A servo drive controls a motor through a stack of three nested feedback loops, each wrapped around the one below it. In the middle sits the velocity loop, the part of the drive that decides how much effort to apply to make the motor run at a commanded speed. It takes a speed command from above, compares it with the actual speed measured from feedback, and drives the difference toward zero. This guide explains what the velocity loop does, how it uses its proportional and integral gains and its bandwidth, and how it fits between the position loop that surrounds it and the current loop it commands.
Velocity Loop in one line: The velocity loop is the middle of the three nested control loops in a servo drive, responsible for regulating the motor's actual speed to match a commanded speed. It compares a velocity command with the velocity measured from the feedback device, and its proportional and integral gains turn the error into a torque command that it passes down to the current loop. It sits inside the outer position loop, which supplies its speed command, and outside the innermost current loop, which carries out its torque request.
A servo axis is usually controlled by three loops arranged one inside another. The outermost is the position loop, which looks at where the axis should be versus where it is and produces a speed command to close the gap. That speed command feeds the velocity loop in the middle, which regulates actual speed and produces a torque command. The torque command feeds the innermost current loop, which regulates the motor current that produces the torque. Each loop trusts the one below it to carry out its request, so the position loop asks for a speed, the velocity loop asks for a torque, and the current loop delivers the current.
This nesting is deliberate and it works because the inner loops are faster than the outer ones. The current loop is the fastest, reacting in a fraction of a millisecond, the velocity loop is next, and the position loop is the slowest of the three. Each outer loop can treat the loop inside it as an obedient, near-instant follower, which keeps the whole system stable and lets each loop focus on one job. The velocity loop's job in this arrangement is purely to make the motor turn at whatever speed it is told, as accurately and quickly as it can.
Not every application uses all three loops. A drive running in pure velocity mode uses the velocity and current loops but no position loop, taking its speed command directly from an external source. A positioning axis adds the position loop on top. Either way the velocity loop is central, because good position control depends on a velocity loop that responds crisply and holds speed steadily, and a sloppy velocity loop shows up as poor following and settling at the position level.
To regulate speed the velocity loop needs to know the actual speed, and that almost always comes from the same feedback device used for position. Most servos measure position with an encoder or resolver and derive velocity from it, essentially by observing how fast position is changing. The quality of that derived velocity signal matters a great deal: at low speeds or with coarse feedback the derived velocity can be noisy or grainy, and that noise limits how hard the loop can be pushed. Higher-resolution feedback yields a cleaner velocity signal and lets the loop run tighter.
The velocity loop is typically a proportional-plus-integral controller. The proportional gain responds to the present speed error, providing immediate stiffness so the motor resists disturbances and follows speed changes promptly; a higher proportional gain makes the loop stiffer and faster but eventually excites noise and resonance. The integral gain accumulates error over time and drives any steady speed offset to zero, so the motor holds exactly the commanded speed under a constant load rather than settling a little short. Together the two make the loop both responsive and accurate.
Tuning these gains is the heart of setting up a servo. Push the proportional gain up for stiffness and bandwidth until the motor starts to buzz, whine, or oscillate from noise and mechanical resonance, then back off to a safe margin. Set the integral action fast enough to remove steady error promptly but not so fast that it causes overshoot or hunting. The load's inertia relative to the motor's, and any compliance or resonance in the mechanics, set the ceiling on how high the velocity gains can go, which is why the velocity loop is where much of the tuning effort in a servo system is spent.
The velocity loop's bandwidth is a measure of how quickly it can respond to a change in commanded speed or reject a disturbance, and it effectively caps the performance of everything built on top of it. A high-bandwidth velocity loop follows rapidly changing speed commands closely and shrugs off load disturbances, which lets the position loop above it be tight and accurate. A sluggish velocity loop, by contrast, lags its speed command and lets disturbances linger, and no amount of position-loop tuning can fully make up for it because the outer loop can only be as good as the inner one it relies on.
There is a practical hierarchy in tuning that follows the loop nesting: get the current loop right, then tune the velocity loop for the stiffest stable response the mechanics allow, and only then close and tune the position loop. Because the velocity loop determines the axis's stiffness against disturbances and its responsiveness to speed changes, it is where symptoms like sluggishness, overshoot, or audible resonance are usually diagnosed and cured. Notch filters are often applied within or just below the velocity loop to suppress a specific mechanical resonance so the gains can be raised further.
The velocity loop itself runs deep inside a drive at speeds no supervisory system touches, but its consequences are exactly the kind of thing a plant watches through cloud SCADA. Speed error, torque demand, and following performance reported by drives can be trended and alarmed by a platform such as Merobix across many machines and sites. A velocity loop that is drifting out of tune as a mechanism wears, or a drive that is working harder to hold speed against a rising load, reveals itself as a change in these signals over time. Bringing drive speed and torque telemetry into a central view lets a maintenance team spot a machine whose velocity control is degrading long before it fails, even at a remote installation nobody visits daily.
In most servos the velocity loop derives speed from the same position feedback device, such as an encoder or resolver, by measuring how fast position is changing over time. Some systems use a dedicated velocity sensor like a tachometer, but deriving velocity from a high-resolution encoder is the common approach today. The cleanliness of that derived velocity signal, especially at low speed, limits how tightly the loop can be tuned.
Nesting a faster inner loop inside a slower outer loop is what makes the arrangement stable and effective. The position loop only needs to command a speed and trust that speed to be achieved quickly, which the faster velocity loop provides. If position were inside velocity the fast loop would be chasing the slow one, which does not match how the physics works, since applying torque changes speed and speed changes position, so the loops naturally stack current, then velocity, then position.
Pushing the velocity loop's proportional gain too high makes the motor buzz, whine, or oscillate as it amplifies feedback noise and excites mechanical resonances, and too much integral action causes overshoot and hunting around the commanded speed. The practical ceiling on the gains is set by feedback noise and by the load's inertia and mechanical resonance, so tuning means raising gains for stiffness until these symptoms appear and then backing off to leave a safe margin.
Merobix reads your field devices into a cloud SCADA - the real thing behind these terms, live in days from any browser.