Connect with us

SMPS feedback and zener distribution curves

Discussion in 'Electronic Design' started by Yzordderex, Aug 21, 2003.

Scroll to continue with content
  1. Yzordderex

    Yzordderex Guest

    I'm designing a 27v 0.75a power supply. I want to provide the best
    regulation short of using the well known TL-431/opto feedback combo.

    I'm going to use a string of zeners in series to do the regulation.
    I'll put the feedback winding on top of the 27v winding. I've already
    built and tested this unit and the load regulation is about 1% from
    25% to 100% load.

    Ok, Question is, what can I expect for the initial voltage
    distribution of say a 5.1v zener? First of all I should ask if anyone
    knows of a manufacture who offers 1% zeners. Many years ago I used a
    very nice Motorola part that was through-hole mount. I'm having probs
    finding a smd device.

    Will a 5% part have a natural bell curve distribution? Will it be
    truncated because of sorting? Will it have a hole at 5.1v because
    those die are used in our 1% part which we don't offer here in USA?
    Will some batches come in on high side and some on low side? Perhaps
    I should use a mix of 4.7v devices and 5.3v devices. Probability less
    that both parts come in at one extream.

    thanks
    Bob
     
  2. Why? This seems a pedestrian way of doing things. Use a proper reference
    and save yourself a lot of agro.


    Kevin Aylward

    http://www.anasoft.co.uk
    SuperSpice, a very affordable Mixed-Mode
    Windows Simulator with Schematic Capture,
    Waveform Display, FFT's and Filter Design.
     
  3. Tim Shoppa

    Tim Shoppa Guest

    Why? 1% adjustable shunt regulators cost a few dimes.
    I know that 5V Zeners are the sweet spot for minimal temperature
    coefficients, but why not one 27V Zener if that's what you actually need?

    27V? Hmm... 13.6*2=27.2... Lead acid charger?

    Tim.
     
  4. R.Legg

    R.Legg Guest

    I should use a mix of 4.7v devices and 5.3v devices. Probability less
    It doesn't matter how many zeners make up the string, or how many
    different values you use, you'll still have to account for the
    labelled tolerance. This occurs at a specified current - vary from
    this and your set-point tolerance gets worse.

    Batches with truncated or dual bell curves that exclude central
    regions are common from any supplier offering A (10%) B (5%) and
    selected tolerance values.

    If you don't want to use the 431 as an active regulator, simply wire
    it up as a zener of the appropriate voltage, and use it as a zener in
    your circuit.

    RL
     
  5. Yzordderex

    Yzordderex Guest

    Yea, your both probably right. The environment I work in is very much
    cost driven. Last two places I worked they threw money around like a
    drunkin sailor. They're both on the skids and laying off. We've
    actually increased the size of our workforce and sales have been
    better than flat.

    In any case supply only has to be close to +-5%. I acheived near 1%
    load regulation from 5w to 20w. I think is pretty amazing considering
    tertiary winding feedback. I used a PQ20-20 coreset and ran it at
    130kHz using a topswitch. PQ20-20 has a big cross section, so few
    turns at 130kHz. Topswitch has a 1 amp 800v mosfet in series with it
    to allow me to run off of 1000v bus.

    I've only designed a half-dozen of these flybacks, but have great fun
    doin it.


    regards,
    Bob
     
  6. Tim Shoppa

    Tim Shoppa Guest

    In order of decreasing economy I see your choices as:

    Most economical: A single 27V Zener. (Cost about a dime in quantity,
    most minimal PCB real estate use)

    Second most economical: A TL431 with two resistors used as a shunt regulator.
    (Cost about 20 cents in quantity, middle PCB real estate use)

    Least economical: A string of 5.1V-ish Zeners. (Cost about 25 cents in
    quantity, most PCB real estate use.).

    Tim.
     
  7. Tim Shoppa

    Tim Shoppa Guest

    In order of decreasing economy I see your choices as:

    Most economical: A single 27V Zener. (Cost about a dime in quantity,
    most minimal PCB real estate use)

    Second most economical: A TL431 with two resistors used as a shunt regulator.
    (Cost about 20 cents in quantity, middle PCB real estate use)

    Least economical: A string of 5.1V-ish Zeners. (Cost about 25 cents in
    quantity, most PCB real estate use.).

    Tim.
     
  8. Yes, I'm aware that theoretically it doesn't matter how many zeners
    make up the string - still going to be 5% tolerance. However (as
    pointed out previously in the thread) the 5.1v zener has the best
    temperature coefficient of the whole bunch. It's almost zero. The
    reason I may choose to use 2 different values near 5.1v (say 4.7 and
    5.3 for purpose of explaination) is that I would be playing the
    probability game. If the reel of 5.3v zeners come in at the absolute
    worse case of +5% and the 4.7v zeners are close, then I'm better off
    than if all zeners on same reel high. In any case power supply
    doesn't have to be very accurate.

    I also realize that It is possible to configure the TL431 part to
    mimic a 27v zener. There may be some other issues tho. There are a
    pair of resistors (one from I think they call it cathode to reference
    pin, and other from ref pin to ground(when wired in a standard
    config)) that will need to be biased, and the TL-431 will also have
    some minimum current. I have to make sure that this total current is
    less than bias current required for Topswitch. I think I got that
    right:)

    I think we pay abt 20c for the Tl-431. Not sure what we pay for a
    zener, but I thought around 3c each. I will double check price on
    zeners for sure on Monday.

    I think Tim's solution of one zener is best bet, if that's all I need
    why over engineer. Sometimes I can't see the forest through the
    trees.

    Reminds me of when I was a young engineer. Production had a problem
    that drives would blow up when doing short circuit test on output. I
    knew layout wasn't so good and provided a lot of stray inductance.
    When short circuit current fell, Darlingtons blew on overvoltage due
    to stray V=Ldi/dt. I stuck a big oil filled cap directly across the
    inverter bus in order to 'investigate' the problem. Old and Grey
    production manager came by and I explained what I was doing. He
    chuckled and said "I think you've already found the solution."

    Regards and thanks,
    Bob




    (R.Legg) wrote in message >
     
  9. R.Legg

    R.Legg Guest

    5.3 for purpose of explaination) is that I would be playing the
    A game? Better to count on worst case and plan on what corrective
    action will be available when it happens. You can fiddle with the
    voltage by adjusting shunt current (removable shunt r values in
    parallel), or you can link in/out 650mV steps provided by a series
    diode string (2.5% increments at 27V).
    Allowing for 80uA of quiescent bias and 20uA of resistive divider
    current is sufficient. I'd run the sensing chain at 500uA, out of
    respect for the 27V operation, and try to keep current below 5mA under
    transient or single fault abnormals.

    You want this current to be more than Topswitch bias, with a suitably
    sized shunt resistor, to make Topswitch input bias irrelevent.
    Lesson learned (?)..I would hope it's...
    "Get it right before it hits production".

    RL
     
  10. Yzordderex

    Yzordderex Guest

    (R.Legg) wrote in message > >

    snippd a bunch hea


    Yea, Actually the oil cap solution was to take care of another's
    engineer's poor design. My first engineering position was as a
    sustaining engineer. The position required me to re-design all the bs
    circuits that the guys who were bringing home twice the pay I was were
    releasing into production. My stuff works!
     
  11. Yzordderex

    Yzordderex Guest

    snipped a bunch hea


    Final solution to zener diode distribution curve was for me to write a
    reasonable power supply specification that would work fine. Other
    (green)engineer who had asked for the power supply had over specified
    (as usual). Asked for 1% when 10% fine.

    Used a pair of 5.1v 5% zeners in series with the Topswitch feedback
    circuit. With the 5.8v or so bandgap reference in Topswitch the thing
    shouldn't move but a few percent. I was able to get the load
    regulation +-1% with just the feedback winding (50% to 100% load). No
    TL-431 or opto.

    thanks for input guys.

    regards,
    Old Grey Engineer
     
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

-