding would have stopped.

Which VFD control mode should you use?

The control mode sets how the drive figures out what voltage and frequency the motor needs, and the right one depends on the torque you need at low speed. Three modes cover almost everything. Volts per hertz is the simple scalar mode that holds a fixed voltage-to-frequency ratio. It is fine for fans and pumps, and it is the one mode that can run several motors off one drive in parallel, but its torque falls apart near zero speed.

Sensorless vector control runs a math model of the motor in real time, watching output current and voltage to hold the torque you asked for and compensate for slip. It delivers strong starting torque, often 150 percent or more, down to a low frequency, and for most industrial loads it is the sweet spot between cost and performance. Closed-loop vector, or field-oriented control, adds an encoder on the shaft so the drive knows the real rotor position. That is what you use when you need full torque at zero speed or tight speed and position accuracy, like a loaded hoist or a coordinated line.

Vector modes need the motor's electrical model, and you get it by entering the nameplate and running an autotune. Enter voltage, full-load amps, frequency, rated speed, and power, then let the drive measure stator resistance and inductance with the autotune routine. An autotune is most accurate with the motor uncoupled from the load, though many drives offer a static tune in place. While you are in the parameters, set the accel and decel ramps to the load, the minimum and maximum speed, and the current limit. The manufacturer's parameter list and defaults govern, and they are not all the same brand to brand.

Start, stop, and safe torque off

How the drive gets its run command and its speed reference is a wiring and a parameter decision, and the two have to agree. The simplest is keypad control, run and speed from the drive face, good for a test bench and rare in a real process. Terminal control runs the drive from hardwired inputs: a two-wire scheme where a maintained contact means run, or a three-wire scheme with momentary start and stop push buttons and a sealed-in run. Pick two-wire or three-wire deliberately, because a drive set for two-wire can restart on its own when power returns, and that surprises people.

Network control runs the drive over a fieldbus, Modbus, EtherNet/IP, or similar, with the run command and the speed reference coming as data from a PLC. It cuts wiring and adds diagnostics, and it adds a dependency on the network being healthy. Many jobs mix them: a hardwired safety and stop, a network reference.

Most modern drives include safe torque off, STO, a pair of hardwired inputs that, when opened, stop the drive from producing torque without removing the main power. STO is a functional safety feature with a rating, and it is wired into the machine safety circuit so an e-stop or a guard switch removes torque through it. STO is not a disconnect. It does not deenergize the drive, so for service you still lock out the upstream disconnecting means. Treat STO as a safe-stop function, not as your lockout point, and prove it during commissioning.

Braking and overhauling loads

A drive ramps a motor up easily, because it just feeds it power. Stopping a high-inertia load fast, or holding back a load that wants to drive the motor, is the harder direction, because that energy has to go somewhere. When a motor is forced faster than its commanded speed, by a fan coasting, a conveyor running downhill, a centrifuge slowing, it becomes a generator and pumps energy back into the DC bus. Left alone the bus voltage rises and the drive trips on overvoltage.

The common answer for short, occasional braking is dynamic braking: a chopper transistor in the drive that switches the rising bus energy into an external resistor bank that burns it off as heat. Size the resistor and its rating to the braking duty, and protect it, because the usual failure of a brake chopper is a shorted transistor that dumps the full bus across the resistor with no control. A fuse sized to the resistor clears that before it spreads.

For a load that overhauls continuously, or one that brakes so often the resistor cannot keep up, the answer is a regenerative or active front end that returns the braking energy to the line instead of cooking it. That is a bigger, costlier drive, and you specify it when the load truly demands it, like an elevator or a test stand. Match the braking method to how the load actually behaves, not to a catalog default.

The bypass: running the motor if the drive dies

When the process cannot stop because a drive failed, you install a bypass: a contactor arrangement that disconnects the drive and runs the motor straight across the line at full speed. A common cooling fan or a critical pump gets a bypass so a failed drive is an inconvenience instead of an outage. The bypass is usually a package of contactors with an interlock so the drive output and the line can never both be on the motor at once, which would destroy the drive.

Two things people forget. Across the line, the motor gets full inrush and full-speed-only operation, so everything the drive was protecting, the soft start and the speed control, is gone while in bypass. The motor and the driven equipment have to tolerate a hard start, and any process that needed reduced speed will not get it. Second, the across-the-line side needs its own overload protection, because the drive's electronic overload is out of the circuit in bypass. That overload sizing follows the motor nameplate the same way it does in a motor control center, and the MCC commissioning guide covers setting it.

Decide early whether a given motor needs a bypass at all. Many do not, and a bypass adds cost, panel space, and another set of contacts to maintain. The ones that earn it are the loads where downtime is worse than running uncontrolled for a while.

Drive heat, enclosure, and derating

A drive is maybe 97 to 98 percent efficient, which sounds like nothing until you put a wall of them in a room. The 2 to 3 percent it loses comes off as heat, and a lineup of large drives can heat a room enough to need its own cooling. Plan the heat load into the electrical room HVAC, because a drive that overheats throttles itself back, trips, or ages its capacitors fast.

The enclosure sets the rules. An open chassis drive needs a clean, cooled, controlled space. A drive in a sealed NEMA enclosure traps its own heat, so it gets a fan and filter, a heat exchanger, or a panel air conditioner sized to the loss. Whatever it is, the filters are a maintenance item the owner inherits, and a clogged filter is a slow path to a heat trip.

Drives are rated at a reference ambient and a reference altitude, commonly 40 degrees C and around 1000 m. Run hotter or higher and you derate, because both reduce cooling. Above the reference altitude the thinner air cools less and the insulation withstands less, so the drive carries less current and sometimes less voltage. Check the manufacturer's derating curves for the actual room temperature and site elevation, and do not assume the nameplate amps survive a hot mechanical room at altitude.

Commissioning the drive

Commissioning is where the install becomes a working machine, and the order matters because you are proving one thing at a time before the load can hurt anything. Start cold. With the motor disconnected from the drive or the drive output isolated, verify the incoming power, the control wiring, and the grounding, then power the drive and enter parameters: the motor nameplate, the control mode, the accel and decel ramps, the minimum and maximum speed, and the current limit. Run the autotune if you are in a vector mode.

Prove rotation before the load turns. Uncouple the motor if you can, or be ready to stop fast, and bump it: a short run at low speed to confirm direction. Getting a pump or a fan backward is easy and on some equipment it is destructive, so this is the same bump test the MCC commissioning guide describes, just commanded from the drive. If rotation is wrong, swap it in the parameters or at the motor leads, not by guessing.

Then run it for real, in stages. Couple the load, run at low speed and watch the current, the speed feedback, and the drive temperature. Bring it up to full speed and confirm the running current sits at or below the motor full-load amps and that the speed holds. Test the overload and the protective trips. Open the safe torque off inputs and confirm the drive actually drops torque. Then back up the finished parameter set to a file or the keypad memory, because the day the drive dies, that backup is the difference between a ten-minute swap and a day of re-tuning.

VFDs on data center cooling

Data centers run on VFDs, and the cooling plant is where most of them live. Chilled water pumps, condenser water pumps, cooling tower fans, and the fans in CRAH and CRAC units almost all ride drives now, because cooling load swings with IT load and the affinity-law savings on a fan or pump that spends most of its life at part speed are large. A tower fan trimmed to hold a setpoint instead of cycling full-on and full-off saves energy and stops the thermal shock of constant starts.

The stakes change the install. These drives sit on critical loads, so the redundancy and the failure behavior get real scrutiny: a bypass on the pumps that cannot stop, drives fed from the right side of the power topology, and a clear answer for what the cooling does if a drive faults. The ride-through behavior on a power event matters, because a drive that trips offline on a sag and stays off has taken cooling with it.

Harmonics get a hard look here too, because a room full of drives on the mechanical side shares a service with sensitive load. The harmonic study and the mitigation, input reactors, chokes, or active filtering against a facility target, are part of the design, not an afterthought. The thermal guidelines the room is held to, along the lines of ASHRAE TC 9.9, set how tightly the cooling has to track, and the drives are the tools that hold it there.

What the owner inherits to maintain

A VFD is not install-and-forget hardware, and the maintenance it needs is the kind that gets skipped until the drive dies on a hot day. The cooling fans and filters are first. Drive fans are a wear item with a finite life, and a clogged filter or a seized fan turns a heat margin into a heat trip. Clean or change filters on a schedule and replace cooling fans before they fail, not after.

The DC bus capacitors age, especially if the drive runs hot or sits unpowered for long stretches. They have a service life measured in years that shortens with temperature, and a drive stored or idle for a long time may need its capacitors reformed before it is trusted on a load. Connections are the other quiet killer: the power terminals at the drive and the motor work loose under thermal cycling, and a loose lug makes heat and eventually an arc. Check and re-torque to the manufacturer's value on a maintenance interval.

And keep the parameter backup current. Every time the settings change, save a fresh copy, because the parameter file is the one part of the drive you cannot buy off a shelf. Hand the owner the backups, the torque values, and the filter schedule at turnover. That is the package that keeps the drive alive past its warranty.

What to document

A drive that runs today and has no record behind it is a drive nobody can troubleshoot or replace cleanly later. The turnover record ties the drive to the motor it runs, the protection on both ends, and the parameters that make it work. When a drive fails at 2 a.m. two years out, this record is what lets the next person put the same machine back without re-engineering it.

Capture the drive make, model, and rating with its duty class, the motor nameplate including full-load amps and horsepower, the input protection and the SCCR basis, the input reactor or DC choke, the output filter if fitted and the cable type and routed length, the control mode and the key parameters, whether a shaft grounding ring is installed, and whether a bypass exists and how its overload is set. Then attach the parameter backup file itself.

Common mistakes

• Running a long motor lead with no output filter, so the reflected wave fails the motor insulation months later.
• No shaft grounding ring, so common-mode current flutes the bearings and the motor fails early.
• Pairing a standard motor with a drive instead of an inverter-duty motor rated to NEMA MG-1 Part 31.
• Skipping the input line reactor or DC choke and dumping harmonics back onto the service.
• Sizing the drive by horsepower instead of motor full-load amps and the load type, then tripping on overload.
• Leaving rotation unverified, so a pump or fan runs backward at first start.
• Treating safe torque off as a lockout point instead of locking out the upstream disconnect.
• Carrier frequency set high on a marginal cable run, shrinking the safe distance and heating the drive.
• No parameter backup, so a failed drive becomes a day of re-tuning instead of a quick swap.

Field checklist

Standards and references

The NEC, NFPA 70, governs the install side. Article 430 covers motors and motor circuits, and Part X of that article covers adjustable-speed drive systems, with the supply conductor to the power conversion equipment commonly sized at 125 percent of rated input current, a rule often cited at 430.122. The exact section numbers move between code cycles, so confirm them against the adopted edition and any local amendments before citing them on a submittal.

The motor side leans on NEMA MG-1, and Part 31 is the section that defines a definite-purpose inverter-fed motor and the voltage stress its insulation has to take. The drive itself is built and rated to the adjustable-speed drive standards, NEMA ICS 7 and the UL and IEC drive listings, commonly UL 61800-5-1, which set the safety and the SCCR basis. For acceptance and maintenance testing of the installed gear, NETA's specifications give the framework. For the facility harmonic limits, IEEE 519 gives the targets a harmonic study works against.

Above all of these sits the manufacturer. The drive manual governs the parameters, the allowed cable length and filter selection, the derating curves, and the torque values, and it overrides any rule of thumb in this guide where the two differ. The project specification controls where it is stricter than the standards. Cite the standard that actually controls the point, verify the clause against the edition in force, and when the manufacturer and a general rule disagree, the manufacturer wins.

Units, terms, and conversions

A VFD goes by several names across a drawing set and a spec, so the same box can read differently depending on who wrote the document. Knowing the synonyms keeps you from chasing two part numbers for one thing.

A VFD is also called an adjustable-frequency drive, an adjustable-speed drive (ASD), a variable-speed drive (VSD), or just an inverter. Frequency is in hertz (Hz), and motor speed in rpm tracks it through the motor's pole count. The carrier or switching frequency is in kilohertz (kHz). The DC bus is in volts DC, near 1.35 times the AC line for a diode front end. Drive and motor ratings come in horsepower (hp) and kilowatts (kW), where 1 hp is about 0.746 kW, but the number that sizes the drive is amps. Torque is in pound-feet (lbf-ft) or newton-meters (N-m).

VFD / ASD / VSD / inverter

Names for the same device that varies motor speed by varying output frequency and voltage

PWM

Pulse width modulation, the switched-pulse output that averages to the AC waveform the motor uses

Carrier frequency

The IGBT switching rate, commonly 2 to 16 kHz, trading motor noise against drive heat and bearing current

Reflected wave

Voltage doubling at the motor terminals from impedance mismatch on a long cable run

Common-mode voltage

Voltage the whole motor sees to ground from PWM switching, the source of shaft and bearing currents

Variable torque vs constant torque

Load type that sets drive duty: pumps and fans (normal duty) vs full torque at low speed (heavy duty)

STO

Safe torque off, a hardwired function that removes motor torque without removing drive power, not a lockout

SCCR

Short-circuit current rating, valid only with the matched protective device, often a Class J fuse

Related tools

Calculators and readiness checks for this work

Compare your options

FAQ

People also ask

• Do you size a motor circuit off the nameplate?
• What is a motor control center?
• What is the difference between Level 1, Level 2, and DC fast charging?
• What is the difference between open and closed transition transfer?
• What is involved in cell tower construction?