Position Measurement on Machine Tools:

by Linear Encoder or Ballscrew and Rotary Encoder?

The discussion as to whether linear encoders or rotary encoders and recircu- lating ballscrews
represent the best solution for measurement on NC machine tools is experiencing a renaissance at
present. This is to some degree certainly an aftereffect of the recession in the machine tool industry,
but it also springs from current changes in drive technology.
After the general recession in the machine tool industry, in several quarters the question arose of
whether the industry's strategy of the last few years has been correct. Could it be that too little
attention was paid to cost and too much to push- ing the limits of the technologically possible? Is it not
preferable to remove those features not required for a basic version of a machine tool, to permit a
competitive starting price for a low-end model? Under the pressure of such currents of opin- ion one
can easily come upon the idea of saving money on the measuring systems by leaving linear encoders
off the lowest-priced models and offering them as an accuracy-enhancing option.
Trend toward digitally driven axes
This line of thought is reinforced by the trend in the drive industry in the direction
of digital axes. A large portion of the new servo motors now feature rotary en-
coders, which in principle can also be used in combination with the ballscrew for
position control. With this method there is no longer a question of choosing between
a ballscrew/rotary encoder system and a linear encoder solution for position control.
With such a drive configuration the decision is rather whether to add a linear en-
coder or simply to use an already existing motor encoder working in combination
with the ballscrew.

In the heat of such discussions one should not forget to consider the problems
known to be involved with position measurement using a rotary encoder/ballscrew
system. They can quickly increase the cost of an "economical" machine if the owner
finds that the accuracy does not suffice in certain applications, or that thermal ex-
pansion problems are causing the machine to generate scrap every morning. Retro-
fitting an installed machine with optional linear encoders is usually much more ex-
pensive than having them delivered with the machine from the start.

Feed mechanism of a milling machining center.

Mechanical feed-drive system

An exact error analysis of position measurement via rotary encoder and ballscrew
begins with a consideration of prevalent mechanical feed-drive systems.

Although machine tool designs vary immensely, the mechanical configuration of

their feed drive is largely standardized (Fig. 1). In almost all cases, the recircu-
lating ballscrew has established itself as the solution for converting the rotary motion
of the servo motor into linear slide motion. The ballscrew is normally fixed in axial
direction at only one end with a preloaded angular-contact ball bearing, which takes
up the axial forces of the slide. The servomotor and ballscrew drive are usually
directly coupled. Toothed-belt drives are also widely used to achieve a compact de-
sign and better adapt the speed.
Fig. 1: Typical drive system of a numerically controlled machine tool with linear scale on the slide and
rotary encoder on the motor.

Fig. 2: Fundamental difference between position control with linear encoder and with rotary
encoder/spindle. The linear encoder includes the feed drive mechanism in the control loop.

A position control loop via rotary encoder and ballscrew includes only the servo-
motor (Fig. 2). In other words, there is no actual position control of the slide, be-
cause only the position of the servomotor rotor is being controlled. To be able to
extrapolate the slide position, the mechanical system between the servomotor and
the slide must have a known and, above all, reproducible mechanical transfer be-
havior.

A position control loop with a linear encoder, on the other hand, includes the entire
mechanical feed-drive system. The linear encoder on the slide detects mechanical
transmission errors and these are compensated by the machine control unit.

Differing terminology
Differing terms are used to distinguish between these two methods of position con-
trol. German-speaking and some English-speaking communities generally refer to
them somewhat inaccurately as "direct and indirect measurement." Here the Japa-
nese concepts of "semi-closed-loop and closed-loop control" seem appropriate,
since they more aptly describe the actual problem.

Kinematic error
Kinematic error in position measurement with rotary encoder and ballscrew results
primarily from ballscrew pitch error. This error directly influences the result of mea-
surement because the pitch of the ballscrew is being used as a standard for linear
measurement.

Reversal error
Reversal error occurs during positioning from differing directions. The causes are
play and elasticity in connection with frictional forces. But also the so-called pitch
loss [1] resulting from a shift of the balls during the positioning of ballscrew drives
with two-point preloading can lead to reversal error in the magnitude of 1 to 10 µm.

Error compensation
Most controls are capable of compensating pitch error and reversal error. However,
to determine the compensation values it is necessary to make elaborate measure-
ments with comparative measuring devices such as interferometers and grid en-
coders. In addition, the reversal error is often unstable over long periods of time
and must be regularly recalibrated (Fig. 3). The causes of this instability include
run-in processes of the ballscrews and changes of the frictional forces in the
guideways. Toothed belt drives can also cause significant positioning errors in the
course of time.

Fig. 3: Circular tests of a machining center without linear encoders in new condition and after one year.
The reversal error has significantly increased in the X axis.

Deformation of drive mechanisms

Forces leading to the deformation of feed drive mechanisms are essentially inertia
forces resulting from acceleration of the slide, cutting process forces, and friction
in the guideways. They cause a shift in the actual axis slide position relative to the
position measured with the ballscrew and rotary encoder.

The mean axial rigidity of a feed drive mechanism as shown in Figure 1 lies in the
range of 100 to 200 N/µm (with a distance between ball nut and fixed bearing
of 0.5 m and a ballscrew diameter of 40 mm).

Forces of acceleration
A typical slide mass of 500 kg and a moderate acceleration of 2 m/s 2 result in de-
formations of 5 to 10 µm that cannot be recognized by the rotary encoder/ballscrew
system. The present industry trend toward accelerations in significantly higher
ranges will result in increasingly great deformation values.
Cutting forces
The cutting forces can quite possibly lie in the kN range, but their effect is distri-
buted not only in the feed drive system, but also over the entire structure of the
machine between the workpiece and the tool. The deformation of the feed drive
system therefore normally has only a small share in the total deformation of the
machine. A linear encoder can recognize and correct only this small portion of the
total deformation. Critical component dimensions, however, are normally finished
at low feed rates, so that the deformation of the feed drive system is negligible.

Forces of friction
The forces of friction in the guideways lie between 1% and 2% of weight for roller
guideways and 3% to 12% of weight for sliding guideways [2]. A weight exerting
5000 N therefore results in feed drive deformation of only 0.25 to 6 µm.

Fig. 4: Circular tests of a machining center that has been retrofitted with linear encoders. With position
control via rotary encoder and spindle, the circle deviates significantly from the ideal path at high velocity.
With linear encoders the contour accuracy is significantly better.

Circular test for inspecting machine tools

A typical example for errors dependent on acceleration and velocity can be re-
corded in a circular interpolation test on a vertical machining center (Fig. 4). Where
position control is by rotary encoder and ballscrew, the circles traversed at higher
velocities deviate significantly from the ideal path. The same machining center
shows significantly better contour accuracy when equipped with linear encoders.

Positioning error due to ballscrew expansion

Positioning error resulting from thermal expansion of the ball screw presents the
greatest problem for position measurement via rotary encoder and ballscrew. This is
because the ballscrew drive must serve a double function: On the one hand it must
be as rigid as possible to convert the rotary motion of the servo motor to linear feed
motion. On the other hand it must serve as a precision measuring standard. The
two-fold function therefore forces a compromise because both the rigidity and the
thermal expansion depend on the preloading of the ball nut and the fixed bearing.
Both the axial rigidity and the moment of friction are roughly proportional to the pre-
loading.
Fig. 5: Causes of the high moment of friction in preloaded ballscrews.

Friction in the ball nut

The largest portion of the friction in a feed drive system is generated in the ball nut.
This is because of the complex kinematics of a recirculating ball nut. Although at
first glance the balls may seem only to be rolling, they are in fact subjected to a
great deal of friction. Besides the microslip resulting from relative motion in the com-
pressed contact areas, the greatest effect is from the macroslip due to kinematic
exigencies. The balls are not completely held in the races and wobble much like
tennis balls rolling down a gutter. The result is a continual pressing and pushing with
occasional slipping of the balls. The friction among the balls is aggravated by high
surface pressure due to the absence of a retaining device to separate them. As in
every angular-contact ball bearing a spinning friction results from a contact dia-
meter that is not orthogonal to the axis of ball rotation. Each ball therefore rotates
about its contact diameter. Recent studies have also shown that the balls can move
in the thread only because of an additional slip component brought on by the thread
pitch [3].

The recirculation system is a special problem zone for ball screws. With every en-
trance into the recirculation channel, just as with every exit, the movement of the ball
changes entirely. The rotational energy of the balls, which in rapid traverse typically
rotate with 8000 rpm, must be respectively started and stopped. In contrast to the
preloaded thread zone, in the recirculation zone the balls are not under stress. For
reasons of energy the balls tend to collect in the recirculation channel. Without ela-
borate measures to reintroduce the balls into the thread at the end of the channel it
tends to congest, causing the familiar jamming of the ball screw drive.

The moment of friction of a ground precision recirculating ballscrew with 40 mm dia-

meter and 10 mm pitch was measured by Golz [4] for various preload forces and
rotational speeds (Fig. 6). The Stribeck characteristic of frictional moment is clearly
recognizable. It confirms the high share of solid-body friction and mixed friction in
ballscrew drives at low speeds. Viscous friction dominates at high speeds. It is in-
teresting to note that for this typical ballscrew the normal machining feed rates lie far
below the speeds at which the moment of friction is at its minimum. The rapid tra-
verse feed rates, however, lie far above it. The feed rates at which this ballscrew is
at optimum efficiency therefore seldom occur. The moment of friction is only slightly
dependent on the axis load of the ball nut [4].

Fig. 6: Moment of friction of a two-point preloaded ballscrew (4). The Stribeck characteristic is
plainly visible.

Frictional heat generated in the ball nut

With a typical preload of 3 kN and allowing for the missing wiper, this results in a
no-load or frictional moment of 0.5 to 1 Nm. This means that in rapid traverse at a
ballscrew speed of 2000 rpm approximately 100 to 200 W of frictional heat is ge-
nerated in the ball nut.
Fig. 7: Simple experimental setup showing thermal growth in a recirculating ballscrew. With a touch
trigger probe, fixed positions within the traverse range are probed and the respective actual position
values of the linear encoder and of the rotary encoder/spindle system are recorded.

Drift of position values measured with rotary encoder and ballscrew

A simple experiment (Fig. 7) shows the influence of frictional heat on the positioning
behavior of a machining center. On a large machining center equipped with linear
encoders and rotary encoders, five positions in the X axis were repeatedly measur-
ed with a touch probe, and the respective actual values of the rotary encoder and
linear encoder at the touch points were recorded. After all five positions were probed,
the axis was moved over the entire traverse range in two cycles at rapid traverse.
In total, 120 of these cycles were performed within 90 minutes.

Fig. 8: Results of the experiment shown in Figure 7. With position control using rotary encoder/spindle
there is a significant drift in the probe position values as a result of thermal expansion in the ballscrew.

Warm-up phase
The drift of the individual probe positions relative to their starting value plainly shows
the thermal growth of the ballscrew (Fig. 8). When the positions are measured with
rotary encoder and spindle, the position farthest from the fixed bearing (x = -53mm)
drifts by approx. 250 µm. It is interesting to note that the drift increases very quickly
immediately after switch-on. Any change in the mean feed rate during a machining
operation therefore immediately affects the positioning accuracy. Similar results
were published by Schmitt [5].
Fig. 9: Typical drift of the actual position values during a work break.
The actual position values - measured by rotary encoder and spindle - drift
by approx. 30 µm within 30 minutes.

Cool-down phase
Figure 9 shows the opposite effect.
The same machining center was stopped during series production of small steel
parts, and one work position was repeatedly probed. The drift in the position
measured with rotary encoder and ballscrew of 30 µm over 30 minutes is again
clearly visible.

Measurement of positioning accuracy according to ISO 230

The ISO/DIS 230-3, at present still a draft standard, describes measuring methods
for thermal shifts in machine tools. To measure the drift of linear axes, it proposes
traversing repeatedly to two points in the end positions of the traverse range and
measuring the position drift from start value. This procedure is in essence a widely
practiced positioning test as described, for example, in ISO 230-2, but with few
measuring points and many repetitions. Figure 10 shows the result of such a uni-
directional measurement relative to their initial value on a vertical machining center
retrofitted with linear encoders. The 1-m long X axis was moved to three positions
at 10 m/min a total of 100 times. At first, position control was through the rotary en-
coder/ballscrew system. In a second experiment under otherwise identical conditions,
position control was through the linear encoders. The comparator measuring device
was a VM 101 manufactured by HEIDENHAIN. In spite of the moderate feed rate
of 10 m/min (rapid traverse 24 m/min, the position farthest from the ballscrew fixed
bearing drifted by more than 110 µm within 40 minutes. A significant growth of the
ballscrew becomes apparent already after the first few cycles.
Fig. 10: Drift of three positions during measurement in accordance with ISO 230 with a large number
of repetitions. Position measurement via rotary encoder and ballscrew shows a significant drift of
positions due to the thermal growth of the ballscrew.

No drift in position values measured with linear encoders

The measured positioning accuracy therefore depends directly on the number of
repetitions, particularly after the first few repetitions. The measurements taken by
the retrofitted linear encoder show no drift.

Fig. 11: Experimental setup for batch production with multiple workpieces. To illustrate drift resulting
from thermal expansion of the ballscrew, the workpieces were not exchanged after machining. Instead,
the part program was run repeatedly at successively increasing depth.

Batch production
The following experiment shows the influence of thermal growth in batch production
with fixed clamping positions. Eight 70 mm x 70 mm workpieces were fixed on a
vertical machining center (Fig. 11). Four pockets and two radii were machined using
4 tools with an infeed of 1 mm in the Z axis (Fig. 12). After the 6-minute machining
operation the 8 parts were not exchanged. Rather, the infeed in Z was increased
by 1 mm and the operation repeated. As a result of the thermal expansion of the
ballscrew, all workpieces show a step pattern on the left side. This pattern is parti-
cularly pronounced on the workpiece farthest to the left. The right sides of the work-
pieces are smooth because with each shift in the positive X direction the previous
step was also removed. In principle the same effect could be observed in the Y
direction as in the X direction, but because of the lesser amount of movement in the
Y axis the step pattern is significantly less pronounced. In the X direction the com-
parative measurement of the step pattern shows a drift of approx. 90 µm with a time
constant of thermal expansion slightly less than an hour (Fig.13).

If additional work is to be done on previously machined workpieces with critical di-

mensions, the machine datum must be continually inspected and corrected. The ma-
chine achieves thermal equilibrium after one hour, but after an interruption in ma-
chining it begins to drift in the reverse direction. If the part program and with it the
mean feed rate are changed, it again takes approx. 1 hour for the ballscrew to re-
gain thermal equilibrium.

Fig. 12: Part program used in the experiment shown in Figure 11. Four pockets and two radii were
machined using 4 tools at an infeed of 1 mm.

The same experiment was conducted with linear encoders. The results of machining
showed virtually no drift.

Various measures are presently being discussed and realized to counter positioning
error resulting from ballscrew expansion.

Some manufacturers offer hollow ballscrews that conduct coolant to prevent thermal
expansion.
Circulation of the coolant through rotating ballscrews requires rotary leadthroughs
near the ballscrew bearings. Besides the sealing problems obviously involved, this
method presupposes the capability of precisely controlling the temperature of the
coolant, which is usually not the case. Also, it reduces the rigidity of the ballscrew
in the direction of traverse. Presumably, the cost of this method is greater than the
cost of linear encoders.

Many studies are presently being conducted on the compensation of thermal defor-
mation with the aid of analytical models, neuronal networks and empirical equations.
In most cases, however, these studies focus on thermal expansion caused by main
spindles.

Fig. 13: Result of the experiment in Figure 11. The left pocket of the left workpiece plainly shows a
step pattern resulting from thermal expansion of the ballscrew.

To compensate the expansion of ballscrews, the temperature must be known as a

function of position, because in some part programs thermal expansion can occur
locally (Fig. 14). Direct temperature measurement of the rotating ballscrew, how-
ever, is very costly. The possibilities of drawing conclusions on ballscrew expan-
sion using temperature data from the ball nut or bearing are limited, partly be-
cause the input parameters of such calculation models change over time.

The application of fixed bearings at both ends of the ballscrew does significantly
increase the axial rigidity of the feed drive, however it can hardly prevent thermal
growth of the ballscrew. For a ballscrew diameter of 40 mm, the bearings would
have to apply approx. 2.6 kN/K of force to suppress the thermal expansion. For the
typical temperature increase of more than 10 K this would require a bearing
strength of more than 26 kN to prevent deformation.
Fig. 14: Local heating of a recirculating ballscrew in the traverse range of the ballnut after six hours of
reversing traverse at 24 m/min between two points 150 mm apart. For this thermographic snapshot, the
machine table was moved aside at the end of the traverse program. The illustration shows the higher
temperatures of the belt drive, locating bearing, and ballscrew.

Conclusion
The primary problem involved with position measurement using rotary encoder and
ballscrew is the thermal expansion of the ballscrew. With typical time constants
of 1 to 2 hours, thermal expansion causes positioning error in the magnitude
of 0.1 mm, depending on the nature of the part program. After every new part pro-
gram the ballscrew requires approx. 1 hour to attain a thermally stable condition.
This also applies for interruptions in machining. A rule of thumb for thermal ex-
pansion is that, over the entire length of a cold ballscrew 1 meter in length, the ball-
screw grows by approx. 0.5 to 1 µm after every double stroke. This expansion
accumulates within the time constant.

As requirements for machine tool accuracy and velocity increase, the role of li-
near encoders for position measurement grows increasingly important. This should
be taken intoconsideration when deciding on the proper feedback system design.

Literature:
1. Schröder Wilhelm, Feinpositionierung mit Kugelgewindetrieben, Fortschritts-
bericht VDI Reihe 1 Nr. 277, Düsseldorf: VDI Verlag 1997
2. VDW-Bericht 0153, "Untersuchung von Wälzführungen zur Verbesserung des
statischen und dynamischen Verhaltens von Werkzeugmaschinen"
3. Weule Hartmut, Rosum Jens, Optimization of the friction behaviour of ball
screw drives through WC/C coated roller bodies, Production Engineering
Vol. 1/1 (1993)
4. Golz, Hans Ulrich, Analyse,Modellbildung und Optimierung des Betriebsver-
haltens von Kugelgewindetrieben, Dissertation Uni Karlsruhe, 1990

5. Schmitt Thomas, Modell der Wärmeübertragungsvorgänge in der mechani-

schen Struktur von CNC-gesteuerten Vorschubsystemen,
Verlag Shaker, 1996
 About the author
Dr.-Ing. Jan Braasch studied mechanical engineering at the
University of Kassel, where he also completed his graduate
work. Since 1994 he has been with the DR. JOHANNES
HEIDENHAIN company in Traunreut, Germany, where he
heads a development department for linear encoders.

For more detailed information on these applications, please contact HEIDENHAIN directly.