Air Compressor Motor Humming But Not Starting: Capacitor and Winding Issues

Single-phase AC makes a magnetic field that oscillates back and forth. Not rotating. Just back and forth along one axis. Equal force both directions on the rotor, so net torque is zero. The rotor vibrates at line frequency, and that vibration is the hum.

Every single-phase compressor motor has a start winding and a start capacitor to get around this. The capacitor shifts the current phase in the start winding relative to the main winding, which fakes a rotating field for the first second or two while the rotor accelerates. Once speed is up past about seventy percent of synchronous, a centrifugal switch or relay disconnects the start winding. Motor runs on the main winding from there.

If the motor is humming, the start sequence failed somewhere.

Going to spend more time on this section than anything else in the article because this is where probably eighty percent of humming problems originate, and it's also where the most diagnostic mistakes happen.

Somebody puts a multimeter on capacitance mode. Gets a reading. "Shows 260, rated 270, that's fine." And then they move on to the windings or the switch or whatever else, and waste an hour before coming back to the cap. Because 260 might be fine. Or it might not be. That depends on what the motor needs.

The cap value on a given motor isn't arbitrary. It's calculated against the start winding impedance at locked-rotor conditions to get the phase angle between the two winding currents close to 90 degrees, which is where starting torque peaks. Shrink the cap, the phase angle tightens, torque drops. How much torque drop matters depends entirely on what the motor is trying to start against. Empty tank? Maybe 260 is enough. Tank has 40psi residual because the unloader valve is sluggish? Now 260 isn't enough and the motor sits there humming and the guy who checked the cap already said it was fine.

A cap that's drifted to around seventy percent of rated value will pass a continuity check, will charge and discharge normally, will look completely healthy on every test except an actual capacitance reading compared against the spec. And the motor does this maddening thing: starts sometimes, doesn't start other times, no apparent pattern. People call it intermittent. The word "intermittent" gets applied to this situation constantly and it's wrong. Nothing is intermittent. The cap is at the ragged edge of being able to start the motor. Whether it makes it depends on how much pressure is sitting on the piston at that moment, which varies depending on where the piston stopped and whether the unloader valve bled off fully. Sometimes there's enough margin, sometimes there isn't.

This part is going to be long. Bear with it because this is maybe the single most common root cause of compressor starting problems and it barely gets discussed anywhere.

Electrolytic start caps carry ±20% tolerance. Printed on the label. A 270μF cap is in spec from 216 to 324. Budget manufacturers build to the bottom of that band. Less dielectric material, less cost. Brand new, still sealed, never installed: meter reads 218. Maybe 235 on a good one. Within spec. Nobody can make a complaint.

Had a situation maybe three years ago, customer bought a compressor, mid-range brand, nothing cheap. Motor started humming on loaded starts within about eight months. Pulled the cap, measured it, 198μF on a 270 rated. Normal aging? In eight months? Pulled a cap off a brand new unit of the same model sitting on the floor, same brand compressor, measured it. 221. Factory installed. That's what the motor was running on from day one. The design engineer's calculations assumed something closer to 270. The purchasing department's cost negotiations delivered 221. The motor worked, because 221 was still enough on a good day with a warm garage and a fully functional unloader valve. Eight months of aging took it below the line.

Put in an eighteen dollar cap from a supplier that doesn't build to the floor of the tolerance band. Measured 278. Compressor hasn't had a starting problem since. That was three years ago.

The eighteen dollar cap versus the six dollar cap: it's foil layers and electrolytic paper thickness. More material, more capacitance, more margin. The six dollar cap has the minimum material needed to pass quality control at the tolerance floor.

Then there are the counterfeits, which are a different level of bad. Marketplace listings, brand name on the label, guts don't match at all. Labeled 270, measures 160 or 170. Labeled 330V, dielectric punctures somewhere around 280 if you're lucky. Install one, loaded start fails. Unloaded it might barely crank, which is just misleading enough to send someone looking elsewhere. Few weeks later the thing blows. Sometimes the top pops off.

The temperature thing catches people every fall and the same guys get fooled by it year after year because the failure condition and the test condition are at different temperatures and nobody thinks about that.

Electrolyte ionic conductivity drops with temperature. How much depends on the specific formulation, but the ballpark is something like 30-40% loss near freezing. The exact numbers vary by manufacturer. Some are worse. The cheap ones tend to be worse because the electrolyte formulation is also cost-optimized.

The pattern goes like this: Compressor ran all summer, no issues. First cold morning, maybe October, maybe November depending on where you are, motor hums. Doesn't start. Owner waits, tries again, nothing. Goes inside, does something else, comes back out at noon, garage is warmer, compressor starts first try. That evening or the next day, pulls the cap, brings it into the house, puts a meter on it. 268μF. Fine. Puts it back. Next cold morning, same problem.

The cap reads fine indoors because it IS fine at room temperature. At 4°C in the garage at 6 AM it was putting out maybe 175. Not enough.

Some people make two, three service calls on this one. New cap, tests fine, works fine for the day because by the time the service call happened the garage was warm. Next cold morning, same thing. Come back out, test cap again, fine again. If nobody thinks to ask "what was the temperature when it failed" versus "what's the temperature right now while I'm testing it," this can go on indefinitely.

Fix: higher rated cap to absorb the cold loss, or a hard start kit that adds parallel capacitance through a PTCR thermistor.

Two situations where a hard start kit belongs: this cold weather scenario, and an unloader valve that doesn't fully relieve pressure. Outside those two, a hard start kit is hiding something.

Quick one. The start winding is inductive. When the centrifugal switch opens and interrupts the current, the inductance kicks back and adds voltage across the cap. On a 230V motor the spike can exceed 350V. That's why 330V rated caps are spec'd.

Using a 220V cap because it's what's on the shelf will work. Each start pushes the dielectric past its rating. Capacitance degrades, ESR climbs, nobody notices. Months later, harder to start. New cap fixes it.

There's also an LC resonance issue. Start winding inductance plus start cap form a series LC circuit. If the cap value is off far enough from original, the resonant frequency shifts toward 60Hz. Near resonance, the voltage across the cap hits multiples of line voltage. 330V cap seeing 600-plus volts. One attempt, done.

Not nearly as much to cover here. The run cap stays in the circuit during operation on CSR motors. It has two failure modes and both of them are diagnostic time sinks, but for different reasons.

Shorted run cap: it's in parallel with the start winding and start cap in series. Short turns it into a bypass that steals current from the start circuit during starting. Test the start cap, it's good. Measure the start winding, it's good. Everything checks out and the motor hums. If someone thinks to check the run cap it's a five minute diagnosis. If they don't, it can be an hour of re-testing components that already tested fine.

Open run cap: motor keeps starting, keeps running, nobody notices for weeks. Efficiency drops, current is up, main winding runs hotter. Thermal protector starts tripping. Each cycle starts from a higher temperature baseline because the cooling time isn't enough. Winding accumulates heat damage. One morning it fails. Looks like a start problem. The run cap went open some unknown number of weeks ago.

Winding diagnosis is where things get less clean. Capacitors are cheap, testable in a minute, and either good enough or not. Windings exist on a spectrum. Partially degraded insulation, minor leakage between phases, a few shorted turns. Readings land in gray areas that require judgment and context that a lot of people doing field diagnosis don't have.

Not going to pretend to cover winding diagnosis as thoroughly as the capacitor section above. Capacitors are straightforward to write about because the diagnostics are straightforward. Windings get murky fast.

C to R: main winding. Low resistance, thick wire. C to S: start winding. Higher resistance, thinner wire. S to R: both windings in series. Should be approximately C-R plus C-S.

That third measurement, S to R, gets skipped so often it's almost a running joke among rewind shop guys. Two measurements, both look normal, move on. But S to R is the one that catches inter-winding insulation leakage. If it reads noticeably lower than the sum of the other two, current is leaking between the windings. Miss this and the motor goes back in and burns again. Three measurements. The third one is the one that matters most and the one that gets skipped most. Go figure.

The other thing about resistance readings: a number by itself means nothing. 3.2 ohms. Good? Bad? Depends on the motor. That's normal on one model and means shorted turns on another. Need factory spec to compare against. Or, better, a baseline taken when the motor was healthy and stuck on a piece of tape in the junction box. Almost nobody takes baseline readings. One of those things that would save hours down the line and takes five minutes to do and still doesn't get done.

Adjacent turns lose their enamel, make contact, form a shorted loop that circulates high current and generates localized heat. Effect on total winding resistance can be negligible when only a couple turns are involved. The resistance change hides in the meter's measurement noise. Numbers look fine.

Surge comparison tester is the right tool. Fast pulse, examine the oscillation waveform, shorted turns shift the frequency. Clear result. But the equipment is specialized and expensive for what it is. Most shops don't own one.

Field method: power on for two or three seconds, kill it, immediately sweep an IR thermometer over the stator housing. A shorted-turn hot spot shows up as a localized temperature difference before conduction has time to spread the heat through the iron. The window is maybe five seconds. After that the temperature equalizes and there's nothing to see. Works for locating shorts that are already severe. A two-turn short might not throw enough heat through the stator mass to register on the outside surface.

Go pick up a 5HP compressor motor from around 2005 and one from last year. Same nameplate. The old one is noticeably heavier. Thicker wire, more turns, deeper stator slots.

Both hit the same specs on a test bench. The difference is thermal margin during normal operation. The old motor's winding temperature sat maybe twenty-five, thirty degrees below the insulation class limit. That gap absorbed problems: low voltage, hot ambient, aging run cap bumping current up. Temperature rose but stayed below the limit.

Current production has squeezed that gap down to ten, maybe twelve degrees. The same conditions that the old motor ignored now push the new one into the accelerated aging zone. Few years of that and the insulation starts going.

It's copper cost. Every manufacturer has done it. Not a quality issue in the sense of defective product. The motors pass validation. The margin just isn't there anymore. What it means practically is that the capacitor, the voltage supply, the ambient temperature, all of those things matter more than they used to. The motor can't absorb sloppy conditions the way the old ones could.

Should probably have put this closer to the cold weather capacitor section since they're related, but moisture is a winding issue more than a capacitor issue, so it landed here.

Compressor in an unheated garage, October through March, goes through dew point cycling. Night temperature drops below dew point, moisture condenses on winding insulation. Day temperature comes up, some evaporates, not all. Over a winter the insulation's moisture content creeps up. Dielectric strength creeps down. Nothing looks wrong from outside.

Spring comes. Owner needs the compressor. Insulation is at peak moisture content. Cap might still be cold. Oil is thick. Switch goes on. Hum. Switch off, wait a few seconds, switch on. Hum. Each locked-rotor attempt pushes five to seven times running current through insulation that's at a fraction of its rated dielectric strength.

Rewind shops see the result every April and May. The intake fills up with burned motors. Volume drops in June. The guys working the counter know what month it is by what's coming through the door.

Running the compressor for a few minutes once a month over winter prevents this. The operating heat drives moisture out. If it sat idle all winter, at least run it unloaded for several minutes before the first loaded start.

A regular multimeter in resistance mode reads "OL" on both healthy insulation and insulation that's about to fail. Same display. Useless for this.

A megohmmeter puts 500V DC across the insulation and measures leakage current. Winding to winding, winding to frame.

Over 100 megohms, clean. Between 5 and 100, aging. Below 2, actively failing. Below 1, condemned.

For borderline readings, a polarization index test: leave the 500V applied, record at one minute, record at ten minutes, take the ratio. Above 2, the insulation is structurally sound and the low initial reading was surface moisture being pushed off. Below 1.5, the leakage is permanent. Drying won't change it.

The tool is not expensive. The test takes ten minutes. It doesn't get used in compressor work as often as it should, mostly because it's not part of how people typically learn this.

Locked-rotor current is several times running current. Heat generation in the winding goes with current squared. A stall doesn't just run warm. The heating is extreme. Thermal protector trips in a few seconds, but by then the winding temperature has already spiked hard.

Thermal protector resets in ten, fifteen seconds. Operator tries again. Winding didn't cool. Starts from the elevated baseline. Three or four rapid stalls can push the start winding past its insulation class rating on a motor that was healthy before the first attempt.

Typical pattern: cap failed, owner tried five or six times in quick succession, gave up, called for service. Tech replaced the cap, motor started, problem solved. Except the start winding took five or six locked-rotor events without cooling. Nothing visible. Resistance won't show it unless a short formed. Winding life is drastically shortened. Burns a few months later. Gets treated as a separate problem.

Cap is fine, windings measure okay, still humming. Check the switch.

Open-frame motors have a centrifugal switch on the shaft. Hermetic compressors use external potential or current relays.

Stuck open is the common one. Spring broke, contacts oxidized, oil mist gummed up the flyweights. Start circuit is disconnected before the motor is even energized. Everything tests fine because everything IS fine. The problem is a five dollar switch nobody looked at.

Contact degradation happens gradually. Every time the switch opens at speed, the collapsing start winding field arcs across the separating contacts. After several thousand cycles the contact faces are pitted and carbon-coated. Mechanism works. Contacts touch. Electrical conduction is unreliable. Motor starts some days, hums other days. No correlation with anything. Whether it works depends on the microscopic contact condition at the instant the surfaces meet. Fine sandpaper fixes it.

A dead relay versus a degraded cap: the relay fails identically every time. Hundred attempts, hundred hums. A degraded cap varies with load and temperature. If failure is perfectly consistent, the switch or relay is more likely than the cap.

Stuck closed is the opposite and it destroys components. Start cap stays in circuit during running. It's a one-to-two-second duty component. Minutes of continuous current overheats and kills it. Start winding overheats too. Motor was fine last week, now hums, because the stuck switch burned the cap. New cap, works a few days, burns again. That repeat failure pattern means check the switch.

Auxiliary cap plus PTCR thermistor relay. PTCR conducts cold, resistance spikes as it heats from current flow, drops the extra cap out after a few seconds. More starting capacitance, more torque.

Parts: ten, twelve dollars. Boxed product: forty to eighty. The margin is in the packaging.

Good for cold weather starting and unloader valve issues. On a motor with winding damage, the kit does nothing useful. The torque deficit on a motor with shorted turns or degraded insulation comes from corrupted field phase relationships. More capacitance can't recover phase angle that shorted turns are consuming internally. The kit lets a damaged motor keep running until the damage becomes total. That's not a fix. That's delayed failure.

After shutdown, the unloader should dump residual pressure from the cylinder head and discharge piping. Stuck closed, the next start fights trapped pressure. Single-phase starting torque can't overcome it on most compressors.

Check: drain tank to zero. Close drain. Try to start. If it goes, unloader is the problem. Simple test that gets skipped when people go straight to electrical diagnosis.

Starting torque follows voltage squared. Ten percent voltage drop costs about nineteen percent torque. That catches people off guard because it seems like too much loss for too little voltage change. But that's how squaring works.

Extension cords, undersized circuits, shared circuits. The voltage drop is invisible at no-load. Only shows up when locked-rotor current, which is several times running current, is flowing through the wire resistance.

The measurement trap: checking voltage at the wall outlet with nothing running and getting a good number. That number doesn't mean anything. The voltage that determines whether the motor starts is at the motor terminals during the stall attempt. That can be 30 or 40 volts below the unloaded outlet reading on a bad installation. Have to put the meter on the motor terminals and read it during the start attempt while the heavy current is flowing. That's the number.

The instinct is to start checking electrical things. Before any of that, try turning the motor by hand with the power off. Grab the flywheel and rotate it. If it won't budge, the problem is mechanical. Seized piston, locked bearing, broken valve plate. Doesn't matter what the capacitor reads.

After that, honestly, the order varies depending on what information is already available. If the owner says "it used to start fine but it's been getting worse over several months," that points toward capacitor aging. If the owner says "it stopped working all at once Tuesday morning," that points toward a switch or relay. If it only fails on cold mornings, temperature-related capacitance loss. If it fails only when the tank has pressure, unloader valve. The diagnostic sequence isn't really a fixed sequence. It starts with whatever the symptom pattern suggests.

That said, if there's no useful history and the motor just hums with no other clues, a reasonable starting point:

Check terminal voltage during a start attempt. If the voltage is way low, everything else is secondary until the supply is fixed. Test the start capacitor, capacitance against rated value. If more than about 15% low, replace. Look for physical damage on the cap body. Check the run cap if the motor has one, mainly for shorts. Look at the centrifugal switch or relay contacts. Power off, check default position, inspect for pitting and carbon. If failures are intermittent, focus here. Winding resistance across all three terminal pairs. If the third measurement doesn't add up, there's leakage. Megohmmeter if everything above checks out. Below 2 megohms is a concern, below 1 is condemned. Drain the tank and try starting to check the unloader valve.