ILLUMINATION DESIGN USING ZEMAX OPTICSTUDIO Edmund Optics®

W hitePa per
Written by Eduardo Gonzalez, Design Engineer at Edmund Optics®, and Sanjay Gangadhara, Chief Technology Officer at Zemax

DESIGN AND EVALUATION OF A TELECENTRIC BACKLIGHT

Telecentric backlight illuminators are ideal for machine vision and spatial distribution of light, from the illuminator with that of the
applications that require high contrast silhouettes, such as precise telecentric lens. The mismatch manifested itself in the image plane as
measurements of small details like the threads on a screw. the non-uniformity seen in Figure 3.
Collimated, highly-concentrated light emits from the telecentric
illuminator, minimizing diffuse reflections at the edge of the part
under inspection. Standard backlights often produce unwanted
diffuse reflections that reduce edge contrast, as shown in Figures
1 and 2. Additional benefits of telecentric backlight illumination
include reduced camera exposure times because of increased light
intensity, faster systems, and larger possible distance between the
object and illumination source.

Figure 1: Collimated light rays from a telecentric illuminator (left) vs. diffuse
reflections from a standard backlight (right)

Figure 3: Test result for the original design, showing a dark ring-shaped region of
non-uniform irradiance

The telecentric illuminator and imaging lens should be modeled

simultaneously to resolve this issue. This would allow for an analysis
of the etendue of both subsystems to ensure there are no uniformity
issues at the image plane. Checking a newly-designed telecentric
illuminator in this way will result in an even irradiance distribution.

The Process
Figure 2: Silhouette from a telecentric illuminator (left) vs. a standard First, a value had to be assigned to the pass/fail metric. It was
backlight (right) decided that the irradiance value anywhere on the sensor could not
decrease from the peak value by more than 10%. The design was set
However, if a telecentric backlight is used with a telecentric imaging up in the sequential mode of OpticStudio, and then incremental steps

® COPYRIGHT 2020 EDMUND OPTICS, INC. ALL RIGHTS RESERVED

lens and they are not properly paired, the irradiance may not be were used to get the design to a place where it could be analyzed
uniform. In this instance, there was a decrease from peak irradiance in non-sequential mode. The design was optimized using RMS spot
of up to 18% across the sensor. Edmund Optics® utilized Zemax size as the figure of merit. In order to keep the manufacturability
OpticStudio® to combat this uniformity issue for a specific illuminator of the design within reason, physical properties of the lenses were
and lens pairing and designed a telecentric backlight that matched a controlled during optimization. The usual suspects of aspect ratio and
specific telecentric imaging lens to produce a more uniform image. edge thickness were controlled. When an optimization was finished,
By using sequential and non-sequential tools to design the telecentric the design was converted from a sequential to a non-sequential (NSC)
illuminator in tandem with an imaging lens, a design in which the model in order to evaluate the radiometric performance. The metric
irradiance decreases by no more than 10% at any point on the sensor being evaluated was uniformity across the sensor. If the uniformity
was achieved. Walking through this design process reveals several was not to specification, then the design was further refined in
lessons that can be applied to other optical designs. sequential mode. It is important to note that, while the illuminator
elements were varied, the telecentric lens with which it was paired
The Problem was static. The design was considered complete when it met the
The large aperture telecentric illuminator performed very well chosen metric for uniformity across the sensor.
when analyzed by itself. The initial illuminator provided a uniform
illumination at the object, which was the intention of the design. Accurate analysis of the design in the non-sequential mode of
When paired with Edmund Optics’ 0.093X Telecentric Lens (part OpticStudio requires defining the correct source distribution, surface
number 34018), there was a mismatch of the etendue, or angular coatings on all optical components, and sensor type. As the source

USA: +1-856-547-3488 | EUROPE: +44 (0) 1904 788600

EDMUND OPTICS ® | www.edmundoptics.com ASIA: +65 6273 6644 | JAPAN: +81-3-3944-6210
 DESIGN AND EVALUATION OF A TELECENTRIC BACKLIGHT
model determines the wavelength and energy distribution of the
light entering the system, choosing the wrong model will provide
incorrect information about the rays that land on the sensor for
evaluation. Applying realistic surface coatings to both the front and
back of all elements being modeled is important to ensure that less
energy is lost when a ray splits due to Fresnel reflections at each of
the surfaces. This is helpful for data gathering, but more importantly,
ensures that the model is accurate when it comes to reflecting “as
built” performance. There are several options for the type of sensor
that can be used, with the “Detector Color” or “Detector Rectangle”
versions being the most useful for our analysis. These two methods
allow the user to determine the number of pixels on the sensor.
“Detector Color” accurately displays the color of the light on the
sensor, while the “Detector Rectangle” sensor does not have this
capability. “Detector Rectangle”, however, has the ability to display
Detector Image: Irradiance
both coherent and incoherent data, while “Detector Color” can only
display incoherent data. The incoherent data was sufficient for our
Figure 5: OpticStudio non-sequential simulation of new design
evaluation purposes.

After setting the system up for analysis, a ray trace was conducted The Result
(Figure 4). In our simulation, we let the rays split, and saved the results After the manufacturing tolerances were checked via a Monte Carlo
to a ray database (ZRD) file in order to make analysis faster for this analysis, the newly-designed lens was assembled. Figure 6 shows
particular run. Please note that these files are relatively large, so the that the assembled lens performance matches the simulated image
compressed version of the complete data is often a good compromise modeled in OpticStudio’s non-sequential mode. Combining the
between data density and used storage space. When the ray trace ability to complete the lens design in sequential mode and evaluate
finishes, you can observe the irradiance at the detector plane. The the real-world performance in non-sequential mode provided the
ray trace allows the use of the detector viewer, which will permit necessary tools to improve the design of the illuminator.
different views of the irradiance. A false color view of the irradiance
is shown in Figure 5. A linear scale was chosen for the analysis, but Not every illuminator needs to be designed in tandem with an
log scales are also available. As Figure 5 shows, the issue with the imaging lens but, in this case, it made the manufactured product
annular drop in irradiance was corrected. This design was chosen for perform significantly better; the variation of irradiance across the
manufacturing. sensor (relative to the peak value) was reduced from 18% to ≤10%.
The type of illumination being used for the design is also important.
This design revealed best practices for correctly modeling sources,
coatings, and sensor type in OpticStudio’s non-sequential mode that
will be beneficial in future designs.

® COPYRIGHT 2020 EDMUND OPTICS, INC. ALL RIGHTS RESERVED

Figure 4: Ray trace control window

Figure 6: Real-world test result for the new design, showing a more even irradiance
profile than the original design

USA: +1-856-547-3488 | EUROPE: +44 (0) 1904 788600

EDMUND OPTICS ® | www.edmundoptics.com ASIA: +65 6273 6644 | JAPAN: +81-3-3944-6210