A proximity probe watching a shaft is supposed to report how much the shaft is vibrating, but it also picks up things that have nothing to do with vibration: a scratch on the shaft surface, a slightly out-of-round journal, or a patch of residual magnetism in the metal. These imperfections make the probe signal wiggle even when the shaft is barely moving, and that false wiggle is called runout or glitch. Slow-roll runout compensation is the technique of measuring that false signal at very low speed and subtracting it, so the corrected reading reflects real shaft vibration rather than the shaft's surface flaws.
Slow-roll runout in one line: Slow-roll runout compensation is a correction applied to proximity-probe shaft vibration measurements that removes the false signal caused by surface and material imperfections rather than real motion. The combined mechanical and electrical runout, also called glitch, is captured while the shaft turns very slowly and is then vector-subtracted from measurements taken at operating speed. This ensures the reported vibration reflects genuine shaft motion instead of being masked or exaggerated by scratches, out-of-roundness, or residual magnetism.
Runout comes in two flavors, and a proximity probe cannot tell them apart from real vibration on its own. Mechanical runout is caused by the physical shape and surface of the shaft: a journal that is not perfectly round, a shaft that is slightly bowed, a scratch, a dent, or a machining mark all change the gap the probe senses as the shaft turns, producing a signal that looks like motion even if the shaft center is not really moving. Electrical runout comes from the material rather than the shape: variations in the shaft's electrical and magnetic properties, such as residual magnetism or metallurgical inconsistencies, alter the probe's reading even over a perfectly round surface.
The key insight behind the correction is that both kinds of runout are properties of the shaft itself and repeat with each revolution, independent of speed. A scratch is in the same angular position on every turn, and a magnetized spot passes the probe at the same point each revolution, so the false signal they produce is locked to shaft angle and stays essentially the same whether the shaft spins fast or slow. Real vibration, by contrast, generally grows and changes with speed as forces build. This difference in behavior is what makes it possible to separate the two.
That separation is done by capturing the runout at slow roll, meaning while the shaft is barely turning, typically at a very low speed where genuine vibration forces are negligible. At slow roll, almost anything the probe sees must be runout rather than real motion, because there is not enough speed to generate meaningful dynamic vibration. Recording the probe signal over one revolution at this low speed gives a clean picture of the combined mechanical and electrical runout as a function of shaft angle, which becomes the correction to apply later.
Once the slow-roll runout is captured, it is removed from operating-speed measurements by vector subtraction. The runout is described as a vector, an amplitude and a phase relative to shaft angle, and the operating-speed vibration is described the same way, so the correction is a matter of subtracting the runout vector from the measured vector to leave the true vibration vector. Because both are tied to the same shaft-angle reference, the subtraction lines them up correctly and cancels the repeatable false signal while preserving the genuine motion.
Getting this right matters because runout can either hide or fake a problem. If the runout happens to be out of phase with a real vibration component, the two can partially cancel in the raw signal and make the machine look quieter than it is, masking a developing fault. If the runout is in phase, it can inflate the reading and make a healthy machine look like it has a problem, prompting an unnecessary investigation or even a trip. Only after the runout is subtracted does the number represent the shaft's actual dynamic motion, which is the quantity that acceptance limits and diagnostics are meant to be applied to.
The correction is usually applied to the one-times-running-speed component and to the overall waveform, since that is where a fixed shaft-angle glitch has the most influence. It is standard practice on the large machines that use proximity probes, where the acceptance limits are in micrometers of shaft displacement and a few micrometers of uncompensated runout could shift a reading across a zone boundary. Establishing the slow-roll runout at commissioning, and checking it again after any work that could scratch or magnetize the shaft, is part of keeping proximity-probe readings trustworthy.
For runout compensation to be applied continuously rather than only during a manual analysis, the monitoring system needs two things stored together: the slow-roll runout vector captured for each probe, and a shaft-angle reference so the runout can be aligned with live measurements. A once-per-revolution signal from a phase reference provides that angle, letting the system subtract the stored runout at the correct phase on every reading. Keeping the runout vector as a saved property of each measurement point is what allows the correction to be automatic instead of a step someone has to remember to do.
A platform that historizes shaft vibration alongside the phase reference can apply and record the compensation as part of normal monitoring. Merobix brings the proximity-probe signals and the phase reference into a browser-accessible history, so a stored slow-roll runout can be subtracted from operating-speed readings and both the raw and the compensated values are available. This means an analyst reviewing a machine after the fact can see the true, compensated shaft vibration rather than a number contaminated by glitch, without needing to have captured the correction by hand at the console.
Storing the runout in the history also makes it possible to notice when the runout itself changes, which is diagnostic in its own right. A new scratch, a bow that has developed, or a change in magnetization will shift the slow-roll runout, and comparing a fresh slow-roll capture against the stored one flags that the shaft surface or material condition has changed. For remote turbomachinery, keeping the runout vector and the compensated vibration together in one record is what lets an operator trust the shaft readings and recognize when the correction needs to be re-established after maintenance.
Mechanical runout comes from the physical shape and surface of the shaft, such as out-of-roundness, a bow, a scratch, or a machining mark, which change the gap the probe senses as the shaft turns. Electrical runout comes from variations in the shaft's electrical and magnetic properties, such as residual magnetism, which alter the probe's reading even over a perfectly round surface. Both repeat with each revolution and together make up the glitch that compensation removes.
At very low speed the shaft is barely generating any real dynamic vibration, so almost anything the probe sees must be runout rather than genuine motion. This lets you capture a clean picture of the mechanical and electrical runout as a function of shaft angle, uncontaminated by actual vibration. That slow-roll runout is then subtracted from operating-speed measurements where real vibration is present.
Uncompensated runout can either mask or exaggerate the real vibration. If it is out of phase with a genuine component it can partially cancel it, making the machine look quieter than it really is and hiding a fault. If it is in phase it can inflate the reading and make a healthy machine look faulty. Either way the number no longer reflects true shaft motion, which is why compensation matters where limits are in micrometers.
Merobix reads your field devices into a cloud SCADA - the real thing behind these terms, live in days from any browser.