Connect with us

Metal Halide Lamps

Discussion in 'Lighting' started by Victor Roberts, May 14, 2007.

Scroll to continue with content
  1. I've had some tests run on metal halide lamps by a certified
    independent laboratory (since I don't have a sphere.) The
    main goal of this work was to quantify the high frequency
    efficacy of metal halide lamps. The common wisdom is that
    there is no high frequency efficacy gain, but there are one
    or two papers that claim otherwise. This is my poster paper
    for LS:11.

    The results are in and they are a bit strange.

    The tests were run on three 320-watt pulse start metal
    halide lamps. there is more money from the sponsoring
    organization to run 6 more 320-watt lamps if necessary.

    As part of the test I measured lamp performance on five
    types ballasts:

    1) Standard reference ballast
    2) Commercial linear inductor ballast
    3) Commercial CWA ballast
    4) Brand A 100 kHz electronic ballast
    5) Brand B 100 KHz electronic ballast.

    There were three of each type of commercial ballast.

    Since I suspected that the commercial ballasts would not all
    run the lamp at the same power, and since lamp efficacy is a
    function of power, I had the lab operate the lamp on the
    reference ballast at rated power, rated power + 15% and
    rated power -15%, so I could construct a correction curve.

    The really strange result is that lamp efficacy on both
    types of commercial ballasts was lower than on the reference
    ballast, even after correcting for the fact that the
    commercial ballasts ran the lamp below rated power. The
    corrected efficacy loss is about 6% to 7%. The only
    explanation I have right now is that these lamps have lower
    efficacy when operated with a high current crest factor
    system. I don't see any other obvious difference between
    lamp operation on the reference ballast and on the
    commercial ballasts that operate the lamps at the same

    Any other suggestions for a cause?

    Vic Roberts
    To reply via e-mail:
    replace xxx with vdr in the Reply to: address
    or use e-mail address listed at the Web site.

    This information is provided for educational purposes only.
    It may not be used in any publication or posted on any Web
    site without written permission.
  2. JB

    JB Guest

    The single biggest improvement I've seen when operating MH lamps at high
    frequencies, and especially ceramic metal halides, is lamp to lamp colour
    variation reduction. With some ballasts I have here we are talking _single_
    digit °K colour temperature variance between a group of 10 lamps when
    operated on 500kHz+ pure sinewave ballasts. Even the engineers at GE's
    Leicester labs here in the UK were amazed at this effect. Also the arc
    itself becomes more 'constricted' (very visible if you project the arc image
    from a quartz burner through a large projection lens onto a wall) and also
    far more stable at these frequencies.

    Another most unusual effect again not predicted, was an overall general
    reduction in CCT of ceramic lamps with increasing frequency (above ~325kHz).
    Typical 3000°K CMH/CDM lamps would have a CCT nearer to 2850°K operating at
    exactly the same lamp power as on normal electromagnetic ballasts or even
    sub-200Hz squarewave electronic ballasts.

    I put many tens of thousands of burning hours on various 20-400W MH and
    70-400W SON lamps operating from a variety of high frequency sinewave
    ballasts over the course of about 8 years, partly as part of a ballast
    development program but also out of curiosity/interest. This is still a
    very, very interesting area of research for me. I'm quite lucky with the
    resources in my lab here in the UK too, as I've just recently finished
    building a 3M diameter sphere with both a spectrometer and luminous flux
    measurement system installed.

    Let me know if you'd like a few ballast samples to play with. I have some
    70W and 150W in dual voltage 120/240V versions here.


  3. TKM

    TKM Guest

    From what I recall, Vic, your crest factor idea is correct. I would go
    further, however, and suggest that the shape of the lamp current wave is the
    controlling factor, not just the crest factor. I'm not a lamp scientist, of
    course, but I do remember that Gil Reiling (and several other MH lamp
    engineers) spent substantial time looking at tuning the lamp current
    waveforms and then designing the ballasts to match those requirements.
    Optimum waveforms changed with the metal halide mixture too.

    It made sense to me because we were seeing something similar with
    fluorescent lamps -- tune the timing and shape of the starting current and
    voltage waveforms (what's now called the "starting scenario") and you end up
    with a system where starting doesn't affect lamp life.

    I've often thought that electronic ballasts could well give us the way to
    optimize waveforms so as to improve several aspects of lamp performance and
    that we should build an "optimize" control into the ballast to better match
    the lamp and the ballast being used just as we now build in automatic
    voltage sensing.

    It's good to hear that you're looking a MH lamp efficacy as a function of
    frequency. That's certainly been a long-debated question.

    Terry McGowan
  4. I would blame/credit current crest factor and not yet expect lower
    current crest factor to help overall luminous efficacy in general.

    In a given metal halide lamp with a given vapor composition (from a
    given temperature of cold spot of inner surface of arc tube), lower
    instantaneous current favors radiation from elements whose radiation
    varies less with electron temperature (such as sodium) and higher
    instantaneous current favors radiation from elements whose radiation
    increases more drastically with electron temperature (such as mercury,
    among MH lamp ingredients).

    If the value for RMS current is decreased for a given power input, then
    sodium (or thallium) radiation could get favored more over mercury
    radiation. If the percentage of each half-cycle of the power line
    frequency that has significant lamp current conduction increases, then
    average and RMS current as determined over the significantly conducting
    portion of each half-cycle of power line frequency decreases, then sodium
    (or thallium) radiation could get more favored over that of mercury.

    Furthermore, lack of cool moments in the arc twice per cycle of power
    line frequency could increase the arc tube temperature, and increase the
    concentration of vapors of the halides.

    So if more sodium (or thallium) radiation and less mercury radiation
    means more efficacy, I suspect making the power input to the arc more
    constant throughout each half-cycle of the power line frequency can
    achieve this.
    I would watch for any higher arc tube temperature and any ill effects
    I would also look for a way to obtain spectral power distribution
    results, and look for shifting towards visible emission features favored
    by higher arc tube temperature, and away from any favored by higher
    electron temperature, as a result of changing to the electronic ballast.

    Now for another matter: High frequency electronic ballasts in
    comparison to line frequency iron core ones are known to increase the
    efficacy of fluorescent lamps via the following mechanisms:

    1. Avoids some "saturation" (or approaching such) of shortwave UV
    production of a low pressure mercury discharge, that results from peak
    power input (whether absolutely or long enough to be a major fraction of
    "imprisonment time") approaching the ability of the discharge to radiate
    such an amount of power from a given discharge surface area, available
    bandwidth of the desired spectral feature, and available from the electron
    temperature (which for that matter usually decreases when power input to
    fluorescent lamps increases IIUC).

    2. Avoids the "oscillatory anode fall" (approx. 2.5 volts
    largely not useful-radiation-producing voltage drop) that most fluorescent
    lamps have when fed either DC or AC of frequency of frequency around or
    less than 1 KHz. I am aware of some fluorescent lamps having an electrode
    structure that impairs this loss in DC and low frequency use, but for now
    I forget about any examples other than US Patent 4,902,933.

    This makes me wonder if MH lamps or the ones in question have any
    electrode-related voltage drops that are nonlinear (in a way where losses
    can be reduced by lower current crest factor) or time dependent/
    oscillatory in a way that can be mitigated by use of high frequency AC.

    - Don Klipstein ()
  5. Well, I don't understand how, as a general rule, increasing
    the crest factor can reduce efficacy while reducing it will
    not increase efficacy. Why should a CCF of 1.414 always be
    I always though that the anode fall was about 5 volts. When
    an F-lamp that operates at 100 volts is switched to FULL
    MODULATED high frequency, the efficacy gain is about 4.5%,
    representing the reduction in anode fall (since the electron
    temperature is still fully modulated.)
    One of my own :)
    More work is needed to resolve this issue.

    Vic Roberts
    To reply via e-mail:
    replace xxx with vdr in the Reply to: address
    or use e-mail address listed at the Web site.

    This information is provided for educational purposes only.
    It may not be used in any publication or posted on any Web
    site without written permission.
  6. I think my brain was running out of gas for the evening!

    I now think that these particular lamps do appear to have efficacy
    varying inversely with crest factor, and I doubt I meant to say that
    improvement in efficacy only increases until crest factor drops to some
    particular value.

    I do suspect that some MH lamps may not have efficacy improved by lower
    current crest factor.

    - Don Klipstein ()

  7. Since I was not present I can't be sure of anything :), but
    the lab is NAVLAP certified for metal halide lamp
    measurements and the power meter they used has an accuracy
    of about 1% at 100 kHz. The accuracy is specified in each
    frequency band, it is not a 3db bandwidth. The lamp was
    also connected to the power meter via a 4-wire Kelvin
    connection, which is the right way to make the connection.

    If Phase 2 is funded I plan to be at the lab when the
    measurements are being taken. I visited the lab on a
    previous occasion when they were making measurements for
    another project I was working on, and was impressed with
    their facilities and staff.

    Vic Roberts
    To reply via e-mail:
    replace xxx with vdr in the Reply to: address
    or use e-mail address listed at the Web site.

    This information is provided for educational purposes only.
    It may not be used in any publication or posted on any Web
    site without written permission.
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