AAC Volume 3 - power supplies

Discussion in 'General Electronics Chat' started by takao21203, Jun 30, 2012.

  1. takao21203

    Thread Starter Distinguished Member

    Apr 28, 2012
    3,577
    463
    http://www.allaboutcircuits.com/vol_3/chpt_9/2.html

    I am not happy with this, even if it is in draft stage currently.

    For instance a distinction between "low-end" switchers,
    and "expensive" switchers is made.

    The former would have ripple as high as unregulated power supplies.
    And the latter would be ripple free.

    This has not much correlation to reality.

    Most switchers nowadays have little ripple currents, however they are still significant, for instance they can disturb certain circuits.

    All switchers can become improved by suitable filters. There are not really "low-end" SMPS, and "expensive" SMPS.

    Important in my opinion are isolation SMPS, using a flyback transformer, and buck-mode SMPS, using only a storage coil.

    As well discretely built, IC based, or complex circuits like for instance ATX supplies.

    There is also a topic called ripple regulated supplies. This is not the usual term. They rather count as Thyristor regulators. And the problems they cause can depend on the load, and the power level (for instance assymetric loading of the grid, as well unwanted currents inside the mains transformer).

    Circuits like these for instance have been used in television sets in the 1970s, but all Thyristor based, never heard of a transistor based regulator of this type.

    Linear regulators by no means have to be brute force, they are often used when only small current are needed. And by no means limited to transistors only. It's correct they are often avoided nowadays for larger currents.

    The article as such is not very useful, not enough correlation to reality, needs major rewrite.

    When "low-end" switchers are mentioned, they should be described properly, for instance abused linear regulator ICs, or based on simple transistor circuits. They are not always of bad efficiency, and application depends on the purpose. Low ripple is not always relevant, sometimes not at all for instance for bulbs. And since the frequency is in the kHz range, filtering is quite simple.
     
  2. Wendy

    Moderator

    Mar 24, 2008
    20,766
    2,536
  3. THE_RB

    AAC Fanatic!

    Feb 11, 2008
    5,435
    1,305
    Bill, assuming the transformer/rectifier is low impedance and always charges the cap to Vpk every half-cycle, can't you get a ball park figure for ripple voltage based on the capacity of the cap vs load current? Or were you talking about something else?

    Re the article I've never seen a "ripple regulated" mains transformer PSU but it looks very much like many automotive regulators that simple turn the output on/off at a voltage setpoint, and they tend to oscillate locked to the input half-cycle waveform so they work very much like a phase angle regulator, off for the first part of the wave then come on when the battery drops a fraction for the last part of the wave.

    One early SMPS regulator type they used to use in old TVs and Mil equipment is not mentioned, the "pulse" type or "packet" type. It sends one switched pulse of power through an inductor to the output when the output is below the setpoint. They could be reasonably efficient, even with the clunky transistors of the day. So it's a variable frequency type and was dropped pretty quickly in commercial equipment because it makes all kinds of nasty whistling noises. :D

    One brand was even done using an SCR as the pulse switch, with an tiny extra pulse transformer to turn the SCR off. From memory those ones would run at about 4kHz at low currents and were unpleasantly noisy.
     
  4. takao21203

    Thread Starter Distinguished Member

    Apr 28, 2012
    3,577
    463
    I built some power supply related circuits for research purpose and/or reference- to see how they look like.

    For instance a switcher based on LM317, and one PNP transistor. Most coils did not work, but the ballast coil from a small CFL turned out to be suitable. It makes noises: http://www.youtube.com/watch?v=xg03Sj3GKM0

    So maybe this could be some kind of low-end approach, maybe used in 3rd world, recycling the coils from CFLs etc., not even having a PCB and things like that. It's not totally bad the transistor remains cold.
     
  5. Wendy

    Moderator

    Mar 24, 2008
    20,766
    2,536
    If you happen to post or point to schematics (simplified preferred, they don't actually have to work) for the article I may use them, and will share credit. I just want to get er done.

    As to schematics, I will redraw them to meet the requirements and keep the feel of the book.

    I've recently come out of an extended deep depression, basically everything stopped for about a year. Now and then I go back into that mode, but I'm working on it. I have several chunks of gear half build to test some of this stuff out. I really want to get it right in the articles, if you know what I mean.

    I would like to see this book finished in my lifetime. When I get the power supply part done I'll pick something else to work on in the book.

    Is that your YouTube post takao21203? If we keep chatting like this I'm going to have to come up with a nickname that doesn't offend.
     
  6. takao21203

    Thread Starter Distinguished Member

    Apr 28, 2012
    3,577
    463
    Oh. I have not known about your half finished projects. But, I have some circuits myself that never were completed.

    Yes this is my youtube. Would you consider crush0nuo10 as offensive? To me it's not more ridiculuous than for instance Ubuntu. There is certainly nothing against the law or public order contained.

    Is there any user handle I should not use?

    The schematic is from the LM317 original user manual*. And I use a ballast coil from a CFL. I also copy it here into the thread.

    Yes why not use it as reference for low-end switching regulator.
    It's directly called low-cost regulator in the datasheet.

    *Each one is different you'd have to search a bit for that one.
     
  7. Wendy

    Moderator

    Mar 24, 2008
    20,766
    2,536
    No, but typing Takao21203 over and over could be a bit tedious. :D I wasn't talking about changing your handle, just a shorthand like Taka.

    The main thing to remember about schematics, it must be obvious. Someone reading the book may not have a clue about a LM317, for example, but a op amp is generic. Think ideal schematics.

    A good example is my PWM article (I wrote this with a bit of help from friends)...

    Pulse Width Modulation

    I'm not sure why, but our editor stuck it here...

    Chapter 11: DC MOTOR DRIVES

    http://www.allaboutcircuits.com/vol_3/chpt_11/index.html

    Why did I write the article? Because I made an experiment for Volume 6, and I needed to refer back to the book. BTW, this one hasn't been released yet...

    555 PWM Oscillator

    Long after all the hassle I discovered this section in the book....

    http://www.allaboutcircuits.com/vol_3/chpt_4/9.html

    about 1/3 down...

    [​IMG]

    Like I said, the book is a little incoherent, but deep. Makes me crazy it does.
     
  8. takao21203

    Thread Starter Distinguished Member

    Apr 28, 2012
    3,577
    463
    This is originally from Beyblade (a cartoon book series). Yes Taka is absolutely agreeable.

    The LM317 is quite common for instance LM317L (100mA version), used as voltage reference. Not as common as NE555, 1n4148, 7805, but I would include it to that list, among others like LM2576, or LM358, as well 2n3904/2n3906.

    Can circuits be done without LM317L? Yes voltage references don't have to be adjustable, too much effort anyway, so the 7805 can be used + 50/50 resistor divider.

    Normally it is not much used as switching controller.
     
  9. THE_RB

    AAC Fanatic!

    Feb 11, 2008
    5,435
    1,305
    Did you miss my post Bill?
     
  10. Wendy

    Moderator

    Mar 24, 2008
    20,766
    2,536
    uhhh, maybe. I'm a mod, I never miss posts! At work at the moment, so I'll get back with you.
     
  11. Wendy

    Moderator

    Mar 24, 2008
    20,766
    2,536
    Sounds easy doesn't it? I'm looking for a rough math model for both half wave and full wave (or CT). It will need to take into account voltage, load, and frequency and be able to calculate within a reasonable range the amount of ripple in volts. With this you will be able to tell if a specific regulator with a known dropout will work.
    It is actually the simplest type. Basically a little hysteresis so the unit stays out of linear mode, and no clock at all.
    Got any references? If it is distinct at all I will include it.
    I am setting myself up big time with this one. There is going to be someone, or more than one, who is unhappy with me or my results. However, if it gets finished, and is reasonably accurate, it can always be tweaked by future writers.
     
  12. THE_RB

    AAC Fanatic!

    Feb 11, 2008
    5,435
    1,305
    Hi Bill, with the "packet" type variable frequency PSUs I have fixed them many years ago but did not know the specific names. Google found this on the wiki SMPS page;
    http://en.wikipedia.org/wiki/Switched-mode_power_supply

    "Quasi-resonant zero-current/zero-voltage switch
    Quasi-resonant switching switches when the voltage is at a minimum and a valley is detected

    In a quasi-resonant zero-current/zero-voltage switch (ZCS/ZVS) "each switch cycle delivers a quantized 'packet' of energy to the converter output, and switch turn-on and turn-off occurs at zero current and voltage, resulting in an essentially lossless switch."

    From what I remember (this was in the mid 80's) they were quite resonant with the pulse transformer and caps working as a team to pass a "packet" of energy each switching time. Regulation consisted of an output setpoint that inhibited another cycle, and I remember watchign the frequency change on the 'scope in response to adjusting the dummy load. Sorry i can't remember which TV used them, it might have been the old '70's Krieslers. Wiki gives the impression these might still be used in very high power apps which makes sense.

    Re the other work, you are not alone Bill you are much loved and appreciated and I'm sure people would be willing to lend a hand if you asked. :)

    Re the capacitor ripply I think the discharge is fairly linear looking at ripple on the scope, and storage is linear to cap size, so you might be able to reduce the problem to something as simple as a capacitance:current ratio which gives a % drop in voltage? So for 1000uF and 1A the ripple will drop 10% from the Vpeak, with 100uF 0.5A it will drop 5% from Vpeak? Would probably get you close enough and be friendly to use as well.
     
  13. MrChips

    Moderator

    Oct 2, 2009
    12,449
    3,365
  14. Wendy

    Moderator

    Mar 24, 2008
    20,766
    2,536
    Vpp = I/(2fC)

    Hmmm, really simple. If that works it is half of it. I am not looking for terrific accuracy, just ballpark numbers.

    I have a capacitance box half built up, with a couple of transformers. I'll be using them (when finished) to test the formula.

    It is interesting, I thought I had researched wikipedia when I started this over a year ago, I wonder if this is new.
     
    Last edited: Jul 3, 2012
  15. MrChips

    Moderator

    Oct 2, 2009
    12,449
    3,365
    I've looked at the equations for the ripple voltage of a full-wave rectifier PSU.
    The approximation of Vpp_ripple = I/(2fC) is very close [if we assume RC >> 1/(2f) ].

    FYI, I did some simulation on Matlab:

    If the transformer Vac_rms center-tap is Vac

    Vdc = sqrt(2)*(Vac/2) - Vdiode

    I = Vdc/R

    Vpp_ripple = I/(2fC)


    There is another simple way of looking at this:

    Vpp_ripple = Vdc/(2fRC)

    Vpp_ripple/Vdc = 1/(2fRC)

    1/(2f) is the half-cycle period = 10ms for 50Hz or 8.3ms for 60Hz

    If you want the Vpp_ripple to be 1/10 of Vdc
    then RC must be 10 times 1/(2f)

    For 60Hz, let's make RC = 83ms

    So for example, a 12Vac_rms CT giving an 8V @ 1A supply,

    R = 8Ω, C = 10,000μF

    will have a peak-to-peak ripple of 0.8V. (I have done some rounding.)

    To summarize and simplify, make RC equal to about 80-100ms for 10% peak-to-peak ripple.

    I have worked out the peak diode current which I will post next (another time).
    Peak diode current will be about 13 times the average load current.
    Edit: (Geo's equation gives 15 times which I am willing to accept.)
     
    Last edited: Jul 3, 2012
  16. THE_RB

    AAC Fanatic!

    Feb 11, 2008
    5,435
    1,305
    Yeah I was pretty sure you could get a good enough result with a ratio of current to capacitance. :)
     
Loading...