Smart is a word often used to describe the emerging information technology paradigm—a vast network of semi-autonomous, AI-driven devices that are globally interconnected and driving us toward a brighter future. But how this word applies to the world of machine vision is not always so clear. What exactly is “smart” in the context of vision technology and its role in factory production today? Let’s take a closer look.
Smart vs. Simple
In any hardware device, there are electronics that process sensor input to deliver raw data into a process. This is the baseline for a simple device. A smart device takes it one step further and adds built-in processing to perform predefined functions upon detection of specific criteria (such as pass/fail control based on specific measurement thresholds).
In a smart device, raw data is transformed to make a decision before transmission over an interface. A smart device incorporates built-in software-defined elements in a pipeline of digitization, data reduction, and communication to external machinery.
With simple sensor technology, the approach is very different. These devices use an interface (direct I/O, Serial, BlueTooth, WiFi, USB, GigE) to transfer sensor input for processing elsewhere (i.e., external to the device itself), which can result in increased latency and a greater incidence of data dropout.
The main motivation for purchasing a simple device is to save money on upfront device cost. However, the “sleight of hand” required to move decision-making from the device to an external location actually costs more in the long run.
The Benefits of Choosing Smart
What many people don’t realize is that inside the hardware device, on the embedded controller, the sensor input and communication interface electronics are the same for both smart and non-smart capability. Adding onboard processing to the embedded controller ends up being a marginal cost increase in light of the significant benefits gained (i.e., faster data processing speed and greater cost-efficiency) by converting your non-smart device into a smart one.
The Cost of Controllers and Pushing the Decision Downstream
On simple devices, data is delivered downstream to some external “smart controller.” So in reality, a simple device requires two controllers to accomplish what a smart device can do on its own. One controller inside the device, and a second one downstream that is used to transform raw data into a higher level result.
The downstream controller is typically an industrial PC or PLC. And of course, these downstream devices do not come cheap. While simple works for things like photocells, and temperature/inductive sensors, it does not work for machine vision—where the demand for speed and resolution is continuously pushing massive amounts of data downstream and overwhelming network bandwidth and CPU demand. That is why being smart by processing data onboard or nearby the device (with dedicated compression and acceleration) is the new norm.
Why Total Cost of Ownership Matters
When you add the cost of configuring both the simple device and its downstream controller to produce the desired result, you've added even greater cost and complexity to your inspection system than you would have had you simply paid to add onboard processing onto that same device in order to make it “smart.”
At the end of the day, the decision comes down to the ease of maintaining systems and the total cost of ownership involved. Given the reality of today’s high-speed production lines, the question is who wouldn't want to be smart?
For more insight on this essential topic, we invite you to download our complimentary Component-Based vs. Smart Sensor factsheet. We also invite you to visit the Firmware section of our website to explore the many smart features and capabilities built-in to Gocator®.