Introduction: A machine DRO can rewrite the number it shows so the display matches the distance the table actually travelled, but only after someone measures how the axis really behaves.
A handwheel dial tells you what the leadscrew was supposed to do. A digital readout tells you what the axis actually did, and on an older lathe or mill those two answers drift apart in patterns you can learn to recognise. Some of that drift grows steadily as the table travels; some of it appears only near one particular stretch of the bed. The console cannot guess which shape it is looking at. Someone measures the machine, and the shape of that measurement decides whether linear compensation or nonlinear compensation is the right tool.
Uniform Pitch Error Along a Leadscrew
A leadscrew is a long threaded part, and no long threaded part is perfect. Along its length the thread pitch can run slightly fine or slightly coarse, and that small error per turn adds up as the nut travels. What matters is the shape it makes: the error accumulates like a rate rather than a step. A familiar shop observation is that a 300 mm move lands a couple of hundredths short and a 600 mm move lands twice as far short, with the mistake growing in proportion to the distance commanded. Measuring at 100 mm and again at 500 mm puts two points on the same sloping line. That behaviour is what a machinist means by uniform pitch error. It is built into the length of the screw, so it repeats pass after pass, it does not care how fast the carriage moves, and it usually survives a change of direction. On a screw with modest, steady error the pattern shows up across the whole travel range rather than in one spot. Because the deviation is a constant share of the distance commanded, one number can describe it well enough to be useful.
Linear and Nonlinear Compensation Correct Different Error Patterns
Two methods exist because machine error is not one shape. A console that offers a percentage correction handles rates. A console that stores a table of points handles places. It is tempting to treat these as beginner and advanced versions of the same feature, but they answer different questions: how much does the error grow as I travel, and where along the travel does the error change character? If you measure an axis and the numbers keep climbing at a steady rate, you are looking at the first question. If the numbers sit flat and then jump near one section of the bed, you are looking at the second. Choosing the wrong shape wastes the correction and can leave the display further from truth than it started.
1. When a Single Correction Factor Fits Uniform Pitch Error
Linear compensation applies one correction factor per axis. The technician enters a value representing the measured error over a known length, and the console scales every displayed position by that factor, spreading the correction evenly from zero to full travel. It fits when the error is proportional: 0.02 mm short at 200 mm, 0.04 mm short at 400 mm, 0.06 mm short at 600 mm. The scaled display then reads close to truth across the whole range without any stored points. The distinguishing test is simple arithmetic — the deviation measured halfway along the travel comes out near half the deviation at full travel. When that relationship holds, a table of points adds setup work without improving the result, and a single factor keeps the console easy to hand over between operators.
2. How Segmented Corrections Address Local Guideway Deviation
Nonlinear compensation stores several points along the axis, each carrying its own measured deviation, and the console interpolates between them. The correction is no longer one rate; it shifts as the carriage moves, which is exactly what a guideway problem needs. A bed that is flat and true for the first 200 mm and then dips slightly over the next 150 mm produces an error curve that rolls away from a straight line. One factor cannot follow that shape, but a segmented table can. The same logic covers a tightened or worn zone near one end of travel, or a section where the scale mounting pulls the reading off. The cost is measurement work: more points mean more setup, and every point has to come from the actual machine rather than from a general rule. The Easson ES-14 is one console that exposes both routes, listing linear error compensation and nonlinear error compensation as separate functions, so a technician can choose the method that matches what the measurement showed. Which values go into either screen still comes from that specific machine's numbers. There are no universal settings to copy from a manual, because two lathes of the same model rarely drift in the same way.
Why Compensation Does Not Replace Mechanical Repair
Compensation changes the number, not the metal. Every correction table rests on one assumption: the error repeats. Loose gibs, a worn nut, a scored guideway or chips packed under the saddle produce readings that change from pass to pass — one traverse ends 0.03 mm out, the next ends 0.01 mm out. No stored table predicts a number that is not repeatable. Wear also steals rigidity, so the cutter can push the table around mid-cut while the display sits at a beautifully corrected position. The finished size still wanders, the surface finish still suffers, and the operator still has to chase the cut. The practical order is mechanical first, electronic second. Check the machine's condition, repair or replace what is worn, then measure the residual error and compensate that. A readout console is a display and measurement device, not a closed-loop CNC control: it reports a corrected position and does nothing to drive the screw or push the axis back to target. Its honest job is removing the repeatable leftover once the mechanics are sound, and on a machine in good order that leftover is often small enough that compensation simply sharpens an already accurate axis.
Conclusion
Uniform error and local error are different animals, and the two compensation methods exist because one correction cannot describe both. When deviation grows in proportion to travel, a single factor does the job cleanly. When deviation appears in one zone, segmented points are the only way to follow the curve. The starting point is always the same: measure the axis, look at the shape of the result, and pick the method that matches it. Compensation then handles the repeatable remainder, while worn screws, slack guideways and tired bearings still belong to the maintenance schedule.
FAQ
Q: What kind of machine error does linear compensation correct?
A: It corrects error that grows in proportion to travel, the classic uniform pitch error of a leadscrew. If an axis reads 0.02 mm short after 200 mm and 0.04 mm short after 400 mm, the deviation follows a straight line from zero, and one correction factor per axis scales the display to match. It is the right tool when the midpoint deviation sits close to half the full-travel deviation.
Q: When does nonlinear compensation make more sense than a single correction factor?
A: When the error changes character along the travel instead of climbing steadily. A dip in the middle of a worn bed, a tight or loose stretch near one end of the guideway, or a section where the scale mounting pulls the reading off all produce a curved error pattern. Segmented points with interpolation between them follow that curve, while a single factor would over-correct one zone and under-correct another.
Q: Why can error compensation not fix a worn leadscrew or loose guideway?
A: Because compensation works on the displayed number and assumes the error repeats. A worn screw or slack guideway produces deviation that varies from pass to pass and steals rigidity, so the cutter deflects the table mid-cut regardless of what the display says. Repair or replacement restores the machine; compensation then cleans up the small, repeatable residual that remains.
Sources / References
CNC Cookbook - The Ultimate CNC Machining Resource
Related Examples
Easson ES14 White Position 3 Axis Digital Readouts Systems For Lathe
No comments:
Post a Comment