What a Variable Speed Compressor Changes and What It Does Not

Why this matters

A variable speed compressor gets sold as an efficiency upgrade, full stop, and it is frequently installed on a plant where it cannot deliver most of what was promised. The machine is genuinely good at one thing. Everything else the customer was told it would fix, it does not touch. Being able to say which is which, before the order goes in, is the difference between a customer who is pleased two years later and one who has an expensive machine and the same bill. This card is mostly about the second list, because that is the one nobody writes down.

What it actually changes

Power follows flow inside the turndown range. The drive changes shaft speed, so displacement per minute follows demand rather than being throttled or dumped. Across the drive's working range, power runs close to proportional to delivered flow with a modest fixed offset for drive, motor and cooling-fan losses. That removes the unloaded-power penalty that dominates every fixed-speed part-load curve.

The pressure band collapses. A fixed-speed machine has to be given a band to work in, because that band is what stops it cycling. A variable speed machine holds a target pressure closely, so the band can be a fraction of what a fixed-speed machine needs. That matters for a reason separate from the drive: it is what lets you lower average discharge pressure.

Cycling stops. Inside the turndown range the machine neither starts and stops nor loads and unloads, so the storage volume that a fixed-speed machine needs in order to keep its cycle rate legal stops being a constraint on control. Storage still matters for demand transients, which is a different job covered separately in this library.

Two general relationships are needed for the arithmetic later, and both carry the condition they were derived under.

First, a commonly published rule of thumb puts the change in compressor power at about 1 percent for every 2 psi of discharge pressure. That figure is derived for a lubricated rotary screw operating near 100 psig discharge; it is not a constant, it does not hold far from that pressure, and the machine's own curve governs where the two disagree.

Second, every fixed restriction in the system, meaning every leak and every unregulated point of use, is passing choked flow at plant pressures, so its mass flow varies in direct proportion to upstream absolute pressure. That relationship and its condition belong to the pressure and flow control card in this library; it is stated here because the example uses it.

What it does not change, which is the longer list

It does not improve full-load specific power. At the top of its range a variable speed machine is usually slightly worse than a comparable fixed-speed machine, because the drive and the fan take their cut and the airend is running at a speed it was not optimised for. A plant that runs near capacity most of its hours is buying a small penalty, not a saving.

Below the turndown floor it reverts. Every drive has a published minimum speed, below which the machine stops, blows off, or cycles like a fixed-speed machine. In that region a variable speed machine is not better than a load/no-load machine and can be worse, because it has to cycle a drive as well as a compressor. A plant whose demand spends much of its time below the floor has bought the wrong machine, and the floor is a published number you can ask for before ordering.

It does not reduce leaks. A leak is a fixed restriction passing air; the drive simply makes that air more attentively. A variable speed machine on a leaky plant runs at a higher speed all shift and reports itself as correctly loaded.

It does not reduce artificial demand. Unregulated tools and blow-offs consume what the pressure lets them consume. The drive holds pressure steady, which is the opposite of reducing it.

It does not fix pressure drop through distribution. If the plant is starving because of undersized pipe, a filter overdue for change, or a restricted coupler, raising the compressor's attentiveness at the discharge does nothing at the tool. That trace belongs to the distribution pressure drop card.

It does not fix a flat base load. This is the one that decides most retrofit decisions and it is the least discussed. A drive can only save on the part of the load that varies. Whatever the plant draws every minute of every shift is served at a speed that never changes, and on that portion the machine is a fixed-speed machine with drive losses added.

It does not remove the need for a controls plan on a multi-machine plant. A drive on the base machine wastes most of its range, because the base machine is the one that runs flat out. The drive belongs on the trim machine, which is the one that has to follow the swing.

Worked example: a retrofit where the drive is not the main benefit

A plant with one machine, capacity taken as 100 units. Demand over the shift runs between 55 and 95 units, averaging 72. The current machine is load and no-load with adequate storage, published unloaded power 0.25 of full load. The proposal is a variable speed replacement.

What the drive is allowed to work on. Demand never falls below 55, so 55 of the machine's 100 units are a base that does not vary. The drive can only act on the top 45 units of range, and the average sits 17 units into that 45.

Part-load power on the existing machine. Using the same linear form the control-strategy card sets out, power equals 0.25 plus 0.75 times output. At 0.72 output that is 0.79 of full load.

Part-load power on the drive. Take the drive machine's curve as power equal to 0.08 plus 0.92 times output; this is illustrative and your machine's published curve governs, with the offset covering drive, motor and fan losses. At 0.72 output that is 0.742.

The speed-control saving alone. 0.79 minus 0.742 is 0.048, on a base of 0.79, which is 6.1 percent. That is the entire benefit of the speed control on this plant, and it is small, because the demand band is narrow and sits high.

Now the pressure. The existing machine runs a 10 psi band, 105 to 115 psig, so its average discharge is about 110. The plant's worst point of use needs a pressure that, with the measured distribution drop added back, puts the minimum acceptable header at 100 psig. The drive machine can hold 102 with a couple of psi of swing, so average discharge falls by about 8 psi.

At the stated rule of thumb of 1 percent of power per 2 psi near 100 psig discharge, 8 psi is about 4 percent of compressor power. And by the choked-flow relationship, absolute upstream pressure falls from about 124.7 to about 116.7 psia at sea level, a ratio of 0.936, so loss flow through every leak and unregulated point drops about 6 percent. Less air to make on top of less power per unit made.

The honest conclusion. On this plant the pressure reduction the tight band permits is worth roughly as much as the speed control itself, and the two together are worth about a tenth of the compressor's energy rather than the transformation the customer was expecting. If the customer's real complaint is the bill, the leak survey and the regulator work come first and cost far less, and this library covers both.

What would flip it. Widen the demand band. If the same plant's demand ran from 20 to 95 units instead of 55 to 95, the average drops and the fixed-speed penalty at the low end grows fast, because the load/no-load curve at 0.20 output is 0.40 while the drive curve is 0.264. That is a 34 percent gap in that region rather than a 6 percent one, and the drive becomes the right answer on its own merits.

Narrow it the other way and the drive gets worse, not merely less good: a plant running 90 to 100 units would pay drive losses at the top of the range for a saving that is close to nothing.

The failure mode. Sizing the drive machine to the plant's peak and never checking the minimum. A machine whose turndown floor sits at 40 units, installed on a plant that drops to 20 units on the second shift, cycles all night, and the customer's complaint arrives as "the new compressor is noisier than the old one," which is what cycling sounds like from the office.

Checking a proposal before it is ordered

Ask for three published numbers and one measurement, and refuse to model anything without all four.

The three numbers are the drive machine's power-versus-flow curve including its offset at minimum speed, its turndown floor as a percent of capacity, and its full-load specific power against the machine it is replacing. The measurement is the plant's own demand distribution, not its peak: at minimum, the lowest sustained demand over a full week including nights and weekends, because that is the figure that tests the turndown floor and it is the one nobody collects.

Then run the demand distribution against both curves before anyone quotes a percentage. If the answer comes out under about a tenth of compressor energy, say so, and put the leak and pressure work in front of it rather than behind it.

Safety note for the retrofit itself

A variable frequency drive holds charge in its DC bus after the disconnect is opened. Work inside the drive enclosure follows 29 CFR 1910.333(b)(2) rather than 29 CFR 1910.147, with dead proved by the live-dead-live sequence of NFPA 70E-2021, 120.5 where your employer's electrical safety program adopts it, and only after the manufacturer's stated bus discharge time has elapsed. Mechanical work on the airend, the receiver or any pressurised joint is the other fork: lock the disconnect open under 1910.147, isolate, bleed the leg, and confirm the receiver's indicating gauge reads zero before breaking a joint.

References

• Manufacturer documentation for the drive machine's power-versus-flow curve, minimum speed and turndown floor, full-load specific power, and DC bus discharge time
• 29 CFR 1910.333(b)(2) for work inside the drive enclosure; 29 CFR 1910.147 for mechanical isolation and stored air energy; NFPA 70E-2021, 120.5 for the live-dead-live sequence as adopted by your employer's electrical safety program
• See related: What a Compressor Control Strategy Is Choosing Between; Why Part Load Behaviour Decides the Energy Bill; What a Pressure Flow Controller Is Actually Doing; How to Sequence Multiple Compressors Without Short Cycling