Connect with us

Galvanically isolated AC voltage + phase?

Discussion in 'Electronic Design' started by Richard Rasker, Jul 2, 2013.

Scroll to continue with content
  1. Hello,

    I'm working on a circuit for true mains AC power measurements, i.e. taking
    cos(phi) between AC current and AC voltage into account. One of the hurdles
    is the need for a good galvanic isolation between the mains AC (230 volts,
    50 Hz) and the low-voltage measurement circuit.

    Current amplitude and phase measurement is done with the ACS712-20 device,
    and that works just fine. No problems there.
    At the moment, AC voltage amplitude and phase are determined using a 1VA 6V
    AC mains transformer, ands even though it's a small transformer and does the
    job, it's quite a bulky solution (approximately 1 cubic inch for
    commercially available types).

    So I wonder if there's a 'smarter' (as in: smaller) way to get information
    about both the amplitude and phase of the mains AC to the low-voltage part
    of the circuit, besides using a transformer.
    Capacitive coupling seems to be no option at all (phase and/or ground levels
    of the separate circuits may or may not be connected), and (forward-biased)
    opto-couplers have way too much tolerance in their current transfer ratio to
    be useful in this respect.

    Does anyone know of any small-size AC voltage measurement alternatives, as
    with this nifty Hall-effect devices?

    Thanks in advance,

    Best regards,

    Richard Rasker
  2. Harry D

    Harry D Guest

    RB, so many ways, so little time. Throw away that ACS712-20 and look at a Si8900 which has 3 10 bit input channels with an isolated output at about $3USD a pop. Input channels; AC voltage, AC current (Sense resistor and op-amp) and maybe Vdc for the third (test). The digital output goes to your favorite MPU to calculate true watts and PF. This will require an isolated supply but that can be your next question.
    Cheers, Harry
  3. This 230vac version works well with 120vac.

    The 120vac version has distortion whenused are 120vac.

  4. Phil Allison

    Phil Allison Guest

    "John Larkin"
    Richard Rasker

    ** The tranny has to be one rated for mains voltage isolation - as the
    primary can still see 240VAC.

    1VA E-core types are about the smallest readily available. 1.6VA toroidals
    are available too, but are a little larger and more expensive.

    Miniature 600 ohm line matching transformers are a candidate, but none seem
    to be specifically rated for mains isolation.

    .... Phil
  5. Guest

    Okay, great, you hacked a way to turn a voltage into a current using a resistor. Now about turning a current into a voltage...?
  6. Phil Allison

    Phil Allison Guest

    "Richard Rasker"
    ** You have one like this - right?

    The ACS712 is VERY noisy, because of which the resolution is only about

    Max current is only +/- 20 amps peak.

    Limits the apps a fair bit.

    The Hall sensors I like to use are made by LEM or Honeywell and have a hole
    you can pass a wire through or wind turns around.

    Peak currents go to +/- 100 amps and resolution to under 1 mA.

    DC to over 100kHz.

    ..... Phil
  7. Guest

    Power meters/monitors are big business these days. You're not going to beat the price and functionality of something like this:

    Leviton is usually cheap:
  8. I know, and I'm actually using a multiplier (AD633 for now, and a PIC
    controller in the future). I just mentioned cos(phi) to make it clear that I
    intend to take the U-I phase shift into account, and not simply multiply |U|
    and |I|.
    Yet another reason to look for a better solution.
    I'm working on a solution based on an HCNR201. The primary supply of a few
    milli-amps is easily built using a capacitor or even 100K/1W in resistors.

    Anyway, thanks for your input,

    Best regards,

    Richard Rasker
  9. Or multiply random sample pairs of their instantantaneous values,
    AIUI. A la Larkin.
  10. Hi Harry,

    Interesting device, and perhaps I'll use it in the future (I'm planning to
    use a controller for the multiplication function anyway). At this point, I
    first want to take the 'simple' route with analog parts.

    Thanks, best regards,

    Richard Rasker
  11. Phil Allison

    Phil Allison Guest

    "Richard Rasker"
    ** I know them well.

    The AD633 is a little winner for such jobs.

    ** Huh ??

    Phase shift becomes academic and irrelevant, once you use a multiplier.

    The ratio of watts to VA gives the PF *precisely* - no matter what the
    current wave is like.

    ** Good luck.

    .... Phil
  12. Bill Sloman

    Bill Sloman Guest

    Only if the V and the I values you are multiplying are the actual values in the relevant bits of the circuit. Any differential phase lag between the two does matter.
    If what you are multiplying is exactly what what's doing the work.

    Dual photodetector opto-couplers have been around for a while, but this one seems to be more tightly specified than usual. +/-5% tolerance on the transfer gain is nice, but still loose enough to need trimming out in a power measuring loop.

    The "1MHz bandwidth" degrades a bit if you want to retain the good linearity ...
  13. Bandwidth is not important, and for now neither is accuracy.
    For those interested, here's a circuit that works pretty well:

    It's only a few percent off, and that is easily adjusted by means of a
    trimmer in series with a slightly smaller R4. Any comments are welcome, of
    course :)

    Richard Rasker
  14. Fred Abse

    Fred Abse Guest

    That assumes sinusoidal voltages and currents. No true anymore, in these
    days of nonlinear loads.

    You need to sample voltage and current, multiply individual samples, then

    PF=cos(phi) is only true for pure sine waves.

    For current samping, look at LEM, Honeywell, and Tamura products.

    Watch out for COTS transformers used at their spec voltage rating. Many
    are designed to run close to saturation, to save iron. They distort
    quite a bit.
  15. Phil Allison

    Phil Allison Guest

    "John Larkin"
    "Phil Allison"
    ** The OP has the wrong idea about phase angle - he thinks it always

    ** PF = watts / VA by definition.

    IF a phase angle exists, it can be quoted.

    There is no debate about that.

    .... Phil
  16. Phil Allison

    Phil Allison Guest

    "Bill Slowman Wanker "

    ** Yawnnnnnnnnnnnnnnn......

    Fucking pedant.
  17. Nico Coesel

    Nico Coesel Guest

    If you create an artificial ground using 2 Y capacitors (like the ones
    used in every mains adapter) for your circuit you can use 2 safety
    rated resistors and a differential amplifier to sense the voltage. In
    most cases the seconday circuit is either floating or somehow
    connected to ground (earth). Some means of overvoltage protection will
    probably be required to protect the resistors (common mode filter +
  18. Fred Abse

    Fred Abse Guest

    The regulatory authorities don't.
  19. Fred Abse

    Fred Abse Guest

    When "on", the PF is unity, when "off",power is zero, so PF doesn't exist.
    Same as switching (tungsten) lights on and off.

    Average true power over 24 hours = 0.5
    Average apparent power over 24 hours = 0.5
    Hence power factor =1

    Makes no difference whether you integrate over one cycle, or 12*3600*60
    cycles. It's unity PF.

    Whether a utility would charge for below unity would depend on the
    ratio of MSEEs to MBAs.
  20. whit3rd

    whit3rd Guest

    If one declares PF (power factor) to be the cosine of a phase
    angle, that's true.
    The square-sum average of a function of time can be defined

    ||F|| == integral_time( |F|**2 dT)

    and there's an important combination of two functions called
    the correlation

    Corr (F1, F2) == integral_time (cconj(F1(t)) * F2(t) dt) )/sqrt (||F1|| * ||F2||)

    which is always in the range {-1, +1}

    It's that correlation, of I(t) and V(t), that is the power factor. If one has only simple
    real sinewave functions, the arccosine of that correlation is an angle, which
    can be identified with a phase shift betwixt F1 and F2.

    So, mathematically, one can define all this stuff, AS LONG AS YOU AGREE ON A TIME
    INTERVAL to do integration. If your signals are non periodic, there's no natural
    reason to agree (and mathematicians will complain bitterly about the scientists'
    habit of choosing a time interval as 'sufficiently long').
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