Running older induction motors on VFDs

[Guineafowl]

TL;DR

What cheap and practical steps could a hobby workshop take to limit the damage caused by VFDs to old motors not designed for them?

The ‘spiky’ output of variable frequency drives (VFDs) can damage the insulation of older motors. They’re often used to run three phase motors on single-phase supplies in hobby workshops. (Modern motors have upgraded insulation and are designated ‘inverter-duty’.) I wonder what cheap and practical steps one can take to mitigate this, given that dedicated sine wave filters are quite expensive. I think most of us quietly ignore the problem and hope for the best.

Apparently, the motor windings themselves act as quite good low-pass filters for the current, and that it’s actually the voltage rise time that causes the problem. The high dv/dt pulses result in reflected waves sloshing back and forth due to impedance mismatch. These waves add together in the cable between VFD and motor, leading to high voltage transients that damage old motor insulation.

^^ I’m sure this is a poor summary, and also sure that you could write a whole monograph on the subject. But, down to brass tacks, what practical steps could we take in a hobby workshop with several old motors on VFDs?

1. In-line chokes, such as old fluorescent light ballasts, providing they could take the rated current?

2. Keep the VFD-motor cable short, and manipulate the VFD’s carrier frequency based on this length?

3. MOVs, or something, to clip the transients? I looked into the surge protection devices (SPDs) now becoming more readily available because they’re mentioned in the new UK domestic mains wiring regulations, but they are apparently not sensitive enough.

4. Some sort of snubber network at the motor terminals?

 

[Baluncore]

TL;DR: What cheap and practical steps could a hobby workshop take to limit the damage caused by VFDs to old motors not designed for them?

4. Some sort of snubber network at the motor terminals?

Not at the motor terminals, but at the inverter output. The voltage switching outputs need to see an inductive load. Wind three inductors on separate ferrite cores, one for each line, then couple them to a delta of pulse rated capacitors, giving three lines with limited dv/dt. A good VFD should have that circuit in its output to reduce radiated EMI.

You could also neutralise the motor power factor with a delta of capacitors at the motor terminals. That would be OK for fixed frequency operation, but difficult if you used a wide range of VFD frequencies.

 

[Guineafowl]

Not at the motor terminals, but at the inverter output. The voltage switching outputs need to see an inductive load. Wind three inductors on separate ferrite cores, one for each line, then couple them to a delta of pulse rated capacitors, giving three lines with limited dv/dt. A good VFD should have that circuit in its output to reduce radiated EMI.

You could also neutralise the motor power factor with a delta of capacitors at the motor terminals. That would be OK for fixed frequency operation, but difficult if you used a wide range of VFD frequencies.

So this would be a dv/dt, or perhaps LC filter, rather than a snubber?

What sort of values would work? Elsewhere it’s been suggested 5mH and 0.47uF. I’ll need a meter that measures inductance, unless I can just buy suitable inductors. I don’t mind spending a bit, since they’d be protecting essentially irreplaceable motors.

The carrier frequency of one of my VFDs can be adjusted from 1-15 kHz, default 3 kHz, if that’s relevant.

How would correcting the motor pf (which you helped me with before) deal with VFD voltage spikes?

 

[Baluncore]

So this would be a dv/dt, or perhaps LC filter, rather than a snubber?

Snubbers in this case would be RC networks that make the output look like a low resistance at high frequencies. That will attenuate the pulse on the line. The LC circuit does not waste energy in a resistor.

I don’t mind spending a bit, since they’d be protecting essentially irreplaceable motors.

Did you look at the lines with an oscilloscope? How do you know there is a problem, or is this just a fear?

How would correcting the motor pf (which you helped me with before) deal with VFD voltage spikes?

It would better isolate the motor inductance from the VFD. Correctly chosen capacitors across the motor will limit dv/dt, and reduce circulating current between the motor and VFD, that might otherwise saturate the VFD output chokes.

 

[Guineafowl]

Did you look at the lines with an oscilloscope? How do you know there is a problem, or is this just a fear?

I have a couple of old CROs, but not the right probes for connecting up. The VFD/old motor problem is described here: https://www.mtecorp.com/wp-content/uploads/SP-006-E_Web.pdf
Also, when buying new motors, they are often specified as ‘inverter-ready’ as mentioned above, implying that older ones are not.

 

[Baluncore]

If you can hear a buzz when a motor runs on a VFD, then that vibration may chafe the field coil insulation. The buzz is the sound of the higher harmonics that are not being attenuated by the VFD output LPF.
Any voltage spikes from the VFD will be due to there being insufficient LPF at the VFD output.

 

[Guineafowl]

If you can hear a buzz when a motor runs on a VFD, then that vibration may chafe the field coil insulation. The buzz is the sound of the higher harmonics that are not being attenuated by the VFD output LPF.
Any voltage spikes from the VFD will be due to there being insufficient LPF at the VFD output.

I can hear the 3 kHz switching frequency, yes.

https://industrialmonitordirect.com...-reflected-wave-trap-for-vfd-motor-protection
This link suggests the reflected waves (would it be a kind of SWR?) are significant only when the cable length exceeds 15m, so likely not a problem for the hobby shop.

For the chokes, would it be acceptable to use something like this:
https://www.amazon.co.uk/dp/B008IGQQFG/ref=sspa_dk_detail_4
Perhaps a bit bigger, with the ordinary single insulated wires supplying the motor wrapped as many times as possible?

 

[Baluncore]

Perhaps a bit bigger, with the ordinary single insulated wires supplying the motor wrapped as many times as possible?

Definitely bigger, but too many turns will saturate the core.
What is peak motor current? You need a core datasheet.
The aim is to add series inductance to the motor windings, to reduce high frequency voltage appearing across motor terminals. Place a delta of capacitors across the motor terminals to keep HF voltage away from the motor.

I can hear the 3 kHz switching frequency, yes.

Is the noise due to movement of windings or magnetostriction of the core?

 

[Guineafowl]

There are several motors that might need this modification, one being a Newman ‘Elf’ from 1961, for which I can’t find a datasheet. It’s a very heavy cast-iron motor.
Its plate says 1.8 A/phase, permanent star, 415V, but I’ve extracted the star point to enable delta 240V, so I’d estimate 3.1 A/phase.

I’m not sure where the sound is coming from. I’d guess the windings, since they’re more mobile than the core, unless the core has started to delaminate.

 

[Baluncore]

Have you considered flooding the field windings with epoxy resin to prevent movement ?

 

[Guineafowl]

No, because the sound isn’t a problem - it’s not annoying - and happens on even the newer motors. On some battery drills, when you just depress the trigger so you can hear the carrier frequency ‘sing’, before the mechanism starts moving and drowns out the noise. It’s that sort of sound.

It’s more what the sound represents. The fundamental of the VFD output is 50 Hz (10-100 Hz in the case of my milling machines, to achieve variable speed), but the PWM means harmonics, and its these that apparently damage motor insulation.

I was leaning towards an RC snubber at the motor terminals, because I probably have the components already, and none of them have to take the full motor current. This means I should be able to make the same snubber for all motors that need it.

In either the RC or LC case, I wondered if the capacitors should be X2 rated, since they’re across the mains, and should be designed for this?

 

[Baluncore]

The fundamental of the VFD output is 50 Hz (10-100 Hz in the case of my milling machines, to achieve variable speed), but the PWM means harmonics, and its these that apparently damage motor insulation.

That is why I suggest fixing the position of the windings, so they are not vibrated and chaffed by the harmonics.

A snubber would reduce dv/dt of the voltage spikes. What circuit would you use for snubbers?

 

[Guineafowl]

Something like this?:

I’m not sure if the star point of the snubber should be earthed; all circuits in my shop are RCD-protected to 30mA so I’d like to avoid earth leakage.

I’m also not sure of values, but assume that they’re not so critical, because of the wide gap between desired (<=100 Hz) and undesired (>3 kHz) frequencies.

 

[Baluncore]

If possible connect the Y of snubbers mid-point to the neutral, avoid a ground connection. If no neutral, design delta a snubbers, or leave the centre floating.
Do you have a circuit for the output of the VFD? I would expect an LC low-pass filter. If possible open the VFD and take a look at that filter. Measure the inductance and capacitance used.

 

[Guineafowl]

If possible connect the Y of snubbers mid-point to the neutral, avoid a ground connection. If no neutral, design delta a snubbers, or leave the centre floating.
Do you have a circuit for the output of the VFD? I would expect an LC low-pass filter. If possible open the VFD and take a look at that filter. Measure the inductance and capacitance used.

No circuit diagram, even in the manual for the decent VFD. At some point, I’ll get around to dismantling one of them as you suggest, but I have to assume the filtering in these devices isn’t adequate to protect older motor insulation, given the literature on the damage they cause.

I have a small box of X2 capacitors, rated for 310 Vac and 0.47 uF, which I used to fit into vintage valve (tube) radios when restoring them.

Starting with this 0.47 uF value, and a -3 dB cutoff frequency of 2 kHz, the Digikey RC filter calculator (https://www.digikey.co.uk/en/resour...sion-calculator-low-pass-and-high-pass-filter) suggests a resistor value of 169 ohm. No comment on the power rating - would it be of the order 1, 10 or 100W?

 

[Baluncore]

Starting with this 0.47 uF value, and a -3 dB cutoff frequency of 2 kHz, the Digikey RC filter calculator suggests a resistor value of 169 ohm.

You do not want any resistance in the path of the current to the motor. A snubber would use a series RC circuit, but not an LPF.

For a 3PH LPF you could use inductors to a delta of capacitors, then the motor. Maybe 5mH and 0u47.

 

[Baluncore]

Here is a crude first guess at what you need.
It has max attenuation at VFD 3 kHz, 0 dB at 50 Hz.
The motor model is a guess, now configured as Y.
Toroids wound for 2 amps, 8mH, wound by the datasheet numbers.
VFD is an AC model, so is a zero impedance voltage source.

 

[Guineafowl]

Here is a crude first guess at what you need.
It has max attenuation at VFD 3 kHz, 0 dB at 50 Hz.
The motor model is a guess, now configured as Y.
Toroids wound for 2 amps, 8mH, wound by the datasheet numbers.
VFD is an AC model, so is a zero impedance voltage source.

Thanks - I see the difference between an RC LPF and the snubber.
I’ve ordered a 20-pack of ferrite rings at those dimensions, and have found my LCR meter.

What I could do with, is a cheap way of measuring THD before and after installing the filter, is that right? I found this:
https://owen-brothers.com/ob975-rogowski-panel-meter.html

 

[Baluncore]

The aim is to remove the sharp steps from the waveform. That will prevent reflections in the cable, and reduce vibration of the motor windings.

Do you have part numbers on the toroids, and a link to the datasheets? The inductors need to carry a couple of amps without saturating. They also need sufficient turns to give 8mH. The 22x14x8mm toroids appears to be the smallest that will satisfy the requirements.

What I could do with, is a cheap way of measuring THD before and after installing the filter, is that right?

I would view the 3PH waveforms with an oscilloscope. Measuring THD may show an improvement, but if it does not improve, it will not tell you why.

 

[Guineafowl]

Do you have part numbers on the toroids, and a link to the datasheets?

Cheap and cheerful, I’m afraid:
https://www.aliexpress.com/item/1005006427487406.html

I would view the 3PH waveforms with an oscilloscope.

I have a Hameg HM203-5 20 MHz CRO, and a Fluke/Phillips PM3384A 100 MHz combiscope, but the digital part doesn’t work. Also, some ordinary 100 MHz probes. I assume I’ll need some better kit.

I’ve been looking at some of the Rigol/Hantek digital offerings. But would I need some special probes?

 

[Baluncore]

Also, some ordinary 100 MHz probes.

Bandwidth or not critical, 1MΩ and 1MHz are sufficient, but the probes should have a x10 voltage switch on them, which also changes them from 1MΩ to 10MΩ impedance.

On the X10 range, the probes can be neutralised with a screw adjustment in the handle of the probe. That is what the 1 kHz square wave test point on the oscilloscope is used for.

Don't disconnect the probe BNC connector from an oscilloscope, or move the X10 switch, while the probe is connected to high voltage.

 

[Guineafowl]

One final problem: the VFD output is mains-referenced, so connecting the probe across phases will trip the RCD. Unless the motor earth will suffice for probe ground, I’ll need an isolation transformer capable of at least starting and running the mill motor at no load. As I understand it, floating the scope is bad practice?

 

[Baluncore]

As I understand it, floating the scope is bad practice?

Yes.

A two channel, A & B scope, usually has a subtraction feature, that makes it possible to view channel A minus channel B. In that case you can view a line voltage with the neutral voltage subtracted, or a delta phase-voltage. That unfortunately, has common-mode problems because the channels are not accurately matched, so you must keep the channel gains carefully balanced. That allows two normal scope probes to be used, while the chassis of the scope is connected to Protective Earth.

The right way to view 3PH lines, is with a three-phase power-scope that has four normal grounded scope-probe inputs, one is a neutral reference that is then subtracted internally from the three channels, all in the front-end amplifiers, before any variable gain stage. That allows the scope to observe 3PH line voltages, A-N, B-N, and C-N, all with a grounded chassis. By alternative cross-connected subtractions, it can also observe the delta phase voltages, A-B, B-C, and C-A.
You can build such a 3PH adaptor by using three op-amps with matched fixed gain. The Y with neutral, or the delta subtraction mode is done by switching the multiple 1M ohm inputs to the amplifiers. That adaptor can be used with any available scope, with all chassis grounded to PE. The differential op-amps can use low-gain for maximum bandwidth, although bandwidth is not usually critical with 3PH power analysis.

This last method will work with MEN power in the UK, where the reference is the neutral, but it is even more dangerous in other systems. Take care.
If you disconnect the mains Protective Earth from the chassis of the oscilloscope, you must place some gas discharge voltage limiter between the PE and the floating chassis inside the scope. That is to protect both the transformer insulation and the operator. You must also flag the instrument as having a floating chassis. The gas discharge arrestors needed are designed for use in twisted-pair telephone networks and data links. A Metal Oxide Varistor (ZnO) is NOT suitable for that task, as it has high capacitance and a limited life.

 

[Guineafowl]

I’m not sure either of these have the subtract function, but willing to be corrected.

My earthing is TN-S, ie N and E linked at the transformer and run as separate conductors to me. No N/E link at my end.

I may have to look for an isolation transformer. I have one mounted in the test bench, but this is a low VA one intended for powering up unloaded SMPSs and the like.

 

[Baluncore]

I’m not sure either of these have the subtract function, but willing to be corrected.

You need to read and understand the manual.

It looks like the Hameg HM-203 has a button to "Invert 1", and a button to "SUM", at the bottom of the panel. That gives you subtraction of the two channels.

I believe the Fluke PM3384A has the necessary switching to make some 3PH observations. You will need to investigate the operator's manual further.

1.1.2 Deflection Modes (Analog Only)
MODES CH1, CH2, CH3, CH4
CH2 and CH4 can be inverted to allow -CH2 or -CH4, CH1 + CH2
CH2 can be inverted to allow CH1 - CH2, CH3 + CH4
CH4 can be inverted to allow CH3 - CH4

 

[Guineafowl]

Yes, it’s been two house moves since I last saw the Hameg manual. As I remember it, the ADD/CHOP button selects dualling options:
ADD: alternate sweeps for 1 and 2
CHOP: switches between 1 and 2 during each sweep.

Will update when I dig out the manual. I’ve never had one for the Fluke.

 

[Baluncore]

Fluke manual.
https://media.fluke.com/c3705bf5-ce6e-4b79-8f55-b10800bf4fdc_original file.pdf

 

[Guineafowl]

Got it. Trace 1 to show CH1+CH2, invert 2 to get CH1-CH2.