Off-chip sensing is technically hard to implement. I completely agree.
By Guido Bertocci, VP, Software Engineering
I read a long article in Avanza from an ornithophile. It might surprise him, but I actually agree with quite a bit of his message. Some technical explanations are wrong and he seems to be a bit confused about what an off-chip architecture is, but the overall message is pretty good. I also agree that some birds are much nicer than others.
His core message is, “off-chip sensing is complicated and hard to implement.” I completely agree. In fact, I would go further and say it is very hard to implement.
He also correctly identified one of the key benefits of IDEX. (apologies for Google translate): “There is a big advantage that IDEX currently has with its solution, separation between chip and sensor grid. It is that you can probably have more functions in your processor. It’s probably not something you can read about, but I’m more based on how their technical solution is.”
TrustedBio represents 4 years and 3 generations of products with a 100% focus on cards. You can read our sensor history here. With each generation we have redesigned critical functions such as the analog front end, the MCU processor system and added power management, AES 256 encryption and more. Each generation is more power efficient, has faster processing, more memory, and TrustedBio now dramatically reduces the cost of a card by eliminating most of the “other stuff” on a card.
We like the fact that off-chip is difficult. Once you solve difficult engineering problems, you have a competitive advantage. Unlike silicon sensors which are relatively easy to copy, as we have seen in Asia, off-chip is much harder. We expect to have a long term sustainable competitive advantage.
With regards to the manufacturing process we do indeed have more steps than silicon sensors and there will be yield loss at each step, even if it is small. We fully understand our yields and they are reflected in our costs and pricing and yet, we are still less expensive than silicon sensors. We use the same manufacturing processes that are used every day in the entire electronic industry. Most of the industry is off chip, separating the user interface, the part you touch or see, from the ASIC. Touch screens on cell phones, HDTVs, Apple watch all connect something large to a small ASIC. It is a very well understood technology, high volume, low cost.
One topic where he is way off target is reliability. His explanations are what we refer to as FUD, Fear Uncertainty and Doubt, claims that sound plausible and hope to confuse people. Let’s talk facts. Our IDEX 3200 products are in cards certified by China Union Pay and a large US based network that have a combined issuance of 70% of global branded payment cards. They have been tested extensively by our customers in Asia, Europe, and US. Reliability is one of their essential criteria. This includes the 3 wheel and wrapping tests where cards are bent thousands of times, and many additional tests. In fact, with the flexibility of our polymer substrate, our small ASIC, and the elimination of all the “other stuff” on a card, reliability will be even higher.
His explanation on costs and volumes are very strange, maybe it is Google translate. I think it is straightforward. There have been billions of silicon sensors sold in a very competitive mobile market. They have already gone down a steep learning and cost curve. The costs quoted for silicon sensors in cards already reflect this learning and volume pricing. It is basically the same sensor and same process. What might concern our competitors is that TrustedBio is already less expensive today, and we are just getting started down the cost volume yield curve.
He wrote “IDEX said that a 50mm2 sensor was the breakpoint” where silicon was less expensive. In other words, if the sensor size is less than 50mm2, silicon would be cheaper. That seems correct for our Gen1 sensor. For TrustedBio the breakpoint is about 16-18 sq mm. This means that a silicon sensor would need to be smaller than 4×4 to be less expensive than TrustedBio at 90 mm2.
Let’s summarize TrustedBio
- Sensor area: TrustedBio 250% larger, 90 mm2 vs 36mm2
- Scan power: TrustedBio <5ma, silicon <5ma.
- Scan acquisition time: TrustedBio < 50ms, silicon < 50ms However, in cards, this is not as important anymore for two reasons. Scan acquisition is now quite short for all sensors, and biometric algorithm processing is the dominant concern, about 80-90% of the total time. It is biometric processing time that really matters now. A larger sensor allows simpler algorithms which reduces match time.
- Image transfer time: TrustedBio ~25ms, silicon ~25ms
- Additional features: TrustedBio includes power management, energy harvesting, 256bit AES encryption, ability to run entire biometric algorithm
- Sensor cost: TrustedBio is less than silicon.
- Card cost: TrustedBio eliminates almost all the “other stuff” which silicon cannot. It works equally well with super-SEs/one-chip and low-cost SEs.
The fact that our bird enthusiast is generally correct should probably concern silicon sensor advocates. Implementing off-chip is difficult, but the benefits are very clear starting with the ability for a much larger sensor at a lower cost than a small silicon sensor and integrating “the other stuff”, reducing the cost of the entire card.
What happens when customers start selecting IDEX because they see the technical challenges solved? That is already happening. Our 2nd generation IDEX 3200 was selected by a major financial institution that is extremely security conscious and is ramping production right now. Asia customers are certifying cards with our IDEX 3200 sensors with China Union Pay. And those are our 2nd generation products.
Now consider TrustedBio which is a big step forward. Why was TrustedBio selected by Idemia, who is extremely experienced and knows how to evaluate technology? Idemia selected TrustedBio while being fully aware of all the products and roadmaps of every sensor manufacturer. There must be some very compelling reasons to select TrustedBio BEFORE it even taped out. Now that samples are shipping, we are much closer to pilots and certifications.
The biometric card market is about to explode and TrustedBio will accelerate the adoption. Pilots with TrustedBio are expected in the next few months, with certifications and product ramps starting 1Q 2021.