Content uploaded by Brian Cahill
Author content
All content in this area was uploaded by Brian Cahill on Jul 27, 2021
Content may be subject to copyright.
LED-based ®bre-optic sensor for measurement of surface roughness
B. Cahill
a,b,*
, M.A. El Baradie
a
a
School of Mechanical and Manufacturing Engineering, Dublin City University, Dublin, Ireland
b
Hewlett-Packard GmbH, Bo
Èblingen, Germany
Abstract
This project deals with the design and development of an optoelectronic sensor system and its possible use in online applications.
Average surface roughness is estimated between 0.025 and 0.8 mm through a light scattering technique. Specular re¯ectivity was measured
at an incident angle of 608. A carrier-frequency system using an electronically modulated LED light source was implemented to improve
the noise rejection of the system. #2001 Elsevier Science B.V. All rights reserved.
Keywords: LED; Fibre-optic sensor; Surface roughness
1. Introduction
With the increased importance of quality control and
micromachining of precision parts in the engineering indus-
try, there is a need for on-line measurement of surface
roughness. Measuring surface roughness between 0.01
and 1 mm is of particular importance for online inspection
of optical surfaces, rolling of aluminium ®lm, precision-
machined parts and substrates for hard disks [1±6]. The
traditional method for measuring surface roughness in this
range is the contact stylus method, but optical methods have
many advantages over this method including the high speed
needed in quality control [7±9]. Optical methods due to their
non-contact nature can perform measurements of surface
roughness very quickly, often while the sample is in motion.
Surface roughness can be measured through the effect of
light scattering from rough surfaces [10,11]. In the transition
from a smooth surface, which transmits light specularly, to a
rough surface, a higher proportion of incident light is
scattered diffusely. This transition can be related to surface
roughness. Surface roughness can be described as a point
quantity using light scattering methods, averaging roughness
over the incident spot of the beam. The ®bre-optic surface
roughness sensor presented here investigates the phenom-
enon of light scattering from surfaces of different surface
roughness.
The sensor is designed to replace the use of stylus
instruments in the control of surface roughness of rolled
sheet, semiconductor wafers, hard disk substrates and pre-
cision-machined parts. In the manufacture of sheet metal,
surface smoothness is an important quantity. Rollers are
used to reduce an input sheet or billet, often at high
temperatures and pressure. In performing this work, the
roller becomes worn out and needs to be reground. A
®bre-optic surface roughness sensor can perform higher-
speed, online, non-contact measurements of surface rough-
ness not only of the end product but also of the roller itself.
Waste can be reduced in comparison to the use of of¯ine
stylus inspection. This ®bre-optic sensor needs give only an
indication of the surface roughness of the ground sheet, not
necessarily a very accurate measurement.
2. Light scattering
A surface can be considered smooth if it re¯ects spec-
ularly. A surface may be smooth or rough for different
wavelengths of electromagnetic radiation. Re¯ection from
a surface depends on the wavelength and incident angle of
the incident beam and the properties of the surface.
The electrical properties of the surface can be considered
as material constants. It is possible to infer some of these
surface features from the light scattering characteristics of
the surface. Eq. (1) [11] was developed to describe the
scattering of electromagnetic radiation from a random rough
surface:
Is
Io
/exp 4pRqcos y
l
2
"#
for Is
Io
>0:6 (1)
Journal of Materials Processing Technology 119 (2001) 299±306
*
Corresponding author. Present address: School of Mechanical and
Manufacturing Engineering, Dublin City University, Dublin, Ireland.
0924-0136/01/$ ± see front matter #2001 Elsevier Science B.V. All rights reserved.
PII: S 0924-0136(01)00954-2
where I
s
is the specular reflectance, I
o
the total reflectance,
ythe incident angle, R
q
the RMS surface roughness and l
the optical wavelength. This equation assumes a Gaussian
distribution of roughness. To estimate roughness from
this equation, the scattering ratio, I
s
/I
o
, must be above
0.6. This equation describes how in the transition from a
mirror-like surface to a rougher surface the fraction of
light intensity is transmitted specularly and how incident
angle, surface roughness and optical wavelength affect
reflectivity.
3. Design
The basic block diagram of the system is shown below in
Fig. 1. Fixturing attached to translation stage positions the
bare optical ®bres above a sample. The ®bres are positioned
at equal but opposite angles above a sample surface. One set
of ®bres emits light while the other receives light. Light
detection is followed by a stage of preampli®cation before
data acquisition by the analog to digital converter (ADC).
The personal computer performs data retrieval, analysis and
presentation using LabView software.
Fig. 2 shows the front view of the XYZ translation stage
and ®xturing for ®bres shown in Fig. 1. Fibres can be held at
an incident angles of 608to normal incidence. Bare ®bres
protruded from the ends of the ®bre holders Ð the absolute
position of the ®bres was uncertain.
4. Data acquisition and data analysis algorithms
This section details the procedure and algorithms that
were implemented for data acquisition and data analysis.
Data analysis was performed using Labview software. The
user interface displays the outputs of the data analysis and
provides control objects, such as switches and variables,
which interfaces with the underlying program. In this way,
the user can interact with the virtual instrument while data is
being acquired.
Fig. 1. Block diagram of fibre optic sensor system.
Fig. 2. System for positioning input and receiving fibres.
300 B. Cahill, M.A. El Baradie / Journal of Materials Processing Technology 119 (2001) 299±306
Use of a carrier frequency system to reject ambient light
improves the performance of a photoelectric sensor [12].
This system uses digital signal processing (DSP) to achieve
noise rejection. Windowing [13] is used in DSP applications
to reduce the effect of sampling conditions on the signal; the
®nite sampling record results in a truncated waveform that
has different spectral characteristics from the original con-
tinuous time signal. Windowing allows shorter sampling
lengths to be used in DSP.
The sensed voltage is primarily affected by low frequency
noise; for this reason, we have chosen a modulation fre-
quency of 930 Hz, well above mains noise and other low
frequency noise. Fig. 3 shows an algorithm that was used to
record the displacement characteristics. The virtual instru-
ment that implemented the program shown in Fig. 3 is shown
in Fig. 4.
The displacement curve shows a characteristic for each of
the three ®bres that sensed the re¯ected signal. The switch
titled ``One Reading'' triggered the addition of an increment
to the graph. The ``STOP'' button caused the program to stop
running once the data ¯ow reaches the switch. The signal
processing inputs could be controlled from the virtual
instrument panel, but were con®gured as shown for all
readings. ``Total sampling time'' is an indicator which shows
the sampling time that has been de®ned in the controlling
program. ``Sampling rate'' is a control that sets the sampling
rate of the device. The low cut-off frequency can be con-
trolled from the front panel. For a bandpass ®lter, the high
cut-off frequency will be 500 Hz greater than the low cut-off
frequency.
Three displacement characteristic curves can be seen in
Fig. 4. These curves correspond to three input signals from
three PIN photodiodes connected to three input ®bres held in
the ®bre holder. The orientation of these ®bres is discussed
elsewhere. These curves show the displacement of the ®bres
relative to the starting position of the stage rather than to the
absolute displacement from the surface. The virtual instru-
ment was con®gured as displayed in Fig. 4 for all measure-
ments with the surface roughness sensor.
5. Results
The light source consisted of a HFE4050-014/BBA high
power LED (850 nm) driven by the square wave circuit at a
frequency of 930 Hz. The signal detection circuit consisted
of unbiased transimpedance ampli®ers using Honeywell
PIN photodiodes, HFE3002-002/BBA. These silicon PIN
photodiodes are sensitive to optical radiation from 400 to
1000 nm, with optimum responsivity close to 850 nm.
Three ®bres collected light re¯ected from the surface. If
the position, i.e. displacement, and the size of the peak of the
displacement curve is consistent throughout a set of read-
ings, it suggests that the orientation of surfaces in question is
similar. Any differences in these criteria may account for
inconsistent results.
Fig. 3. Block diagram of program to record displacement characteristic.
B. Cahill, M.A. El Baradie / Journal of Materials Processing Technology 119 (2001) 299±306 301
The experimental samples were surface roughness com-
parator scales, manufactured by Rubert with the following
nominal R
a
values: 0.025, 0.05, 0.1, 0.2, 0.4, and 0.8. The
accuracy of these specimens is quoted as 12%/17% of
the nominal value. The surface roughness characteristics
shown in Table 1 were measured using the Mitutoyo Surftest
402 contact stylus; after all the experiments were con-
cluded, R
z
measurements were made over a pro®le length
of 10 mm.
Eq. (1) is used to plot a theoretical curve in ®gure using
608and 850 nm light source as the incident angle and optical
wavelength, respectively.
Fig. 4. Front panel of virtual instrument described by block diagram in Fig. 3.
Table 1
Surface roughness characteristics of Rubert comparator scales
Nominal R
a
value (mm)
Measured roughness characteristics
R
a
(mm) R
q
(mm) R
z
(mm)
0.025 0.04 0.05 0.2
0.05 0.05 0.06 0.4
0.1 0.10 0.13 0.9
0.2 0.21 0.27 1.5
0.4 0.47 0.61 2.6
0.8 0.85 1.1 4.6
302 B. Cahill, M.A. El Baradie / Journal of Materials Processing Technology 119 (2001) 299±306
The vertical displacement characteristics of each surface
were measured using ®bres oriented at incident angles of 608
for each of the samples shown in Table 1. The following six
graphs show the displacement characteristics for each sam-
ple at an incident angle of 608. This set of measurements
suggests that the three ®bres are oriented with the ®bre
supplying PIN photodiode 2 in the centre and those supply-
ing PIN photodiodes 1 and 3 at either side.
Figs. 5±7 correlate each peak voltage for the vertical
displacement characteristics presented in Figs. 8±13 against
RMS surface roughness, average surface roughness, and 10-
point height. For this set-up there appears to be two distinct
lines on these graphs: ®rstly at roughness up to 0.1 mm R
a
and secondly above that value. In all the ®gures, PIN 3, the
photodiode connected to the ®bre receiving the highest
intensity, gives the most consistently successful readings.
Eq. (1) was used to generate the theroretical curve shown in
Fig. 6.
Fig. 5. Peak voltage values versus average roughness, R
a
, for incident
angle of 608.
Fig. 6. Peak voltage values versus RMS roughness, R
q
, for incident angle
of 608.
Fig. 7. Peak voltage values versus 10-point height, R
z
, for incident angle
of 608.
Fig. 8. Vertical trace of 0.025 mmR
a
surface roughness sample at incident angle of 608.
B. Cahill, M.A. El Baradie / Journal of Materials Processing Technology 119 (2001) 299±306 303
Fig. 9. Vertical trace of 0.05 mmR
a
surface roughness sample at incident angle of 608.
Fig. 10. Vertical trace of 0.1 mmR
a
surface roughness sample at incident angle of 608.
Fig. 11. Vertical trace of 0.2 mmR
a
surface roughness sample at incident angle of 608.
304 B. Cahill, M.A. El Baradie / Journal of Materials Processing Technology 119 (2001) 299±306
6. Conclusions
It can be seen that only the ®bre that receives the highest
signal (PIN 3) is particularly useful for measuring surface
roughness. This shows the disadvantage of using bare ®bres
to collect light. Nevertheless, the signal measured by PIN 3
shows a good correlation with all three chosen surface
roughness parameters.
It is clear that each of these curves can be divided into two
sections, corresponding to above and below 0.1 mmR
a
. The
section below 0.1 mmR
a
is more sensitive, while the section
above is less sensitive, but displays relatively good linearity
for all three ®gures. This corresponds roughly with the
condition for Eq. (1) that the scattering ratio, I
s
/I
o
, must
be above 0.6. The graphs show that light scattering may be
used to indicate the surface roughness of rougher surfaces.
In general, the stand-off distance of the endface of the
optical ®bres from the surface of the sample plates was
between 2 and 3 mm. For on-line use using present con-
struction of the experimental rig, this is unsatisfactory as the
®bres are unprotected. If protection of the ®bre was satis-
factorily incorporated as part of the mechanical design of
the sensor, this stand-off distance may be adequate.
To increase the stand-off distance, the optical design of
the system needs to change; an emitting ®bre with lower
numerical aperture could be chosen. The defect sensor
used 62.5/125 mm ®bre with numerical aperture of 0.275,
y16, while standard 50/125 mm ®bre has numerical
aperture of 0.2, y11:5, which would allow a slightly
higher stand-off distance. The incident angle also affects
stand-off distance, higher incident angles having lower
stand-off distances for the same lateral separation of the
®bres. The choice of incident angle also affects the ability
of the sensor to sense blind holes. A higher intensity light
source could be chosen. This would introduce further
consideration of laser safety, in particular for on-line
Fig. 12. Vertical trace of 0.4 mmR
a
surface roughness sample at incident angle of 608.
Fig. 13. Vertical trace of 0.8 mmR
a
surface roughness sample at incident angle of 608.
B. Cahill, M.A. El Baradie / Journal of Materials Processing Technology 119 (2001) 299±306 305
use.Theoptimumsolutionwouldbetousealownumer-
ical aperture ®bre as the input ®bre and a ®bre bundle, as
used by Domanski et al. [14], to collect the re¯ected
signal.
Eq. (1) uses RMS surface roughness; results are presented
using average surface roughness, RMS surface roughness
and 10-point height and it is dif®cult to quantify which
property correlates most closely with surface re¯ectivity.
The curves for each different roughness property differ very
little. As the sensor does not measure any roughness prop-
erty absolutely, it may be most properly termed a surface
®nish sensor.
References
[1] P. Cielo, Optical Techniques for Industrial Inspection, Academic
Press, New York, 1990.
[2] M.M. Moslehi, Sensor for semiconductor device manufacturing
process control, US Patent No. 5,293,216 (1994).
[3] R. Brodmann, O. Gerstorfer, G. Thurn, Optical roughness for fine
machined surfaces, Optic. Eng. 24 (3) (1985) 408±413.
[4] K. Mitsui, In-process sensors for surface roughness and their
applications, Precis. Eng. 8 (4) (1986) 212±220.
[5] M. Miller, R.S. Lakes, S. Conner, Optical testing of hard disks, Opt.
Laser Technol. 28 (3) (1996) 151±156.
[6] R. Brodmann, G. Thurn, Roughness measurement of ground, turned
and shot-peened surfaces by the light scattering method, Wear 109
(1986) 1±13.
[7] M. Stedman, K. Lindsey, Limits of surface measurement by stylus
instruments, Proc. SPIE 1009 (1988) 56±61.
[8] R. Brodmann, M. Allga
Èuer, Comparison of light scattering from
rough surfaces with optical and mechanical profilometry, Proc. SPIE
1009 (1988) 111±118.
[9] K.J. Stout, Optical assessment of surface roughness, Precis. Eng. 6
(1) (1984) 35±39.
[10] J.M. Bennett, L. Mattson, Introduction to Surface Roughness and
Scattering, Optical Society of America, 1989.
[11] P. Beckmann, A. Spizzichino, The Scattering of Electromagnetic
Waves from Rough Surfaces, Pergamon Press, Oxford, 1963.
[12] Handbook of Photoelectric Sensing, Banner Engineering Corp.,
Minneapolis, MN, 1994.
[13] C.D. McGillem, G.R. Cooper, Continuous and Discrete Signal and
System Analysis, Saunders, Philadelphia, PA, 1991.
[14] A.W. Domanski, T.R. Wolinski, T.J. Rzysko, Optimal distribution of
optical fibers in surface roughness sensor, Proc. SPIE 1009 (1988)
134±139.
306 B. Cahill, M.A. El Baradie / Journal of Materials Processing Technology 119 (2001) 299±306
