IDEX Biometrics – Let’s Talk Testing
In our fifth 2021 Let’s Talk post on how IDEX Biometrics is more than just a component supplier, Andrew Robinson, Director of Program Management at IDEX Biometrics talks about the extensive range of testing that IDEX carries out on its biometric sensors.
The development of biometric authentication technologies clearly requires a significant amount of testing, and we test each fingerprint sensor type in a variety of real-world scenarios, helping to ensure that they function as required in the myriad situations that could be encountered during their life-time.
We use computer simulations to improve performance, reliability and time to market, we validate the performance of the fingerprint sensors against the customer specification and we carry out card testing by mimicking real life in our labs.
Using Computer Simulations to improve Performance, Reliability and Time to Market
Before our biometric sensors are released into the manufacturing stage, we use computers to show us how the design will perform in real life. By performing computer simulations, we can find out potential issues and fix them in a matter of hours. This speeds up the design cycle and ensures the sensors are reliable, and when we release the sensor design to manufacturing, we know with a high degree of certainty that it will perform as we have predicted.
The computer simulations we perform fall into two categories:
- Electrical functionality and performance
Before we could begin to accurately simulate the complex nature of our biometric sensors, we had to develop our own proprietary simulation tools as nothing powerful enough existed in the market. These tools, coupled with the expertise of our Engineers, allow us to simulate the capacitive relationships within the whole fingerprint sensor, from the layers of the substrate to the individual bond wires connecting to our sensing Application Specific Integrated Circuit (ASIC), including layers within the ASIC itself. Unwanted coupling (or pathological coupling as it is also known) causes a reduction in biometric sensor performance, but this can be designed out, by making changes to the physical structure of the sensor within the simulation model itself.
The level of capacitance we model down to is measured in atto-farads; thousands of times smaller than the modelling needed for traditional mobile phone fingerprint sensors. We typically need 12 machine days of simulation time to run a full set of simulations running on 24 core, 256GB Random Access Memory (RAM) cloud-based computers.
Add to the mix human finger and biometric algorithm modelling, and we have everything we need to simulate the entire system with exact correlation to the real world.
- Mechanical Reliability
Computer simulations can predict how our fingerprint sensors will perform when subjected to various physical stresses seen during their lifetime. Whether it’s the bending stress which occurs when a card is carried in your back pocket, or the force of the rollers on the top of the sensor surface in an ATM machine, we can compute the best position for the components on the sensor, and the best position for the sensor in the card to ensure we meet or exceed the industry standards designed to ensure the longevity of the product being tested.
Simulations are the key to realizing the best possible product in the shortest space of time. The tools at our disposal allow us to design cutting edge sensors operating at the limit of what is physically achievable, in terms of ultra-low power consumption, speed of operation and quality of fingerprint image, in a thin and reliable package. They give us the best possible chance to get the design “right first time” and therefore significantly reduce the time to market.
Validation of Fingerprint Sensor and Biometric Reference Designs
Once the first devices roll off the pre-production line, with IDEX having completed the fingerprint sensor computer simulations, and having approved the design for sample manufacturing, we then begin the process of Design Validation and System Validation.
- Design Validation Testing
Our fingerprint sensor Design Validation test regime is focused at sensor level testing, and ensures we understand how our sensors perform in various scenarios such as at different operating voltages and at temperature extremes.
The range of tests is diverse, covering not only electrical tests as you would expect but also mechanical and environmental tests, such as bending and twisting, chemical resistance, abrasion resistance, impact testing and electrostatic discharge.
We also subject multiple sensors from different manufacturing batches to different stresses over extended durations (of several months) and check that the device performance after the stress is applied is the same as when the device was manufactured. The types of stress applied are designed to find weaknesses in the design, the materials used and the manufacturing processes. This is key to ensure our products will perform reliably every time they are used, regardless of the usage or storage conditions.
- System Validation Testing
Our biometric sensors perform the complex task of acquiring a fingerprint image, but it is down to other components to perform the biometric fingerprint matching process and to communicate with the outside world. IDEX has developed a series of reference designs that incorporate our fingerprint sensors and allows our customers to develop biometric smart cards for various applications, including touch-free authentication and contactless payments, as these designs are tuned to meet the specifications of key certification bodies around the world.
Clearly, biometric performance is a priority focus in our testing strategy. We use smart cards based on our reference design to collect and build a diverse database of several thousand fingerprint images which can then be used to calculate the False Accept Rate (FAR) and False Reject Rate (FRR) of our biometric algorithm. Each database run requires millions of match and cross-match attempts to measure performance.
We can use this database to fine-tune the algorithm, the speed of the matching algorithm and to optimize operation of the biometric sensor itself. The extremely low FRR and FAR of our solutions ultimately leads to high security, convenience and peace of mind for our customers.
Smart Card Testing – Mimicking Real Life in the Lab
Smart cards are part of our everyday life, and our reliance on them means we encounter a great deal of inconvenience should they fail in the field. Payment cards are expected to have a working life of several years, and during this time they are exposed to a range of ambient temperatures, are handled by wet hands, or hands contaminated with chemicals such as sun cream or gels. They sit in our wallets and handbags, and are bent, twisted and flexed. The environment in which our fingerprint sensors must operate under is extreme and we need mimic this environment in the lab to ensure they keep working without any degradation in performance.
Many of the environmental tests can be carried out at biometric sensor level, but as for the mechanical stress tests, it only really makes sense to perform these on cards with the sensor attached, as the position of the sensor in the card plays a part in the overall reliability of the sensor and its connection to the supporting electronics.
Firstly, the cards are assembled, locating the biometric sensor in the optimum position as indicated by our computer simulations. We use a heat-activated thin film to “glue” the sensor into position, and we then send a command to the sensor to calibrate it, adjusting its operation to its new environment. The baseline performance of the sensor is measured, and the results are recorded. The tests we then perform are automated, and we use a mixture of bespoke and proprietary systems for the test stresses applied, conforming to industry standards laid down for the application in question.
We have machines that can simulate the repeated action of touching a sensor and we do this hundreds of thousands of times with the equivalent of a 1kg mass bearing down each time. We can bend and twist by varying amounts on multiple cards simultaneously, and for thousands of cycles. We can replicate the action of an ATM machine’s rollers, simulating the insertion and removal of a card from one in a repeated manner, with increasing amounts of force on the card. We even “wrap” cards around a 50mm diameter metal drum. Like the cards they sit in, our biometric sensors are flexible too, and this is a key factor in their survival.
To conclude each test, we remeasure the sensor performance on every card and check it against the original results to make sure nothing has changed. A comparison with results from our computer simulations also confirms our understanding of how the sensor has fared.
This type of accelerated life testing gives us confidence that our fingerprint sensors will carry on working long after the expiry date of the card is reached and allows consumers to carry on using their biometric payment card in the same manner that they always have done, without regard for the new complexity hidden within.
The previous post in our series on more than a component supplier looked at the Security Architecture of IDEX Biometrics’ Fingerprint Sensors and Biometric Solutions.
