With integrated devices, there is often a quandary over what resources to include and what to leave out.
The field programmable analog array (FPAA) has broken into the market more than a decade ago, but at the time, it did not seem to capture the imagination the way the field-programmable gate array (FPGA) had done before it.
That said, Anadigm Inc. (Mesa, Arizona), has stuck to its FPAA guns. So what is it that has enabled Silego Inc. (Santa Clara, Calif.) to ship more than 1 billion units of its configurable mixed-signal ICs (CMICs) in the last two years? The answer would seem to be: by taking account of application knowledge and specific functionality.
With integrated devices, and particularly an array of analog and mixed-signal functions, there is often a quandary over what resources to include and what to leave out. In addition, in the analog case, when these resources are configured performance can vary depending on how the resources are hooked up. This pushes back on the design software and often requires an iterative approach to prototyping the FPAA/CMIC.
And as a result there is the potential to miss both the performance requirement and volume sales windows for these base devices. In these circumstances the technology-push product offering—putting down arbitrary numbers of op amps, voltage comparators, LDO regulators, level shifters and so on with an interconnect matrix, because we can—is likely to fail.
What appears to succeed is finding an application in production with a number of discrete analog, logic and power devices and designing an IC with these resources and more for engineers to configure. There is a known market with known specifications in terms of performance, PCB real-estate, power consumption and so on and if the FPAA/CMIC version can score against the discrete status quo ante surely happy days lie ahead, except for vendors of discrete components.
I think of it as the difference between a box of conventional Lego bricks that are application-agnostic and a themed box of Lego that provides all you need to build a pirate ship and populate it with pirates. The choice of how many masts, how many cannon on each side and how many pirates is configurable. I may like the purity of the agnostic Lego bricks that allow me to build anything but the variants in the pirate box and the reference design on the box lid speed time to toy.
In this manner Silego has been able to use its experience of the notebook motherboard to craft a series of superset analog and mixed-signal base die devices that can sweep up the discrete analog and logic devices while catering for many different ways multiple customers might wish to customise how they do that.
Now Silego wants to branch out into the more fragmented world of IoT and industrial electronics. It's hard to bet against a company that has climbing unit sales and revenues and is profitable, but the fragmentation of the markets it is now attacking does pose challenges in terms of getting the die resource configurations correct. Things should continue to go well for Silego, as long as it uses its contact with customers, which is now considerable, and their applications and functions to drive the design of next-generation CMIC die. Providing real integration within real applications is the key.
Surely six pirates with cutlasses will be enough?
It seems this same thinking is current at Anadigm. For example Anadigm offers FPAAs specific to baseband signal filtering for a universal RFID reader and an audio chipset that simplifies the analogue design of subwoofer signal conditioning.