Predictive Power: Abrar Akram on Building Circuits That Anticipate Demand

The observation seems simple, but its implications reach deep into hardware design. As personal electronics have grown more capable and thinner in form factor, the batteries inside them have not evolved at the same rate. 

The circuits that regulate how battery power is distributed inside a device are known as power management integrated circuits (PMICs). As demands increase, so does their importance. 

"No matter how fast we can make our computers and how accurate we can do our computing services, still there is a constraint of power and energy." 

A central component in these systems is the low dropout regulator (LDO), which maintains a stable voltage supply even as demand shifts. “Low dropout” refers to its ability to function when the difference between input and output voltage is very small, making it suitable for battery-powered devices near their limits. 

Traditional LDOs rely on external capacitors, passive components that act as small charge reservoirs, to smooth sudden changes in demand. As devices shrink, this becomes harder to maintain. A device with twenty LDOs needs twenty capacitors, each taking up space and adding cost. 

Akram’s doctoral work focused on removing these capacitors. The challenge is that they release stored energy almost instantly during demand spikes, giving the regulator time to adjust. 

That adjustment happens through a feedback loop, a continuous process where the circuit monitors its output and corrects deviations. Without capacitors, the loop must become faster and more precise. 

"I introduced the concept of non-linear, event-driven control. Whenever there is an event happening of any workload, the feedback loop will sense that event and immediately respond. And the multi-loop control was very beneficial." 

Photo: Antti Yrjönen

From Reactive to Predictive 

This push for speed has led to a different approach: anticipating demand instead of reacting to it. In work under review, Akram’s lab applies an open-chain hashing algorithm, a form of reinforcement learning where a system builds memory of past situations and responses, to LDO design. 

Modern workloads, especially in artificial intelligence, often follow recognisable patterns. Once a regulator has seen a pattern and stored the correct response in a lookup table, it can reuse it instantly the next time, bypassing the feedback loop. 

"The first time it will respond as the typical LDO, but the next time if the same kind of repetitive workloads are happening, it will take the information from the lookup tables and just pump the information at the output." 

In prototypes, this reduces response time to under one nanosecond, far faster than conventional analog circuits. The reason is structural. Analog systems are limited in how quickly they can respond to signals.  

Digital logic operates differently, processing discrete binary values at gate delays, the tiny switching times of transistors. These are much faster. Combining analog sensing, which interfaces with real-world signals, and digital control defines mixed-signal IC design, the area where Akram works. 

If validated and extended, the approach could eventually be scaled to the higher current demands of data-centre-class computing hardware.  

"The analog guys should know how the digital control will run their analog circuit. Digital control is the future." 

Power Without a Battery 

After his PhD, Akram turned to biomedical implants, where conventional power solutions are not viable for long-term use inside the body. 

“We cannot put batteries inside the body for long-term operation. Replacing them is not feasible, and biocompatibility is also a concern. So we deliver the power wirelessly, from outside the body to the implant through inductive links.” 

Inductive links transfer energy through oscillating magnetic fields, allowing power to cross tissue without physical contact. Akram’s work explores transmitting that power in intervals rather than continuously, while ensuring compliance with Specific Absorption Rate (SAR) limits, the standard that defines how much electromagnetic energy human tissue can safely absorb. 

“We are not sending power continuously. We send it in bursts. For example, at a 20 percent duty cycle, it is active only for a short time, but the energy is sufficient for operation in between. At the same time, we have to make sure it satisfies SAR limits, which we verify in electromagnetic simulations before we go to silicon.” 

Inside the implant, sensing circuits operate under nanowatt power budgets. In this setting, Akram developed a potentiostat for a subcutaneous glucose monitor without relying on analog amplifiers. A potentiostat measures electrochemical signals generated when glucose interacts with a sensing electrode. 

“The power budget is extremely low, in nanowatts. We removed all the amplifiers, so the entire architecture is amplifier-free. We achieved 129 dB dynamic range while consuming less than four nanowatts.” 

Dynamic range describes the span between the smallest and largest signals a circuit can detect. In electronic systems, decibels express signal ratios, where each 20-decibel increase represents a tenfold change. 

 

Photo: Antti Yrjönen

The IPSS Lab 

Akram describes the opportunity at Tampere as well-aligned with where his research was heading. He describes the available startup support and the ability to recruit doctoral researchers as essential for a programme at this stage. 

While the university has established strong international reputations in fields like materials engineering and wireless communications, Akram sees a clear opportunity to expand its footprint in the global integrated circuit design community. A major objective for his new group is to execute independent lab-level "tape-outs", i.e. the process of finalising a chip design and sending it for physical fabrication, and to represent Tampere University at top-tier IC design conferences. 

The IPSS Lab is currently recruiting its first doctoral cohort, with the aim of building a group that works across the lab's four guiding themes: integrated, power, sensing, and systems. 

"Power is the main fundamental thing which is quite essential for everything in modern day electronics. You can make it fast, you can make it whatever you want, but until and unless the energy efficiency is not addressed, the device is almost useless." 

The lab's scope spans from power regulation circuits for artificial intelligence hardware to implantable biosensors, contexts that differ widely in scale and application but share the same underlying constraint. That continuity, from the circuit level up to the system, is what the IPSS framework is built around.