Changing Microamps to Nanoamps

Ambiq Micro is redefining “ultra-low power” with the development of the World’s most energy efficient semiconductor components. Using its unique and proprietary Subthreshold Power Optimized Technology (SPOTTM) Platform, Ambiq Micro delivers unprecedented energy efficiency without compromising on functionality or price. Ambiq Micro’s AM0800 and AM1800 Real-Time Clock (RTC) families are the first products built on the SPOTTM Platform, and they deliver a 7x improvement in energy efficiency over other RTC products. The low power advantage delivered by the SPOTTM platform enables longer battery life, the use of smaller and less expensive batteries, and the incorporation of new (and previously power hungry) functions.

Breakthrough Subthreshold Design

The SPOTTM platform uses transistors biased in the subthreshold region of operation to achieve unmatched energy efficiency. Rather than using transistors that are turned all the way “on,” subthreshold circuits use the leakage of “off” transistors to compute in both the digital and analog domains. With most computation handled by leakage current, total system draw on the order of nanoamps is easily achieved.

Subthreshold operation has been used sparingly in select markets like wristwatch and RFID chips, but Ambiq Micro’s SPOTTM Platform represents the most comprehensive and aggressive demonstration of subthreshold to date. Ambiq Micro’s patented approach to subthreshold was developed over an 8 year research effort by University of Michigan engineers and Ambiq Micro engineers. Ambiq Micro’s approach relies heavily on unique circuit designs and system architectures, and chips may be manufactured in a standard commodity process to ensure minimum cost.

Implemented in Standard Low Cost CMOS

Despite the comprehensive use of subthreshold in the SPOTTM Platform, Ambiq Micro’s products behave identically to their superthreshold competitors. The subthreshold circuits in Ambiq Micro products are surrounded by boundary circuits running at a full I/O voltage as in any other chip. Consequently, SPOTTM-enabled Ambiq Micro chips can be used in products just like any other competing product. (figure 1)

Figure 1 – Voltage domains in superthreshold and subthreshold designs

Digital circuits implemented with the SPOTTM Platform are designed to run at supply voltages at or below the transistor’s threshold voltage. A superthreshold digital circuit in a typical manufacturing process operates with a 1.8V supply voltage and a switching threshold of approximately 0.9V. In the same process, a subthreshold digital circuit may operate with a supply voltage near 0.5V and a switching threshold near 0.25V. With such a small voltage swing, considerably less charge (and thus energy) is used to run computations. Operation at such low voltages is complicated by susceptibility to noise, high sensitivity to temperature, and a variety of challenges. In developing its SPOTTM Platform, Ambiq Micro has addressed all of these challenges by redesigning every digital circuit in the chip, from standard cell building blocks to internal voltage regulators. (figure 2)

Figure 2 – Voltage switching waveforms in superthreshold and subthreshold circuits

SPOTTM-enabled analog circuits typically run at higher supply voltages than their SPOTTM-enabled digital counterparts but are biased in the subthreshold regime. The transistor current-voltage curve shows nearly all conventional analog circuits use superthreshold bias currents on the order of 1µA. Analog circuits designed with the SPOTTM Platform use subthreshold bias currents on the order of 0.1nA. As with digital design, operation in this regime is complicated by several challenges including high sensitivity to temperature/voltage fluctuations and manufacturing variations. The SPOTTM Platform uses proprietary analog building block circuits that are redesigned exclusively for subthreshold operation and are resistant to the aforementioned fluctuations and variations. (figure 3)

Figure 3 – Current-voltage (IV) curve for a typical transistor with superthreshold and subthreshold bias points identified