Design Article
Getting the most out of power conversion stages
Shane Callanan, Excelsys
8/7/2012 10:40 AM EDT
Predictive dead time control
Predictive dead time control attempts to compensate for any uncontrolled component variations such as propagation delays, temperature effects, load variations, etc. The predictive approach uses feedback controlled gate timing to allow the circuit to "learn" the power stage components. It will do this over any given combination of line voltage, load, operating temperature, FET characteristics or any dynamic part of the circuit. As a result, body diode conduction and reverse recovery losses are virtually eliminated.
A PWM signal is fed into control blocks. The control blocks use power stage voltages to adjust gate timing delays from PWM, with the goal being to actively control the turn off the freewheel. We see the reduction in body diode conduction as shown in figure 5. The optimum turn on can be achieved.
The turn-off operates in a similar fashion to turn-on only this time a comparator must be used instead of a NOR gate. High di/dt's causes ringing on the drain to source of the freewheel FET as a result current commuting from the forward synchronous FET to the freewheel. A comparator must be used to prevent false triggering, which could possibly result in shoot-through.
The algorithm works as follows…
Vds and Vgs are fed into a NOR gate. The output from the NOR gate is latched, inverted and used to increment or ecrement a delay counter accordingly. To adjust the delay, a reference can be used to detect when both are simultaneously low. A high from the NOR gate indicates the delay was too long and this signal is then used to decrement the delay counter. This means that for the next cycle the turn off time is decreased. Any other combination indicates that the turn off time is too short and the counter is incremented, thus the turn off time is increased for the next cycle. The feedback loop will continuously adjust until turn-on/turn-off delay minimized.
Adaptive control-driven approach
The Collins English dictionary defines the word 'adapt' as 'to adjust to different conditions'. So, intuitively, an adaptive controller is one that modifies its output after changes in the control loop occur. Research in adaptive control started in the early 1950s. Control Engineers tried to agree on a formal definition of Adaptive Control, and some vocabulary used over the years has included 'self-organizing control (SOC) system' and 'parameter adaptive SOC'. However, the most pragmatic definition comes from Astrom & Wittenmark in their book on adaptive control and is defined as follows. As adaptive controller is a controller with adjustable parameters and a mechanism for adjusting the parameters. We see how we can draw this representation in figure 8. We show a control system, and a parameter system, which varies the controller, based on a series of input signals.
In a similar manner to the Predictive algorithm, a counter is again manipulated to determine minimum dead-time. The difference, however, is how the counter is manipulated. A high frequency clock starts to increment a counter on the rising edge of the duty cycle voltage-mode control. This continues until the falling edge of this signal occurs. This then allows the controller to determine the timing for the switching of the upper & lower MOSTFETs. A fixed value x is subtracted from this counter and this value is predetermined by design of the controller to prevent shoot-through in an application. This value will change from cycle to cycle as the controller 'adapts' to its existing operating condition. Regardless of the algorithm invoked. Reducing dead time means losses are reduced. Each method tries to eliminate body diode conduction.
The benefit of using an intelligent based algorithm means that the controller can compensate for uncontrolled component variations such as propagation delays, temperature effects and load variations.
Both approaches attempt to "learn" the power stage components and reduce the dead time accordingly.
Conclusion
Precise control-driven technique can eliminate body diode conduction in synchronous rectifiers. Intelligent control driven techniques attempt to squeeze out maximum efficiency. They can achieve this by 'learning' the power stage components, and adjust its control of turn-on and turn-off of the forward and freewheel FETs accordingly. The benefits in achieving this are not only an increase in efficiency but also component and temperature variations are compensated for, which results in minimal losses over converter lifespan. The design techniques as described above are regularly implemented in the Excelsys product range in order to achieve the highest power densities in our designs.
About the author:
Shane Callanan joined the Excelsys Technologies team in early 2006, and currently holds the role of director of applications engineering. He can be reached at shanecallanan@excelsys.com
Visit Excelsys
Courtesy of EETimes Europe
Predictive dead time control attempts to compensate for any uncontrolled component variations such as propagation delays, temperature effects, load variations, etc. The predictive approach uses feedback controlled gate timing to allow the circuit to "learn" the power stage components. It will do this over any given combination of line voltage, load, operating temperature, FET characteristics or any dynamic part of the circuit. As a result, body diode conduction and reverse recovery losses are virtually eliminated.
A PWM signal is fed into control blocks. The control blocks use power stage voltages to adjust gate timing delays from PWM, with the goal being to actively control the turn off the freewheel. We see the reduction in body diode conduction as shown in figure 5. The optimum turn on can be achieved.
Figure 5: The NOR gate is used to determine min dead time during turn on
The turn-off operates in a similar fashion to turn-on only this time a comparator must be used instead of a NOR gate. High di/dt's causes ringing on the drain to source of the freewheel FET as a result current commuting from the forward synchronous FET to the freewheel. A comparator must be used to prevent false triggering, which could possibly result in shoot-through.
The algorithm works as follows…
Vds and Vgs are fed into a NOR gate. The output from the NOR gate is latched, inverted and used to increment or ecrement a delay counter accordingly. To adjust the delay, a reference can be used to detect when both are simultaneously low. A high from the NOR gate indicates the delay was too long and this signal is then used to decrement the delay counter. This means that for the next cycle the turn off time is decreased. Any other combination indicates that the turn off time is too short and the counter is incremented, thus the turn off time is increased for the next cycle. The feedback loop will continuously adjust until turn-on/turn-off delay minimized.
Figure 6: Comparator is used for min dead-time during turn off
Figure 7: Output from NOR gate is used to increment or decrement a delay counter
Figure 7: Output from NOR gate is used to increment or decrement a delay counter
Adaptive control-driven approach
The Collins English dictionary defines the word 'adapt' as 'to adjust to different conditions'. So, intuitively, an adaptive controller is one that modifies its output after changes in the control loop occur. Research in adaptive control started in the early 1950s. Control Engineers tried to agree on a formal definition of Adaptive Control, and some vocabulary used over the years has included 'self-organizing control (SOC) system' and 'parameter adaptive SOC'. However, the most pragmatic definition comes from Astrom & Wittenmark in their book on adaptive control and is defined as follows. As adaptive controller is a controller with adjustable parameters and a mechanism for adjusting the parameters. We see how we can draw this representation in figure 8. We show a control system, and a parameter system, which varies the controller, based on a series of input signals.
Figure 8: Controller is adapting cycle by cycle as defined by parameter adjust
In a similar manner to the Predictive algorithm, a counter is again manipulated to determine minimum dead-time. The difference, however, is how the counter is manipulated. A high frequency clock starts to increment a counter on the rising edge of the duty cycle voltage-mode control. This continues until the falling edge of this signal occurs. This then allows the controller to determine the timing for the switching of the upper & lower MOSTFETs. A fixed value x is subtracted from this counter and this value is predetermined by design of the controller to prevent shoot-through in an application. This value will change from cycle to cycle as the controller 'adapts' to its existing operating condition. Regardless of the algorithm invoked. Reducing dead time means losses are reduced. Each method tries to eliminate body diode conduction.
The benefit of using an intelligent based algorithm means that the controller can compensate for uncontrolled component variations such as propagation delays, temperature effects and load variations.
Both approaches attempt to "learn" the power stage components and reduce the dead time accordingly.
Figure 9: Flowchart for an adaptive algorithm
Conclusion
Precise control-driven technique can eliminate body diode conduction in synchronous rectifiers. Intelligent control driven techniques attempt to squeeze out maximum efficiency. They can achieve this by 'learning' the power stage components, and adjust its control of turn-on and turn-off of the forward and freewheel FETs accordingly. The benefits in achieving this are not only an increase in efficiency but also component and temperature variations are compensated for, which results in minimal losses over converter lifespan. The design techniques as described above are regularly implemented in the Excelsys product range in order to achieve the highest power densities in our designs.
About the author:
Shane Callanan joined the Excelsys Technologies team in early 2006, and currently holds the role of director of applications engineering. He can be reached at shanecallanan@excelsys.com
Visit Excelsys
Courtesy of EETimes Europe
|
|
|
|
|
Navigate to related information
shawnmk
8/10/2012 6:47 PM EDT
I was asked the question: what are some of the losses because of switching elements in a DC-DC converter? My interviewer helped me through to getting the answer and now, I see this article on the same topic. It is a good article. Thank you,
Sign in to Reply