Hysteretic Control: An old technology makes a comeback

One reason I like to write this blog is because it allows me to research new areas of power and analog electronics. Another reason I value the research is I have the opportunity to simplify electronics especially for the hobbyist and beginner. Recently, I found that the area of hysteretic control offered me a change to look in a new area that simplifies control loop design to a degree.

Way back before Pulse-Width Modulators [PWM] were an off-the-shelf part, control was largely performed with discrete components. A favorite control technology at the time was hysteretic control due in part to its ability to operate without a dedicated frequency and therefore clocking circuit. In those ancient times, a bang-bang self-oscillating circuit served the purpose of generating its own frequency while converting power in a switched mode fashion. This offered advantages over the losses and limitations of linear power supplies. Reference 1, Hysteretic-Mode Converters Demystified, Part 1, tells the history of how hysteretic control has again gained popularity.

As technology progressed, fixed frequency designs enabled PWM. This simplified filter design for the input as well as the inductor as it is much easier to design for a fixed frequency. However, PWM also had short comings as it requires complexity in order to compensate. Most of the references refer to a Type 3 compensation network for overcoming the double pole due to the LC filter corner frequency when using voltage mode control. Current mode control does reduce the Bode plot to a single pole compensation puzzle however there is often an additional need for slope compensation in current mode control designs.

In both current mode and voltage mode, a high bandwidth error amplifier and compensation network is beneficial. Hysteretic control in its most basic form doesn’t need an error amplifier. Instead, hysteretic mode control can use a simple comparator. Also, fixed frequency designs can cause issues with slower response speeds whereas the variable frequency of hysteretic control offers a faster response time. This is advantageous for high current transient loads.

Screen capture of output voltage: Undershoot close-up.

Screen capture of output voltage: Undershoot close-up.

One other advantage of hysteretic control is that it uses parasitics as an input signal. Because the control relies on the triangular waveform of the inductor current, control schemes exist for sensing the series resistance of the inductor or the Equivalent Series Resistance [ESR] of the output capacitors. Thus, the use of low ESR ceramic capacitors actually create a disadvantage when using hysteretic control based on the ripple current in the capacitor.

Now that the basic advantages have been introduced, it’s best to focus on the information that provides more detail than a concept introduction in a blog. The basic buck converter design is covered well in the three part series by Masashi Nogawa of Texas Instruments. Part 1 introduces the concept and gives some insight on the advantages. Part 1 also compares the phase contributions for the three modes of control. This is an excellent, basic introduction into control loop theory especially for the beginner. Part 2 compares the performance of various voltage mode, current mode, and hysteretic mode controls in a similar application having a large transient load current. Part 3 focuses on the associated Bode plots for the three modes of control. The three part TI series even compares the tradition of piling on output capacitance at the load adding a twist of reality to the explanation.

Reference 5 adds a new twist by analyzing the stability of current mode control in a hysteretic mode control application.

For topologies beyond the buck converter, Reference 4, Voltage-Mode, Current-Mode (and Hysteretic Control) Technical Note TN-203 Sanjaya Maniktala goes into detail on the plant gain Bode plots and related equations that form the poles and zeros.

Click here for larger image

Microsemi’s Voltage-Mode, Current-Mode (and Hysteretic Control) Technical Note TN-203 Sanjaya Maniktala application note also illustrates the three types of compensation networks. Graphical representation of the Bode plots for the plant as well as the control loop is the best way to understand stability without having to perform a lot of messy calculations. Also, phase calculations are much easier when one realizes that:

  • Poles have a 90 degree phase lag [downslope] starting a decade before and extending a decade after the pole frequency
  • Zeros have a 90 degree phase lead [upslope] starting a decade before and extending a decade after the pole frequency

Again, the graphical analysis for phase lag simplifies the math a great deal over computing the trigonometry and imaginary/real number calculations. With a little practice, one can become quite knowledgeable in predicting system stability from a few, simple Bode plots.

In closing, it’s always interesting when technologies come full circle. For hysteretic control, large load transient applications are offering a new way to use an old technology.


  1. Hysteretic-Mode Converters Demystified, Part 1, Masashi Nogawa, Senior Systems Engineer, Texas Instruments | May 27, 2016
  2. Hysteretic-Mode Converters Demystified, Part 2: Voltage and Current Mode Control, Masashi Nogawa, Senior Systems Engineer, Texas Instruments | Jun 23, 2016
  3. Hysteretic-Mode Converters Demystified – Part 3: Regulator Stability, Masashi Nogawa, Senior Systems Engineer, Texas Instruments | Sep 28, 2016
  4. Voltage-Mode, Current-Mode (and Hysteretic Control) Technical Note TN-203, Sanjaya Maniktala, 2012
  5. Stability Analysis & Design of Hysteretic Current mode Switched-inductor Buck DC–DC Converters, Carlos J. Solis, Graduate Student Member, IEEE, and Gabriel A. Rincón-Mora, Fellow, IEEE Georgia Institute of Technology, Atlanta, Georgia 30332 U.S.A

6 comments on “Hysteretic Control: An old technology makes a comeback

  1. D Feucht
    June 12, 2018


    I share your interest in the return of hysteretic control of converters, but I don't think loop compensation need be as complicated as Maniktala makes it out to be. I am presently designing a “Volks-inverter” for off-grid residential use and the control subsystem has a hysteretic supply. (I wish I could post it to you.) It has no loop compensation circuitry because these supplies are inherently unstable.

    However, what is nontrivial is the design for a given frequency range, and that takes some derivations resulting in design formulas that don't go well in text responses. The equations are needed to determine the inductor value, for instance. Perhaps I'll write an article with it all in the future, inspired by your article.

  2. Steve Taranovich
    June 13, 2018

    Hi Dennis,

    I would love an article on the “Volks-inverter” for off-grid residential use—we could use that on EDN. Would this be something that you would use in your remote location in the 'Boondocks” (Originally meaning a mountain)?

  3. D Feucht
    June 13, 2018


    I live on a jungle hilltop, miles from a paved road or utilities. Consequently, the inverter is of direct interest. I cannot bring myself (like my non-engineer fellow neighbor homesteaders) to buy an expensive “boat anchor” inverter with 60 Hz magnetics. The low-cost Asian imports cut corners in the design and are unreliable. A Swiss company has a good though expensive unit with too many bells and  whistles – and remote support. Some American brands are okay, but the real problem is maintenance. When it breaks, you must reverse engineer the ciruit to know what it is, to fix it. So if you want it done right …

    This open-source inverter design is well along and is unlike any other I have seen in that it optimizes the battery converter, using a boost-push-pull circuit. The input inductor is an added magnetic component but it also prevents high di/dt that destroys MOSFETs in other designs. It also has a novel overcurrent protection circuit. The prototype design is all through-hole (though need not be manufactured as such), modular, scalable, and uses commodity parts available from multiple suppliers. The only custom parts are the circuit-board (mfg files to be made available) and the magnetic  components, which are constructed of low-cost cores in a way that does not require a winding machine or any specialized equipment (other than a variable-speed drill).

    I should have something to present by this fall, I am hoping. I have been working, off and on, on this project for years and it is time to end it! I need a back-up inverter! It can be the source of multiple articles.

  4. Steve Taranovich
    June 14, 2018

    Many thanks for your insights Dennis. We look forward to hearing more about this project

  5. Effective-Technical-Writing dot com
    June 19, 2018

    Thanks D,

    I am reverse engineering a similar design that's in a “deterrent”.  The more I look into the hysteretic control, the more I realize that there isn't much written about this interesting technology.  Anything that you can contribute will surely provide more insight.  Work with our editor Steven on this contribution.  By the way I can't pull up the article while responding to see if I mentioned a bang bang converter however this is the bang bang converter technology.  


  6. D Feucht
    August 2, 2018

    “… a truly archaic technology or computing language turns out to have more to it than meets the eye”

    The language is Forth. Those of us who have been using it find it in a bigger league than C or other fixed-definition languages. And it easily fits into 16 kbytes, including interpreter and incremental compiler.

    In my previous comment, I forgot to mention that the control subsystem of the “Volksinverter” that I am designing has its own control supply which has a hysteretic controller. I did not find in the available literature any design formulas for it, so if you want it done right …

    I will include the design formulas in the forthcoming inverter articles. (Maybe Scott will beat me to it!) The interesting one calculates the frequency, and it comes out about right as measured in a prototype unit.


Leave a Reply

This site uses Akismet to reduce spam. Learn how your comment data is processed.