Connect with us

Switching power supply question

Discussion in 'Electronic Design' started by [email protected], Feb 23, 2008.

Scroll to continue with content
  1. Guest

    A swithing supply is a feedback control system. Therefore why doesn't
    it oscillate when you put a capacitive load on it? It should bugger up
    the phase margin.

  2. John Larkin

    John Larkin Guest

    Probably because some clever person designed the loop so that it
    becomes more stable as you add capacitive loads.

  3. If the dominate pole for the control loop is made up by the
    impedance of the supply feeding the output capacitor, then
    adding more capacitance just shifts the 3db corner frequency
    down. Then there are hysteresis controls that just bounce
    the output voltage between two slightly different voltages.
    These are unstable, by design, but the oscillation
    amplitude is controlled. Adding capacitance just slows the
    oscillation cycle.

    But there are lots of other loop designs that are
    destabilized by adding enough pure capacitance across the
    output. Many need a little resistance in series with a
    significant part of the total capacitance to add a
    stabilizing zero to the total response.
  4. Joerg

    Joerg Guest

    Does that mean there can't be many clever persons in groups tasked with
    LDO design? SCNR ...
  5. legg

    legg Guest

    Actually, it will do so, unless you intend otherwise.

    This is one of the basic tenets of electronic design.

  6. John Larkin

    John Larkin Guest

    Apparently not. A couple of people make opamps that tolerate any
    capacitive load. The concept is pretty simple: the output pin is one
    end of the compensation cap; adding load capacitance doesn't add a
    pole to the loop, it just shifts the dominant pole down. The guys who
    do LDO's ought to be told about this.

  7. Jim Thompson

    Jim Thompson Guest

    Not quite as simple as that. There's some feed-around techniques used
    to isolate the load.
    They can't, without an extra pin, and an external R/C... LDO's can't
    afford to waste power with implementing the R on-chip... a small R
    forces a large un-integrable capacitor.

    I recently did one on-chip, but the current ability was only 15mA,
    charging an external capacitor, which was then dumped with 2.2A.

    ...Jim Thompson
  8. John Larkin

    John Larkin Guest

    This is what I had in mind, per Fig 1.

    The load isn't isolated, it becomes part of the compensation. The
    transistors operate in "LDO" mode, namely with the output coming off
    the collectors, not the emitters.

    The 8261 is a very handy gadget to have around.

  9. Jim Thompson

    Jim Thompson Guest

    It's a "simplified" schematic. Add up the poles if it were
    implemented purely as shown, and you'll see why a zero is required
    somewhere... for capacitive loads it needs to be associated with the
    output configuration.

    BTDT (almost daily :)

    ...Jim Thompson
  10. Marra

    Marra Guest

    It depends on the feedback function.

    You get the same with motor speed control.
    Thats why someone had to use PID to control a motor.
  11. Robert

    Robert Guest

    So LDO Designers aren't as stupid as might be assumed?

    Robert H.
  12. Tim Wescott

    Tim Wescott Guest

    Cruise control is a feedback control system, yet I think I could load up
    my wife's station wagon with boxes and boxes of capacitors and have the
    cruise control work fine.

    Just because many (but not all!) op-amp circuits go unstable when they're
    capacitively loaded doesn't mean that all electronic feedback circuits
    will go unstable when capacitively loaded.

    Tim Wescott
    Control systems and communications consulting

    Need to learn how to apply control theory in your embedded system?
    "Applied Control Theory for Embedded Systems" by Tim Wescott
  13. Joerg

    Joerg Guest

    But it's a Dollar a pop. Yikes.
  14. Joerg

    Joerg Guest

    That the point. We discrete designers do not mind at all to have to add
    a cap or even a more elaborate external loop. We are used to that from
    PWM controllers. It can't be too expensive to go from a SOT23 to a SOT23-5.

  15. MooseFET

    MooseFET Guest

    It depends on how stupid you assume they are. There are also some
    mistakes that it takes very smart people to make.

    In the past I have made some "do it your self" LDOs. It is very easy
    to make them stable with any load capacitance you you accept a crappy
    load regulation.
  16. MooseFET

    MooseFET Guest

    Let me be the first to say "use a PIC" in this thread. ADC->MICRO-
    chips have enough bits, people may not notice the chatter.
  17. Joerg

    Joerg Guest

    First post must have been picked up by a wind gust (it's ugly out here
    right now).

    We discrete designers would not mind an extra pin one bit. We are used
    to external parts from PWM chips. What's the big deal?

  18. Guest

    Yes but with a motor the type of load is normally well defined.(a
    rotating mass of some sort) and is designed for a specific application
    in mind.

    Not so with power supplies where you can be connecting up to almost
    I have seen a power supply osicllating once.

  19. Guest

    I never said it would - you can read my earlier comments. You
    obviously understand little aboyt control engineering.
    (If it's in a book it's either elementary,irrelevant or wrong!)

  20. Some of them do oscillate if you put a large, low ESR capactive load
    on the output for exactly the reason you give.

    DC/DC converter modules suffer from this because they have some fixed,
    compromise compensation that's usually designed for a small capacitive
    load. They ring like a bell when you put lots of low ESR C on the
    If you're designing the DC/DC converter you can compensate it properly
    for any load, but this doesn't normally apply to the prebuilt modules.
    I first noticed this effect when I started using OS-CON and POSCAP
    capacitors in my designs. These caps have lots of capacitance with
    low ESR.

    A few years ago TI / POLA released a series of DC/DC converter modules
    which have programmable compensation (called TurboTrans), which allows
    you to tune the thing to have good performance for a particular
    (large) C load. I haven't used one yet, but they seem promising.

    It's described in this PDF:

Ask a Question
Want to reply to this thread or ask your own question?
You'll need to choose a username for the site, which only take a couple of moments (here). After that, you can post your question and our members will help you out.
Electronics Point Logo
Continue to site
Quote of the day