Terry Arden, CEO, LMI Technologies
In industries such as battery, consumer electronics (CE) and solar, the race for faster scanning, measurement and control is critical to delivering 100 percent inspection of small parts moving at production speed. Sensor manufacturers in countries around the world are constantly striving to deliver this groundbreaking performance, but have yet to produce a game-changing solution. This article provides a strategic approach to designing a faster, smarter sensor for small parts inspection.
For several years now, the industry standard for inspection rates has plateaued between 2 and 4 kHz at 800 columns resolution. This is simply not fast or resolute enough for many of the emerging small parts inspection applications facing today’s CE quality engineers.
In the past, manufacturers have relied heavily on making subtle modifications to existing sensor design in the hope of achieving an increase in speed, but have failed to achieve the desired results. It has become evident that breaking through the 4 kHz speed barrier requires a significant technology innovation.
Calculating sensor speed
There is a common perception that scan rate is the only metric that matters in a laser sensor’s speed specifications. The reality is more complicated than that. In order to be meaningful in the real-world factory environment, a sensor’s speed calculation must factor in (1) scan rate (i.e. the speed of data acquisition), (2) data processing/measurement, and (3) control decision-making stages required to fully inspect a part or assembly on the production line. These three stages make up a single inspection cycle, which is the true measure of effective speed in today’s factory. (Image 1)
The three stages for calculating sensor speed.
Custom components and smart sensor design
Achieving a major breakthrough in effective inspection rate, while maintaining micron-level resolutions, is made possible through a combination of custom components and smart sensor design.
Sensor building-block components such as custom cameras and advanced optics deliver the necessary resolution, repeatability and sensitivity for defect detection on fast-moving small parts. In smart design, the entire inspection cycle occurs onboard the sensor, which allows it to collect 3D profiles to build a 3D point cloud, perform metrology and transmit results––all at high speed. Depending on their application requirements, users can leverage this high speed in several ways, namely:
- Increasing Y resolution for detection of smaller features along the direction of motion;
- Increasing Z accuracy for more accurate height measurement, and achieving tighter dimensioning tolerances as a result; and
- Using multiple exposures (high dynamic range) to handle a wider variety of reflective targets (shiny black to white), even if they are in the same scan, without a loss in effective speed.
Benefits of onboard data processing
Building data processing directly into the smart sensor is one of the key methods used to achieve increased sensor speed. For example, a custom embedded dual-core controller is able to support raw data processing without latency, which cannot be achieved when using an external controller or industrial PC. Embedded processing gives the sensor accelerated computing performance, and the ability to undertake onboard 3D surface generation and alignment, in addition to anchoring to correct part movement, and 3D feature fitting using built-in measurement tools. Most importantly, onboard data processing allows the sensor to handle the entire processing pipeline, from raw image data to 3D results, with minimal latency.
An embedded controller allows the sensor to process highly detailed 3D surface scans and communicate decisions onboard without the need for an external processing device. With no dependencies on external hardware, latency between scan data acquisition and the decision output is minimised, allowing the sensor to keep pace with inline production speeds. This degree of sensor autonomy is the key differentiator between smart solutions and other 3D profilers on the market that require transfers of raw data to a PC for processing. In cases where acceleration is required to meet cycle time, the pre-processed data from a smart sensor can be compressed and sent to dedicated processing nodes (PC or vision accelerator) on a network, while keeping latency to a minimum.
Standard sensors that have to send raw pixel or profile data downstream to PCs or external controllers to carry out their 3D measurement dramatically increase latency and performance jitter, with possible drops in processing altogether. The situation is exacerbated when multiple sensors must stream and synchronise datasets before measurement can occur.
The Gocator 2510 laser profile sensor performing a high-speed scan of small parts.
Sophisticated software
Smart sensors come with sophisticated onboard software that provides surface generation, measurement and pass/fail decision-making. As a result, new sensor designs are able to achieve 2.5x the resolution at twice the speed of the current industry standard.
This software is included at no extra cost, providing powerful built-in tools for filtering, profile and surface analysis, multi-sensor alignment, and support for various programmable logic controller (PLC) and robot protocols. With no additional software to install, a smart sensor’s out-of-the-box setup and configuration is straightforward and intuitive.
A custom camera chip
Web browser interface
Another key design feature of a smart sensor is a web browser interface. This provides flexible configuration of settings and measurement tools using any browser running on PCs or mobile devices.
Custom optics
Small footprint
A smart sensor that has small dimensions can be easily mounted or retrofitted into virtually any machine environment. However, the sensor must still achieve an IP67 rating, which ensures that it is resilient to common environmental stresses such as moisture and dust.
Other important smart sensor features include: bandpass filters (BPFs), which limit the effects of ambient light; and a material choice such as precision-machined aluminium and custom optical design, which allow the sensor to handle vibration and limit the negative effects of temperature drift.
Large field of view
In addition to having small dimensions, sensors used in inline inspection need to have a large field of view (FOV) and large measurement range (MR) in order to achieve maximum scan coverage with the fewest number of sensors, while still being able to capture the finest surface and edge details of electronics and small parts.
A smart sensor can also provide built-in support for multi-sensor networking (including automatic alignment and built-in stitching) to deliver high-density 3D models for users who want to expand the FOV or acquire multiple angles on the same target while maintaining ultra-high resolution.
Gocator 3D smart sensors can scan, measure and make control decisions.
Blue laser projection
Ideally, a blue laser sensor should be used to scan small parts. Due to its shorter wavelength, blue laser light performs better than red or green laser light on the highly specular metal surfaces commonly found on electronic and shiny parts. Blue lasers produce cleaner profiles (i.e. less speckle) and, as a result, deliver superior data on these challenging surfaces.
Smart technology and a commitment to innovation
Advancements in smart technology are at the heart of today’s movement toward better sensor design. The result is a complete machine vision solution that provides exceptional 3D scanning, onboard software and a web browser user experience for rapidly setting up, scanning, measuring and communicating control decisions to factory machinery.
Increasing demand for higher resolution and speed in detecting small features must be met with equal innovation on the part of sensor manufacturers. In the end, significant gains are made by paying attention to the details. An investment in a custom camera chip with special optics, for example, can deliver a major impact on moving the base technology forward.
LMI Technologies