Connect with us

Isolating PS Capacitance

Discussion in 'Electronic Design' started by Mark, Apr 13, 2004.

Scroll to continue with content
  1. Mark

    Mark Guest

    I'm working on a VME chassis with multiple cards. In the chassis is a
    power supply card that distributes 5 volts to the other cards (digital
    power for logic devices). The power supply is a DC/DC converter and has
    a maximum capacitive load it can operate into. The other cards in the
    chassis were designed using heavy filtering and thus we are exceeding
    the maximum capacitance allowed on the output of the DC/DC converter.

    We are adding another supply to deal with the problem, but this led me
    to consider how small RC filters would look to the DC/DC converter.

    I have seen and used an RC bypass circuit instead of just a bypass
    capacitor for critical devices where I wanted to assure low noise
    operation. I've used a 10 ohm resistor is series with a 0.1 or 0.01 uf
    cap to ground and have had good results with this type of circuit in low
    or medium frequency operation.

    My question:

    If you have LOTS of bypass capacitors on the board, and used the RC
    circuit instead of just capacitors, would this tend to isolate the large
    value of capacitance the the DC/DC converter "sees"?

    If so, is there a way to estimate or calculate this?

    Thanks,

    Mark
     
  2. legg

    legg Guest

    It depends on why the maximum C value was imposed on the converter by
    the mfr. It could have something to do with start-up and overload
    recovery, rather than dynamic stability. You should really ask the
    manufacturer.

    Your description of 'heavily decoupled' doesn't jive with the
    extensive use of 0.1 uF parts. Upper limits on C are usually in the
    hundreds or thousands of microfarads. What symptoms have actually
    shown up and what decoupling is actually used?

    Adding supplies in parallel, without partitioning, isn't a guaranteed
    solution to either effect. Repacing with a more robust device is a
    better bet.

    Some mfrs have taken these problems seriously, when they were made
    aware of them and offer solutions, including impedance definitions,
    sharing circuitry and start-recovery synchronization methods.

    You can test bus stability by loading it transiently and monitoring
    the bus voltage response over the operating temperature range.

    RL
     
  3. Joerg

    Joerg Guest

    Mark, it is not a good sign if a power supply has problems with large
    capacitive loads. I would get a better one. I have worked on ultrasound
    machines that contained a dozen boards literally peppered with tantalums but
    the power supplies never had a problem with that.

    Regards, Joerg.
     
  4. Mark

    Mark Guest

    It depends on why the maximum C value was imposed on the converter by
    the mfr. It could have something to do with start-up and overload
    recovery, rather than dynamic stability. You should really ask the
    manufacturer.

    Your description of 'heavily decoupled' doesn't jive with the
    extensive use of 0.1 uF parts. Upper limits on C are usually in the
    hundreds or thousands of microfarads. What symptoms have actually
    shown up and what decoupling is actually used?

    Adding supplies in parallel, without partitioning, isn't a guaranteed
    solution to either effect. Repacing with a more robust device is a
    better bet.

    Some mfrs have taken these problems seriously, when they were made
    aware of them and offer solutions, including impedance definitions,
    sharing circuitry and start-recovery synchronization methods.

    You can test bus stability by loading it transiently and monitoring
    the bus voltage response over the operating temperature range.

    RL[/QUOTE]

    There are 5V to 2.5 volt converters on 3 other cards. All three
    converters have two 47uf input capacitors on them for a total of 282uf,
    thers are also some 4.7uf and 10uf tantalums spread around.

    Supplies were not added in parallel, system was partitioned.

    Mark
     
  5. R.Legg

    R.Legg Guest

     
  6. Ban

    Ban Guest

    There are 5V to 2.5 volt converters on 3 other cards. All three
    converters have two 47uf input capacitors on them for a total of
    282uf, thers are also some 4.7uf and 10uf tantalums spread around.

    Supplies were not added in parallel, system was partitioned.

    Mark[/QUOTE]

    Mark,
    I'm pretty sure your power supply is broken or too weak. The inductance of
    the traces leading to the other cards should be sufficient to isolate the
    input capacities. What happens with the PS, does it oscillate or is there
    too slow a reaction when the loads are switching in or out? Very often
    converters have a big inrush current, many times higher than the normal
    operating current, it might upset your PS and force it into shutdown.
     
  7. Mark

    Mark Guest

    We have inrush current limiting. The manufacture lists 300 uf as the
    maximum load cpapcitance for the DC/DC converter. With the dual 47uf
    caps in front of the three 5v to 2.5 volt converters makes a total of
    282uf. There are also many 0.1uf, and a handfull of 4.7uf and 10uf caps
    spread through the three boards that the power supply board is
    supplying. So we are indeed exceeding the the manufactures suggestion
    of maximum load capacatance.

    Under the heavy capacitance load, the converter would somtimes not
    start, if it did start though, it worked fine and did not oscillate.

    This is a "prototype" and we have partitioned the power system and
    added another converter so both converters are now operating within the
    manufactures load capacitance specifications.

    Lesson! Need a "systems engineering" approach. We have 4 engineers each
    designing their respective cards. We did a good job tracking I/O and
    every thing worked whne the boards were plugged into the WME chassis and
    if the power supply turned on.

    However, we did not keep track of each engineers zeal by heavily bypass
    their designs.

    Thanks for the help.

    Mark
     
  8. Mark

    Mark Guest

     
  9. Fred Bloggs

    Fred Bloggs Guest

    The principle of "systems engineering" most applicable to this situation
    is that of basic risk reduction by allocating tasking to qualified
    personnel- individuals possessing a skill set which includes the ability
    to comprehend a data sheet among other things.
     
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

-