Power supply, but no power???

Discussion in 'IBM' started by U. Cortez, Apr 7, 2005.

  1. U. Cortez

    w_tom Guest

    Even in the very first IBM PC, power supply included that
    voltage monitoring function (that is now part of the single
    IC). A function completely different from the overvoltage
    protection circuit - which was also in that IBM PC. Some
    power supply controller failures can cause overvoltage.
    Totally unacceptable 30 years ago as today. The crowbar (or
    equivalent) is required today as it was 30 years ago on switch
    mode power supplies. Even the Intel spec (which is only for
    switch mode power supplies) says
    Nothing new in this requirement. However when dumping
    product into a marketplace dominated by 'bean counter'
    mentalities, then profits are just too great. How to suspect
    a defective power supply? No long list of numerical
    specification and no specific listing of overvoltage
    protection? Then it is a power supply for the market of 'bean
    counting' computer assemblers.

    That schematic is further proof that many power supplies are
    missing essential functions. Some power supplies fail and
    then destroy computer peripherals. A failure directly
    traceable to a human who buys on price rather than learn basic
    electrical principles. In a market where so many 'experts'
    don't even know how how electricity works, then many clone
    computers do have $25 power supplies. When a supply fails due
    to a manufacturing defect, that 'computer assembler' instead
    blames disk drive and motherboard damage on mythical
    transients. Why? Most don't even know what electronic part
    has failed; let alone know why. Better to just blame
    something mythical that others will quickly believe.

    Does the power supply provide a long list of numerical
    specifications? Many do not provide specs which means
    essential overvoltage protection is not required. Its called
    dumping. Made profitable by computer assemblers don't even
    understand why component manufacturers should provide
    numerical specs.

    Specs from a responsible power supply manufacturer are quite
    long. Abridged election from one responsible power supply
    vendor whose supply could never sell for $25 retail - because
    it includes essential functions:
    Specification compliance: ATX 2.03 & ATX12V v1.1
    Short circuit protection on all outputs
    Over voltage protection
    Over power protection
    100% hi-pot test
    Efficiency; 100-120VAC and full range: >65%
    EMI/RFI compliance: CE, CISPR22 & FCC part 15 class B
    Safety compliance: VDE, TUV, D, N, S, Fi, UL, C-UL & CB
    Hold up time, full load: 16ms. typical

    Does your power supply specifically, with numbers, make
    these claims? Why not? Symptoms of a supply that does not
    have these essential functions; which can create future
    computer failures. Demonstrated are how those with minimal
    education recommend and buy power supplies. Schematics from
    Franc Zabkar only demonstrate that power supplies without
    essential functions are being manufactured. Some computer
    failures can be created by power supplies missing these
    essential functions. Many computer assemblers may then invent
    other suspects to blame. If a power supply failure coincides
    with disk drive damage, then first look for a missing
    overvoltage protection circuit.
    w_tom, Apr 12, 2005
    1. Advertisements

  2. U. Cortez

    Franc Zabkar Guest

    I understand that separating the two functions (OVS and OVP) gives
    added security, but I can't see the need for a brute-force approach to
    OVP. By this I mean that it is not necessary to clamp the output with
    an expensive high-current zener or SCR - one can much more elegantly
    achieve the same end, and still comply with Intel's spec, by turning
    off the PWM controller, or by turning off its drive transistors. See
    this DTK PSU which implements OVS, OVP and OPP with about $2 worth of


    I have a 1000W minicomputer SMPS whose control module senses the
    output and shuts off the oscillator in the event of an OV fault. There
    are no expensive OVP parts, and this is in an SMPS that has 4 or 5
    screw terminal capacitors the size of soft drink cans, stud mounted
    diodes on massive heatsinks, four TO3 chopper transistors, and an AC
    fan. The +5V cables (+5V @ 150A) are thick enough to start my car.
    Some TV sets use 130V protection zeners (eg R2M, R2KY, $1.85 retail)
    on their 100-115V supply rails. These designs have no OVS. Ironically
    it appears that the designers have chosen this sledgehammer approach
    because it costs *less* than OVS. I'm not really comfortable with such
    a design because its success depends on the failure mode of the
    protection device. IME the zener always fails SC, and therefore
    protects the TV, but if it fails OC (unlikely, but possible), then
    there is no protection at all. Killing the oscillator would be much
    safer. In fact, HV protection and beam current limiting is usually
    implemented by shutting down the horizontal oscillator.

    I guess a comparable analogy may be MOVs in "surge protected" power
    boards. They may sacrificially absorb the first surge, but thereafter
    they are useless.
    Sorry, I don't see it. You state that the original IBM supplies had an
    OVP circuit, but you don't elaborate. How exactly did they do this? I
    have the original IBM AT Tech Ref Manual but it doesn't adequately
    spell out the PSU spec. Can you recommend one upmarket PSU that
    handles OVP by brute-forcing the output(s)?
    True. I witnessed a discussion at aus.electronics where an SMPS failed
    in such a manner that AC leaked into the DC side of the switchmode
    transformer, causing catastrophic damage to the PC. I suggested that
    an external crowbar circuit could provide protection against such
    disasters but the responses were negative.
    What I find hard to comprehend is how even the upmarket ATX PSUs can
    deliver their claimed power given the relative size of the components
    in my 1000W SMPS. For example, when comparing the mains filter caps,
    one is the size of a Coke can, the other is smaller than a C size
    battery. Has technology really improved that much?

    - Franc Zabkar
    Franc Zabkar, Apr 12, 2005
    1. Advertisements

  3. U. Cortez

    w_tom Guest

    You are assuming a power supply controller will see all
    failures and then shutdown accordingly. What happens when
    controller's feedback fails? A feedback circuit that will be
    discussed again below. Power supply controller outputs more
    power trying to raise a voltage that never rises. IOW power
    supply has output excessive and destructive voltages. Just
    another reason why "separate and distinct" overvoltage
    protection has been required 30 years ago as it still is
    required today.

    Neither you nor I care how a particular power supply meets
    this 'well proven to be necessary' OVP function. We don't
    even care if the function uses a simple and easily constructed
    crowbar or uses something different. The point remains a power
    supply must provide that defined function. Schematic from
    electro-tech.narod.ru quite obviously violates the industry
    requirement. It has no overvoltage protection. It
    demonstrates how power supplies are sold for well under $60
    retail. They forget to include essential functions such as
    OVP. That supply from http://electro-tech.narod.ru is
    designed to be dumped into a market driven by 'bean counter'

    Provided were numeric specs for a supply that does provide
    the OVP circuit. Back to the original point. Properly
    designed supply does not and can not sell for $25 retail. So
    how do others sell power supplies for $25 retail? They forget
    to include required and necessary functions such as
    Overvoltage Protection.

    Is this done by a crowbar circuit or by some other means?
    Neither you nor I care. Industry standards demand that OVP
    function exist for 'long proven' necessary reasons. Some
    supplies do provide such functions. But these minimally
    acceptable supplies cannot sell for $25 retail - again
    repeating the bottom line point.

    In the meantime, that DTK PSU from www.pavouk.comp.cz
    violates another essential function. The power supply must
    provide galvanic isolation of at least 1000 volts. Therefore
    the "Feedback" circuit must contain an optocoupler or
    something equivalent. The DTK PSU has no such isolation. It
    violates another industry standard. And so another essential
    function would be 'forgotten'.

    Earlier a power supply controller would output overvoltage
    because the feedback circuit failed. Controller never knew it
    was outputting excessive voltage. What feedback components?
    Same optocoupler that is required with galvanic isolation.
    Just another missing specification to sell at $25. Again the
    point. Power supplies missing essential functions and
    routinely dumped into the market because so many computer
    assemblers don't even have basic electrical knowledge; never
    learned about galvanic isolation, overvoltage protection,
    feedback current limiting, or overpower protection.

    When a computer assembler looks only at power and price,
    then a game of specmanship is afoot. For example, a Dell or
    HP power supply may claim only 250 watts. The 'bean counting'
    computer assembler then claims that is woefully too small.
    And yet if the same power supply was being marketed by others
    to computer assemblers, then suddenly the same supply is rated
    at 375 watts. Why? They don't list the output power. Rated
    is maximum power that a power supply might consume. The 'bean
    counting' computer assembler then declares the HP and Dell
    supplies are undersized.

    It gets even more interesting. A power supply must be
    completely shorted out and still must not be damaged. And yet
    here are power supplies, designed for a 'bean counter' market,
    that self destruct even before reaching 100% load:

    First thing to look for in any power supply: if the supply
    manufacturer does not provide a long list of numerical specs,
    then bet it is a scam. More responsible power supply
    manufacturers provide numerous numerical specs. One need not
    even know what those specs mean. Just having written numerical
    specs is a first requirement. Manufacturer commits; says to
    the 1% who know technology that this supply does provide these
    functions. However a manufacturer who is dumping inferior
    supplies into a market of computer assemblers must disempower
    the 1%. He must provide no written specifications. Then a
    knowledgeable 1% cannot warn the other 99% of a defective

    Welcome to a world where so many power supplies don't
    provide specs and are then recommended based only on the price
    and watts.

    #1 requirement for a power supply: it must provide a long
    list of numeric specs. If that power supply does not
    specifically state overvoltage protection, then another
    essential function is missing.

    We care less how he provides overvoltage protection. We
    care more that he claims to provide OVP. That missing
    function demonstrates why so many supplies sell for only $25
    and $40 full retail.
    w_tom, Apr 12, 2005
  4. U. Cortez

    Franc Zabkar Guest

    I did write that "I understand that separating the two functions (OVS
    and OVP) gives added security", so I understand that there are some
    caveats with a single-chip approach.
    In most cases the OVS circuitry will detect an OV condition and shut
    down the oscillator. I've actually seen such faults in other PSUs, eg
    when optocouplers or feedback transformers or sense resistors go OC.
    One potential problem with single-chip solutions is if the internal
    reference voltage is shared by both the OVS logic and the PWM
    regulator. In such cases a fault in the voltage reference may cause
    catastrophic damage. Having said that, it may still be possible to
    separate these two functions within the same IC. In any case the DTK
    circuit I alluded to *does* separate these functions and *does* comply
    with Intel's spec. And it does this at the cost of a few small-signal
    transistors, diodes, and passives. In fact the BOM for the OV, PG, and
    PWM circuits would cost about $1 or $2.

    $0.23 /1000

    $0.16 /1000
    It *does* have OVP. It's just that the implementation does not appear
    to rigidly follow Intel's spec. For example, if I interfere with the
    +5V feedback by removing R62 from pin 14, the output voltages will
    rise to a point where the OVS pins (3,4,6) will trigger a shutdown of
    the oscillator.
    Specs can lie, or at least be deceiving. A high price does not
    guarantee quality or spec compliance.
    On the contrary, I do. At the very least, knowing how a certain
    function is implemented can explain any differences in price. For
    example, if each of the major rails were shunted by high current SCRs
    or protection zeners, then one could expect to pay a much higher
    price. OTOH, if OVS and OVP were implemented with 50c ICs, then one
    would not expect such functions to impact noticeably on the overall

    In the absence of specific examples from you, I'm left to examine
    various PSUs at this website:


    Unfortunately the site is in Japanese, but there is just enough info
    to ascertain the inner workings of both low-end and high-end ATX PSUs.

    For example, Antec's TruePower 550 is pictured and described here:

    The ICs of interest are the UC3844 PWM regulator and the TPS3510 PSU


    The costs are:

    $0.45 /1000

    $0.80 /1000

    The TPS3510 IC implements OVS, UVS, OVP, UVP, and PG. I suspect its
    Fault Protection Output pin controls the UC3844 regulator. There
    appear to be no other OVP parts.

    Another manufacturer, PC Power & Cooling, uses the TPS5510 supervisor
    IC in its Turbo-Cool 450ATX:
    That may be true, but if we were to use your definition of OVP, then
    some, if not all, branded PSUs do not provide this function either.
    The regulator and the feedback circuit exist on the same side of the
    PSU, ie on the DC side. Therefore no "galvanic isolation" is required
    between the two. In any case, isolation between the AC and DC sections
    is provided by transformers, at least when high currents are involved.
    AFAICS, cheap optocouplers are only used in the 5VSB circuit, if at
    all. In the DTK case, no special feedback is required in the 5VSB
    section because the 78L05 linear regulator doesn't require it. OTOH,
    the LC-B250ATX uses an optically isolated TL431 voltage reference to
    achieve the same end. The Antec Truepower PSUs appear to do it this
    way, too. All three PSU designs are "galvanically isolated" where
    That works both ways. Ignorant people assume that a higher price means
    higher quality. They have never heard of "badge engineering". That's
    where an up-market vendor/manufacturer rebadges a down-market product
    and extracts a price premium by leveraging his reputation rather than
    his technical expertise.
    I have to agree. In fact my own measurements lead me to believe that
    typical Athlon and P4 systems will never draw more than 150W.
    The first thing that caught my eye was this:

    "We've ... now decided to tackle this volatile subject with testing of
    21 different, high-end power supplies in the THG lab in Munich,
    Germany. In spite of the high end-user retail prices, our lab
    technicians were surprised by the test results. No fewer than 6 power
    supplies struggled under full load: 3 of the candidates simply went up
    in smoke, while the other 3 shut down prematurely."

    It seems that price, specs and reputation are no guarantee of
    performance. ;-)

    This also caught my eye:

    "Our attempt to test the Noise Magic power supply had disastrous
    consequences. This is a modified Enermax unit fitted with a Papst fan.
    It stopped working after one minute. When we tried to switch it back
    on again after a suitable delay, it simply burned out."

    I would have thought that even a badge engineered Enermax was not a
    "bean counter's" PSU ;-)

    Anyway, you have introduced a new subject, namely overload protection.
    OVP is cheap to implement, but sensing current is more difficult than
    sensing voltage, so I would expect that many manufacturers would take
    shortcuts here. The aforementioned LC and DTK PSUs both sense a
    voltage at the drive transformer, whereas other PSUs monitor the
    current flow in the chopper transistor(s). I'm not sure that's the way
    the Antec 550W PSU does it, but its UC3844 PWM controller *does* have
    a current sense pin. The application circuit shows this pin connected
    to a current transformer. My 1000W minicomputer PSU also uses a
    current transformer in the chopper circuit to indirectly sense the
    load current. I suspect that the OPP circuits in cheaper PSUs (such as
    the LC and DTK) would be able to respond correctly to dead shorts, but
    not necessarily to overloads.

    - Franc Zabkar
    Franc Zabkar, Apr 13, 2005
  5. U. Cortez

    Jon Guest

    Jon, Sep 1, 2005
  6. U. Cortez

    CJT Guest

    If your MB has the 4-pin (square) auxiliary power connector, and you
    haven't connected it, you will get that result.

    Another possibility is that, while changing the PS, you accidentally
    disconnected the power button connector. It's small and easy to
    pull off its header.
    CJT, Sep 1, 2005
  7. A mistake people make every now and then (including myself, and I've
    assembled lots of computers) is to set the FDD power connector one
    step to the side (only connecting three of the four pins), and this
    will cause a shortcut.

    Another possibility is that something else is dead too. All defects
    aren't visible to the naked eye...

    Disconnect all power cables except the one to the motherboard, and
    also disconnect the IDE and FDD cables (a defective drive can prevent
    a computer from starting).

    Then try to start the comp again.

    If it works now, start reconnecting components one by one, and do a
    test start between each...

    If it doesn't start even with all drives etc disconnected, your mobo
    might be fried. Or maybe your new PSU is defective too. Or something
    Peter Emanuelsson, Sep 1, 2005
  8. U. Cortez

    PWY Guest

    On the back of the power supply, there is a switch where voltage can be set
    to either 120 volts or 240 volts. Make sure it is set to the correct
    voltage. This often overlooked.
    PWY, Sep 2, 2005
    1. Advertisements

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.