Lower-cost MCUs with DSP functionality open up more opportunities for digital power, reducing the prohibiting factors of cost and complexity associated with digital power

BY
DR. MICHAEL HALLWORTH
Senior
Design Engineer, XP Power
www.xppower.com

Digital
power has experienced rapid growth over the last five years with industry
taking advantage of the benefits that digital control brings to a traditionally
analog domain. This has been supported by the many new products from various microcontroller
vendors that specifically target the digital power market.

Fig. 1: An analog
controlled power supply.

In
a traditional analog power supply, the control is implemented using a control
IC consisting of operational amplifiers and comparators combined with a series
of carefully selected external capacitors and resistors forming the
compensation network. These give the power supply the desired transient load
performance and stability in the frequency domain, the s-domain. The
compensation network is fixed and often compromised by the presence of the bandwidth-limiting
optocoupler in the feedback path. This is shown in Fig. 1.

PS0818_XP-Power_Fig-2

Fig. 2: A digital
controlled power supply.

When
we refer to digital power, what we mean is a digital control loop, which
regulates and stabilizes the power supply. This replaces the analog control ICs
that have been used in switch-mode power supplies since the 1980s. In the
digital power supply shown in Fig. 2, the analog
control IC and its associated analog compensation network has been replaced by
a microcontroller.

The
microcontroller is used to close the feedback loop of the power supply. In a typical
digitally controlled power supply, the analog-to-digital converter (ADC) module
on-board the microcontroller samples the output voltage or current. This is
compared to a demand reference value with the result being an error term. The
error term is then used as an input to the discrete-time controller (usually a
two-pole, two-zero or a three-pole, three-zero controller), which has poles and
zeros in the discrete-time domain, the z-domain. The discrete-time controller
is executed at exact and pre-defined intervals — each time, there is a new ADC
sample available.

PS0818_XP-Power_Fig-3

Fig. 3: A discrete-time
two-pole, two-zero controller.

An
example of this discrete-time controller is shown in Fig. 3.
The controller consists of five multiply-and-accumulate operations, which are
referred to as multiplier-accumulator (MAC) instructions on-board a digital
signal processor (DSP). The input to the controller for this sampling period
x[n], the error term, is multiplied by the controller coefficient B0. The z-1
term is a unit delay and results in the previous input to the controller,
x[n-1], being multiplied by the coefficient B1. Following this, there is another
unit delay, so the error term from the two previous sampling periods, x[n-2],
is multiplied by B2.

On
the right-hand side of Fig. 3, the same
process is applied to the outputs of the controller. The previous output of the
controller, y[n-1], is multiplied by A1 and the output from the two previous
sampling periods, y[n-2], is multiplied by A2. These multiplications are accumulated
together and the result is the new output of the controller for this sampling
period.  The output of the controller,
y[n], is the new value of duty cycle for pulse-width-modulated (PWM) converters
or switching frequency for resonant/pulse-frequency modulation (PFM)
topologies.

Just
like the analog compensators in the s-domain, the discrete-time controller
shown in Fig. 3 will have a frequency response in the
z-domain. It is the controller coefficients that determine the frequency
response and, hence, the stability of the power supply. Therefore, the engineer
must analytically calculate the controller coefficients to stabilize the power
supply.

Microcontrollers
have been used in power supplies for many years for the purposes of
implementing basic functionality such as PMBus and fan-speed control using
relatively simple and low-cost microcontrollers. However, full digital control has
previously been most prevalent in the server and telecom market with uptake in
the industrial and medical markets lagging.

The
prohibiting factors to the switch to digital control have been predominantly the
cost and complexity associated with digital power. The good news is that the
cost of a modern microcontroller with the DSP functionality needed to implement
full digital control has decreased dramatically in recent years, making their
use viable for many more designs. However, the complexity remains an issue.
This complexity stems from the need for a mixed-domain approach to designing
the power supply. Engineers need to combine their power supply design knowledge
with the ability to write efficient code and stabilize the discrete-time
control loop.

So
what are the reasons for switching to digital control? The digital control loop
has many advantages over its analog counterpart. The digital power supply is
insensitive to the environment, temperature, aging, and tolerances of the
control loop components. It allows the system to monitor the performance of the
power supply in real time and adjust parameters to tune the performance as required.

Furthermore,
advanced discrete-time control techniques allow us to achieve higher
performance compared to analog compensators, recovering from transients in a matter
of a few switching periods. This has been of particular interest for the point-of-load
(POL) converter market, which has been a big adopter of digital power. One high-performance
microcontroller can be used to stabilize and regulate multiple power stages — negating
the need for individual analog control ICs for each power stage.

The
ever-increasing demand for high-efficiency converters is an area in which the
flexibility of digital power offers solutions beyond the capability of typical analog
control schemes. This can involve adjusting the operation of the PSU to achieve
optimal zero-voltage or zero-current switching, reducing switching losses and
increasing the overall efficiency.

Alternatively,
we can also consider the impact of improving the efficiency of the overall data
center or system in which multiple power supplies are used. This could be
achieved by responding to requests to shut down or enter low-power mode based
on information from a master controller within the system.

PS0818_XP-Power_Fig-4

Fig. 4: Typical control
loop duration vs. PWM switching period.

The
latest microcontrollers for digital power applications contain DSP
functionality, allowing the digital control loop to execute within a fraction
of a single PWM switching period every switching period. Fig. 4
shows the PWM switching period of a typical digital power supply. In this
simple example, the output voltage is sampled once per switching cycle. An ADC
conversion time of a few hundred nanoseconds is typical for a microcontroller
designed for digital power applications.

Following
the ADC conversion, the interrupt service routine is called to execute the
discrete-time controller. This is a time-critical routine, and therefore, the
controller can be written in assembly code to make use of the MAC instructions
and optimize the use of every single instruction cycle.

As
shown in Fig. 4, for this example, the time that the
MCU does not spend executing the controller is our spare bandwidth. This spare
bandwidth can be used to perform other tasks or functions specific to a
customer application. Any low-priority tasks are run in a slow loop and will be
interrupted whenever the high-priority tasks occur such as the ADC interrupt to
run the control loop code.

The
development of robust and efficient firmware for a power supply can take a significant
amount of time depending on the complexity of the design. Added to this is the verification
and testing process and the documentation required for various safety
approvals. Therefore, substantial resources need to be invested into the
development of a digital power supply. However, once the initial investment has
been made, the firmware can be reused across many different products. For
example, the change in firmware for products with different output voltages
within one series may be a simple case of changing the controller coefficients.

Given
the flexibility that a microcontroller adds to the design, digital power lends
itself well to custom power applications in which standard products may not
satisfy all of the customer’s requirements. There may be specific communication
requirements such as controlling the power supply over USB, I2C, or EtherCAT
combined with the possibility of updating the communication protocol at a later
date using a live firmware update. The customer may require the output voltage
or current limits to be adjusted on the fly or require real-time monitoring,
power rail sequencing, or accurate current share between output modules.

Certainly,
the high-performance microcontroller used for digital power will be costlier
than the analog IC that it replaces. However, the digital controller opens up
the opportunity to implement other functionality within the MCU rather than
using discrete components. This can lead to a reduced component count and more
compact solutions, especially for designs with complex signal requirements or
multiple power rails, which could be controlled using one microcontroller. The
result may be an overall solution that is more cost-effective when implemented
using a microcontroller over an analog solution. Of course, for some complex
requirements, going digital may be the only solution.

XP
Power has the capability to implement complex custom power solutions utilizing
the diversity of our standard product range. With an MCU at the heart of the
power supply, the possibilities for custom power applications are far-reaching.
Given the many benefits that digital power offers, we are likely to see an
ever-increasing presence of digital power over the coming years.