Connect with us

voltage sag when switching on circuit

Discussion in 'Electronic Design' started by starfire, Jun 11, 2007.

Scroll to continue with content
  1. starfire

    starfire Guest

    First a bit of the background:
    I am putting together a circuit that runs on a 3.6VDC, 19Ahr battery. The
    circuit has a LDO voltage regulator (TC1108) to output 3.3VDC for a local
    PIC microcontroller (PIC18LF2620). The PIC has an output pin that controls
    an N-Channel MOSFET (IRLML2502), which controls a P-Channel MOSFET (FDS6675)
    to switch the 3.6VDC to an external DC-DC converter. The DC-DC converter
    will be used to turn on a 12VDC device with a current draw of about 1A, max.
    This translates to about 3.7A at the 3.6VDC side of the DC-DC converter.

    Now the problem:
    I have the circuit put together and have a piece of test code in the PIC to
    exercise the on/off function for the MOSFETs. Initially there is no load
    connected to the switched output side (the DC-DC converter is disconnected).
    When I try to turn on the MOSFET switch, it temporarily tries to swtich on
    but then immediately switches back off. The PIC gets reset and
    re-initializes. I traced the problem down to the voltage from the 3.6VDC
    input side sagging (actually a series of transients going down as far as
    1.2VDC and up to 4.1VDC and lasting for about 2us) when I attempt to switch
    on. I've tried putting a 100uF or 1000uF cap on the 3.6VDC power supply
    side with no effect. Scope traces show the same response. I'm using a
    HY3003 bench power supply for the tests. The supply can output up to 3A at
    any voltage up to 30VDC. The normal circuit draw at 3.3VDC is about 110mA.
    I have breadboarded the switching circuit and verified it works correctly.
    Is there any thing that can be done about the initial droop of the voltage
    from the power supply causing this problem? Is there any way to desensitize
    the 3.3VDC circuit? If I issue a command to switch the circuit on in the
    PIC, it immediately resets. If I immediately send another command to switch
    the circuit on again, it does switch on. I can't get the circuit to switch
    on at all, though, if there is any load at all on the switched output side.

    Any insight or constructive suggestions would be appreciated. I am trying
    to avoid having to put in a separate 12VDC power supply if I can. I realize
    this would probably be a cleaner solution, though.


  2. J.A. Legris

    J.A. Legris Guest

    If there's no load on the MOSFETs, then you have to ask, where's the
    current going that makes the power supply droop? Remove all the MOSFET
    circuitry and see what happens. The problem may have nothing to do
    with them.

    And while you're at it, please post the *complete* circuit so we can
    see it.
  3. Joerg

    Joerg Guest

    Electrolytics of that capacitance most likely won't be able to do much
    about current spikes in the usec range. Sprinkle some 1uF ceramics and a
    few 0.1uF. Short leads, best is SMT. I hope you have a full ground
    plane. If not I guess it may be time for a re-layout.

    WRT the LDO I'd be careful. Don't know this type but I've seen lots of
    grief with LDOs. A switcher may be much better and many processors can
    start on a slower clock as low as 2V. At least the MSP430 can.
  4. Jamie

    Jamie Guest

    obviously you have a heavy load coming on and maybe from capacitance ?
    you could try using a series R to the gate of the Fet for one thing..
    and, maybe a diode to the Vcc side of the Pic from the supple with a
    cap there to ground. this way when a load takes place, the PIC will
    still have power from the cap and the rest of it will have to suffer
    for that moment.
  5. starfire

    starfire Guest

    Thanks for the response, Joe.

    I was thinking of posting the circuit (it's fairly small) as a .jpg but I
    didn't want to offend anyone.

    Is it acceptable to attach a small .jpg?

  6. starfire

    starfire Guest

    Thanks for the response, Jeorg.

    I chose the TC1108 (Microchip part) for the very low drop-out voltage and
    pin-out with the pads on the circuit board. Since I'm working with a 3.6VDC
    source and need a regulated 3.3VDC for the PIC and associated circuitry, I
    needed a fairly small drop-out device.

    The circuit has 1uF ceramic 0805 SMD caps on the input and output of the
    TC1108 regulator. I also have a 0.1uF ceramic cap across all components in
    the 3.3VDC circuit. The total current draw from the 3.3VDC circuit varies
    from 80mA to about 110mA. It goes to 110mA at power-up/reset.

  7. starfire

    starfire Guest

    Thanks for the response.

    Could you elaborate a little on what you mean by series resistor to the gate
    of the MOSFET? Do you mean a small (100-ohm or so) resistor from the PIC
    pin to the gate of the N-Channel control MOSFET? And the diode/cap... do
    you mean diode anode to the 3.3VDC supply, diode cathode to the PIC Vcc
    input, cap from PIC Vcc input to ground? I have a 0.1uF cap from Vcc to
    ground now.

  8. Donald

    Donald Guest

    Do you have a web site to post you .jpg ?

    There are many available for just this sort of thing.

    More information about the sag would be a good thing.

    Like a scope trace if you have it.

  9. First of all. I'd skip the regulator. The PIC should handle
    the full battery voltage. Add an RC low pass filter between
    the supply and the PIC. Something like 100 ohms and 100 uF
    should do it.

    It sounds like the regulator in the supply has a fairly low
    frequency response, so keep the 1000 uF across the supply to
    help it with the load transients. You may need even more
    for the battery. Keep in mind that not all electrolytic
    capacitors are made equal. Some have lots lower or higher
    equivalent series resistance and inductance.
    This series is pretty good for the money.
  10. Post it on alt.binaries.schematics.electronic and then post the subject
    line in this thread.
  11. Jamie

    Jamie Guest

    I mean both..
    You shouldn't drive the gate directly.. it has capacitance, how much I
    don't know how ever, It would help the PIC out in current inrush on the
    output. something like a 1k R to the gate should be fine.
    And for the VCC on the pic, I would use a diode to isolate from the
    main source to prevent back feed when supply voltage is getting dragged
    down and add a semi large by pass cap there to maintain the charge for
    that brief moment, that should take care of the PIC rebooting. 100 Uf
    should do it ..
    You'll get a little drop in voltage to the PIC using a Diode... You
    could go with a resistor and cap. like maybe a 150 ohm and a 470 uf cap.
    it depends on the current demands of the PIC.
  12. Guest

    if the system runs on 3.6 and the voltage regulator is 3.3 may be
    there is the problem you need more voltage
  13. neon


    Oct 21, 2006
    you are sagging the voltage due to inrush power a dc -dc converter is a dead short for a while and that makes the inrush current to be whatever dc inpedance it see many times more then the power required therefore making it sag and finaly try to correct. same thing with a car 12v can be <8 v until it starts. nothing is perfect.
  14. starfire

    starfire Guest

    Per a suggestion by Paul Hovnanian (I just realized I mispelled your last
    name in the binaries post, Paul... sorry:) ), I posted the applicable
    circuit section and scope trace on alt.binaries.schematics.electronics with
    the heading of "3.6VDC voltage sag on switching".

    A suggestion was made to just run the circuit on 3.6VDC, as the PIC will
    tolerate this voltage. While this is true, the circuit has some additional
    circuitry (a GPS, and RS232 interface chip, a voltage monitoring circuit,
    etc.) that I did not include in the schematic. The GPS is pretty picky
    about the operational voltage. Please understand I can't post the complete
    schematic as this is an on-going project, etc. The GPS is active at the
    time of the switching attempt, though, and is drawing about 90mA, max, at
    that time. The PIC itself and all remaining circuitry draw less than 20mA
    at 3.3VDC.

    Another suggestion was to include a small series resistor to the N-Channel
    MOSFET gate from the PIC output. Good suggestion. I will try that.
    Another response stated the votlage regulator might not have enough margin.
    The TC1108 is designed to work with a very small margin (typical around
    240mV at rated output of 300mA... less with a lesser load). A lower
    drop-out regulator would probably work better, though, and is planned for
    the next revision.

    The cap and resistor (or cap and diode) to the power supply input line is
    also a good suggestion. I will see if I can cut the trace to the Vcc pin
    going to the PIC to insert this. I'm a little concerned about this lowering
    the available Vcc to the PIC as I use the Vcc for a reference on an ADC
    input. I can always derate the value for the full-scale Vcc, though. The
    analog value is not that critical anyway.

    If anyone sees anything blatantly standing out after examining the trace and
    circuit, please post a reply.

    Thanks for all the help so far.

  15. John Fields

    John Fields Guest

  16. Joerg

    Joerg Guest

    Dave, you may have a layout issue then. Enough of those ceramic caps
    should definitely buffer such weak 2usec spikes. Might be the time to
    haul in an expert, preferably someone close by.
  17. Guest

    Had a similar PIC "self reset" problem the other week. Was switching
    100A from a battery. Because of this and from past experience the
    design had included the usual bits and bobs to stop this kind of
    reset. Still happened though.
    Cured with 0.1uF right at the PIC power pins. Problem was the tiny
    series inductance of 2" of tracking from the PIC power pins (40 pin
    DIL package)to the 'real' decoupling capacitor.
    The few nH of stray inductance was sufficient to negate the effect of
    the decoupling capacitor during the fractions of a uS supply
    disturbance at the 100A switch point..
    Still not sure whether it was due to a current loop effect, or the pin
    current at point of switching.
    Maybe a mixture of both but I certainly couldn't be arsed
    investigating further :)
  18. starfire

    starfire Guest

    Thanks for the response. I do have a 0.1uF cap right at the Power/Ground
    pins but what you've stated made me re-examine my layout. This is a
    two-sided board design (I know that's at least part of the problem... no
    internal power and ground planes) and the power and ground traces are all
    over the board. I may try to route the power for the PIC a little more
    directly from the +3.3VDC regulator or (as was suggested in another
    response) through a Schottky diode and to the PIC from the +3.6VDC side
    directly with a fairly large (10uF to 22uF) cap to ground at the power pin
    of the PIC.

  19. ehsjr

    ehsjr Guest

    Use jamie's idea.
    Dump the LDO and isolate the power like this,
    usng a Schottky diode.

    + 3.6----+-----to other stuff
    +--->|---+---To PIC

    That is step 1. Of itself, it won't keep spikes out
    of the PIC, but it will ensure that the PIC's Vcc
    doesn't sag. Right now, that is your major problem.

  20. starfire

    starfire Guest

    .... snip...

    I've posted some resultant traces on the alt.binaries.schematics.electronics
    newsgroup under the heading "3.6VDC voltage sag on switching" after
    implementing some fixes.

    Thanks much to all who responded.

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