A normal servo closes all its loops around a single encoder mounted on the motor, which means it knows precisely where the motor shaft is but only assumes where the load is. Whenever there is slack or flex in the mechanism between the motor and the load, that assumption is wrong. Dual-loop feedback fixes this by using two encoders: one on the motor and one on the load. This guide explains how a dual-loop servo splits the job between the two devices, why doing so lets it correct for backlash and mechanical compliance, and how it ends up positioning the actual load rather than just the motor shaft.
Dual-Loop Feedback in one line: Dual-loop feedback is a servo control arrangement that uses two feedback devices: a motor-mounted encoder that closes the fast velocity loop and a load-side encoder that closes the position loop. Because the position loop measures the actual load rather than the motor shaft, the servo corrects for backlash and coupling compliance between the two, positioning the load accurately even when the mechanism between motor and load is not perfectly rigid.
In a conventional single-loop servo, one encoder on the motor provides feedback for all the control loops. The velocity loop derives speed from it and the position loop reads position from it, so the servo controls the motor shaft with great precision. The unspoken assumption is that the load moves exactly as the motor does, which is true only if the coupling between them is perfectly rigid and free of play. In the real world, gearboxes have backlash and shafts, belts, and couplings flex under load, so the load does not perfectly mirror the motor, and a single-loop servo cannot see that difference.
A dual-loop servo adds a second encoder at the load, and it divides the work between the two devices according to what each is good at. The motor-mounted encoder, being rigidly attached to the motor with no mechanism in between, gives a clean, fast, well-behaved signal ideal for the velocity loop, which needs high bandwidth and freedom from the delays and wobbles that mechanics introduce. The load-side encoder, mounted where the actual work happens, tells the position loop where the load truly is, which is what the machine ultimately cares about.
This division is the key insight of dual-loop control. The fast inner velocity loop stays on the motor where the feedback is crisp and stable, so the loop can be tuned tight and remains well-behaved. The slower outer position loop closes on the load, so the quantity it regulates to a target is the real load position. The servo therefore gets the best of both: a stable, high-performance velocity loop and a position loop that controls the thing that matters.
The reason dual-loop is worth the extra encoder is that it sees through the imperfections of the mechanism. Consider backlash in a gearbox: with a single motor-side encoder, when the axis reverses the motor turns across the backlash gap while the load stays still, but the encoder reports motion, so the servo believes the load has moved when it has not. A load-side encoder reports the truth, that the load is not yet moving, so the position loop keeps driving until the load actually reaches its target, effectively taking up the backlash as part of closing the loop rather than merely guessing at it.
Compliance, the flexing of shafts, belts, and couplings under torque, is handled the same way. Under load a compliant coupling winds up like a spring, so the load lags the motor by an amount that depends on the force. A motor-side encoder cannot know how much the coupling has flexed, so it positions the motor correctly while the load sits short. A load-side encoder measures where the load has actually ended up including the wind-up, so the position loop corrects for it and drives the load to the true target regardless of how much the coupling has deflected.
This makes dual-loop the natural choice for machines where accuracy at the load is paramount but the mechanism cannot be made perfectly rigid, such as geared axes, long or belt-driven axes, and machines with significant compliance. It is a more complete answer to backlash than a compensation offset, because rather than predicting and subtracting a fixed amount of lost motion, it actually measures the load and closes the loop on it. The trade is the cost and complexity of the second encoder and the extra care needed in tuning, because the compliance and backlash that dual-loop corrects also introduce dynamics that the loop must be tuned to handle without instability.
Choosing dual-loop is a decision about where accuracy is needed and how good the mechanics are. If the coupling between motor and load is stiff and free of play, a single motor-side encoder is simpler, cheaper, and entirely sufficient, and adding a load encoder buys little. Dual-loop earns its keep when there is real backlash or compliance between the motor and the load and the application demands that the load, not the motor, be positioned accurately. Machine tools, precision stages, and geared positioning axes are typical homes for it.
Tuning a dual-loop system deserves respect because the very flexibility it corrects for also complicates the control. The compliance between the motor and the load creates a resonance, and closing the position loop on the far side of that resonance can make the system harder to stabilise than a rigid single-loop axis. Good dual-loop tuning keeps the velocity loop fast on the well-behaved motor feedback and closes the position loop on the load carefully, sometimes with filtering, so the servo gets load accuracy without exciting the mechanical resonance into instability.
The two feedback signals in a dual-loop axis are also a rich source of diagnostic information for a plant monitored through cloud SCADA. The difference between where the motor thinks it is and where the load actually is directly measures the backlash and compliance in the mechanism, and that difference tends to grow as gears and couplings wear. A platform such as Merobix can trend load position error and following performance reported by machines across many sites, so a mechanism whose lost motion is creeping up over months shows itself in the data before it spoils product or fails. For distributed equipment, watching that motor-to-load discrepancy centrally turns the second encoder into an early warning of mechanical wear at sites a maintenance team rarely visits.
The fast inner velocity loop closes on the motor-mounted encoder, because that feedback is clean and free of the delays and wobble the mechanism introduces, which lets the velocity loop be tuned tight and stable. The outer position loop closes on the load-side encoder, so the quantity regulated to a target is the actual load position. This split gives a stable high-performance velocity loop while positioning the thing the machine really cares about, the load.
Backlash compensation predicts the lost motion and adds a fixed offset on each reversal to take up the slack, without ever measuring the load. Dual-loop feedback actually measures the load with a second encoder and closes the position loop on it, so the servo drives until the load itself reaches the target, correcting backlash and compliance directly rather than by prediction. Dual-loop is more complete and handles varying compliance, but it costs an extra encoder and needs more careful tuning.
A single motor-side encoder is enough when the mechanism between the motor and the load is stiff and essentially free of backlash, so the load moves as one with the motor and there is no meaningful discrepancy for a load encoder to reveal. It is simpler, cheaper, and easier to tune. Dual-loop is worth its extra encoder and complexity only when real backlash or compliance exists between motor and load and the application requires the load itself to be positioned accurately.
Merobix reads your field devices into a cloud SCADA - the real thing behind these terms, live in days from any browser.