Paolo Nora Paolo Nora, co-author of the paper “A novel sufficient condition to avoid sub-harmonic oscillations for Buck converters with constant on-time control” talks to us about their work page 305. I am part of the Architecture Definition Team of the Analogue Power and Interface Division (APID) of Microchip Technology Inc., and I collaborated with Profs. Bizzarri and Brambilla of the Politecnico di Milano on the topic of stable operation of Constant-ON-Time (COT) control, which is one of the most common control methods for analogue point-of-load (POL) Buck regulators. Indeed, my main area of interest is the architectural definition of analogue switch-mode regulators. A few years ago, I became aware of the topic of power integrity for my field of work. Both areas are very closely related, since POL regulators are often deployed in applications where power integrity considerations become increasingly critical for a successful power system design. The topic of Constant-ON-Time (COT) control, and – more generally – of ripple-based control of Buck converters has been a major focus for many power electronics researchers in the last few years, and different analysis methods have been used to determine its stability conditions. The importance of COT control (and its variations) of Buck POL converters is huge, due to the popularity of this control method in the industry. The reasons for such a wide popularity mainly stem from its superior load-transient performance over fixed-frequency architectures, and in the apparently simple design process it takes to develop the application. Simply put, the key design aspect is the phase relationship between the output voltage ripple fed to the controller and the Buck inductor current ripple. In addition, the total magnitude of the ripple seen by the controller plays an important role and must be adequate to ensure jitter-free operation. However, over many application cases encountered in practice, we've only seen a few of them being affected by confirmed instability issues (pulse-bursting operation) despite all recommended design practices being followed. This pushed us to consider different analysis methods to assess stability conditions, and here is where the nonlinear dynamical systems analysis expertise of profs. Bizzarri and Brambilla came to the fore. By using the state-variables methodology and state-plane graphical representations, and by leveraging nonlinear dynamical systems theory results, a new and more articulated sufficient condition for stable operation of the COT Buck converter was derived. This condition includes both circuit parameters and operating conditions. We have also shown that the well-known output capacitor time constant stability criterion is encompassed by the new sufficient condition. State-space analysis is a powerful method; however, its complexity rapidly explodes as the number of state variables increases. For the sake of manageability, the source of ripple fed to the COT control engine had to be limited to the resistor Re in series with the output capacitor, neglecting other well-known methods of ripple self-generation which would have added more state variables (i.e. more capacitors). The big question was then whether the simplified configuration would still be relevant from the practical perspective. Here is where some power integrity concepts, such as flat-impedance design of the POL stage and the power distribution network (PDN), played a key role. Indeed, resistor Re in series with the output capacitor may not just be a poorly controlled parasitic element – such as the equivalent series resistance (ESR) of an ordinary multi-layer ceramic capacitor (MLCC). In power-integrity-aware POL and PDN designs, the value of Re is intentionally controlled and – in some cases – it could even be implemented with a physical resistor. Also, if the POL stage is dedicated to some digital loads which are not particularly sensitive, the magnitude of the output voltage ripple generated by Re could also be adequate for a real-world application. We can conclude that the analysis of the simplified configuration presented in this Letter will become increasingly relevant as power integrity considerations will affect more and more POL/PDN designs. Another major challenge was experimental verification, whose description unfortunately exceeds the scope of the Letter. The state-variables method and the state-plane representation lend themselves very well to model rapid load transients and can predict the maximum load step that can be applied without saturating the COT control engine (i.e. engaging the minimum OFF time). This is also a new result and it is of utmost practical importance. Simply put, the outcome is that it should only take a very small load step to saturate the COT engine. If this does not happen, then it means that the design – for some reason – is not exploiting the full potential of COT control, and this should raise a red flag to the designer. As previously mentioned, this topic has attracted the attention of many researchers in power electronics over the last 10–15 years and there is a wealth of excellent, highly technical literature out there. I have to say that the perception of COT control is too simplistic for many practicing engineers: just make sure the controller is seeing enough voltage ripple, and that voltage ripple is in-phase with the inductor current ripple, and you're done! Unfortunately, experience reveals that the situation is not so simple. The challenge is how to convey all the highly technical results to a large community of users who simply do not have the time to dig into the fine details of POL design. The focus of many practicing engineers is not analogue design, not to mention POL design. Therefore, in the future, I see the highly technical analysis methods and results becoming available from the websites of many semiconductor vendors in the form of software tools as an extra aid for POL design verification, beyond the simple sets of equations provided by most datasheets. I also believe that the topic of digital implementation of COT-control and ripple-based control and their modeling will also grow exponentially.
Read more