Connect with us

Help with power supply noise., Please.

Discussion in 'Electronic Design' started by Mike, Aug 1, 2007.

Scroll to continue with content
  1. Mike

    Mike Guest

    I am building a homebrew bench power supply and it is working very well except for too much noise on the
    output. It is a single output with adjustable CV from 0 to 40V and adjustable CC from 0 to 4A. I have posted a
    schematic of the regulator section and a digital photo of the noise waveform under no load on a.b.s.e in PDF
    format. I posted it using the same suject line. I would greatly appreciate it if someone could give me some
    ideas of what to try in order to reduce the noise. I fear it is doomed to have a noisey output since it is
    point to point wired on perfboard, but I sure hope not. I tried very very hard to get the ground connections
    right and have rearranged them to try minimize the noise, but those changes didn't affect it much either way.
    This noise is very stable in that it doesn't change much from no load to the full 4A load, It gets fuzzy under
    full load, maybe from some low level high frequency oscillation. It doesn't change with output voltage. I
    increased the value of C5 to .047uf to get the noise down to where it is now. With C5 at 2200pf the noise is
    about 2 1/2 times what it is now and increasing C5 further decreases the noise very little.

    There is another curious thing that happens that has me stumpped. I intended to use an OPA2277 for U1. With
    the 2277 installed the transition from CV to CC mode in very abrupt, I.E. with a 10ohm load resistor
    connected, the voltage set to 20v, and the current set to 2.5A it draws 2A as expected. When I lower the CC
    setting it has no affect on the output untill it is set to about 400ma below the current being drawn or 1.6A
    in this case. The current then snaps down to 1.6A and decreases smoothly from there down. It did this no
    matter what current level I used to test it. It also does it if I increase the output voltage to excede the CC
    seting. With an LM6132, LF412, or a TL082 installed it transitions to CC mode quite smoothly.


    Mike


    When truth is absent politics will fill the gap.
     
  2. ehsjr

    ehsjr Guest

    A couple of thoughts: increase C1 to 470, and add a small
    cap in parallel; add a small cap directly across the
    output. Looks similar to 120 hZ ripple mixed with noise.
    Can't tell what the noise frequency is.

    Ed
     
  3. Andrew Holme

    Andrew Holme Guest



    Have you brought the ground terminal of the 2.500V reference; U3 pin 7; and
    the earthy ends of R26, C9 and C10 to a common point? Any noise between
    these nodes would be bad.
     
  4. Guest

    Think I'd be schrieking with fear if my own supplies ran that
    clean :).
    It's capacitor ripple breaking through. Seems a very non linear
    control loop from sampling the output to taking corrective action.
    lags. I'd be inclined to lose D2 and drop R8 R17 to maybe 1/10 their
    present values and then add some local linearising with a (say) 220ohm
    in Q4's emitter.
    Puzzled by U3, it seem way overkill.
     
  5. legg

    legg Guest

    The fact that the noise is largely independant of load suggests that
    it does not originate in the unregulated input rail, as this source
    would increase with load-induced ripple.

    You may be able to spot it's earliest source - it should not be
    present in the outputs of U2. The higher frequency stuff might benefit
    from a little miller capacity around Q4 (picofarads) or an emitter
    resistor there, (not enough to affect source rejection).

    You might try re-orienting your mains transformer, with respect to
    your circuit board, while monitoring the output noise. You may find
    that the lower frequency line undamental and harmonics vary favourably
    under some circumstances. The steel leadframes used in many
    semiconductor packages (including the reference, tend to be affected
    by magnetic fields that have a predictable orientation, and which
    reduce rapidly in strength with increasing distance.

    It was once demonstrated feasible to reduce output noise of some
    models in a commercial series of linear power supplies into the
    microvolt region, simply by mounting the regulator board assemblies at
    right angles to convention, on the unchanged industy-defined chassis's
    metalwork.

    All the longer connections to pots, or larger-bodies pots themselves
    are subject to pick-up. Perhaps R11 should be grounded, and not used
    as a voltage pick-off point and C5 should have a small resistor in
    series. C4 or a similar part might be more effective on the actual
    output terminals.

    A more smooth transition between current and voltage regulation might
    be possible if U3 had reduced influence. It does, after all, only have
    to over-ride a milliamp from the output of U2B, using it's full output
    voltage compliance, to zero U1A input pin3.

    RL
     
  6. Mike

    Mike Guest

    I have solved the noise problem and I want to thank all who provided input.
    As best I can tell the noise is down to about 600uV P-P and mostly 120Hz. The 50KW AM broadcast transmitter
    located close by has a tendency to inject RF into my scope probes which makes for pretty fuzzy traces. The
    biggest single difference came when I put a very large bypass cap (2200uf) on the -12V line. I thought that
    was a bit strange because the +12v line has considerable noise riding on it that's coming from a brushless 12V
    fan and it just barely shows up in the output. The remaining noise does not change under a full 4A load
    either. At least I can't see it on my scope. The noise may even get lower once I get everything mounted in a
    metal case and shorten the cables to the front panel pots. I picked up some 2 conductor shielded cable at a
    hamfest on saturday and using that to connect the pots may help a bit also.

    Btw, It concerns me that I can't measure any difference in output voltage between no load and a 4A load. I
    used an HP 3478A 5 1/2 digit meter and the 40.000V reading did not change when when I connected my 10 Ohm load
    resistor. I wonder what I have unknowingly traded off for such good regulation. I would have guessed it would
    be noise and/or instability. I suppose that somewhere down the road I'll find out. It does drift a few
    millivolts as it warms up tho. That's probably not helped by having to use the LM6132.

    Now to make a gizzmo to switch my load on and off so I can measure the time it takes the thing to respond to
    step change in load.

    Thanks again.

    Mike



    "The scientist is possessed by the sense of universal
    causation...His religious feeling takes the form of
    rapturous amazement at the harmony of natural law,
    which reveals the intelligence of such superiority
    that, compared with it, systematic thinking and acting
    of human beings is an utterly insignificant reflection."
    Albert Einstein (theoretical physicist)
     
  7. legg

    legg Guest

    You should probably also examine the condition where the
    microprocessor switches the input voltage. A sudden change of this
    magnitude could have it's own problems. You might need to borrow a
    storage scope to do this.

    It's fairly easy to examine load transient response by simply
    switching the 10ohm resistor at the load end using a mosfet. With pnp
    output buffers, you will find that output terminal decoupling is
    critical. Larger electrolytics may be needed. A measurable load
    regulation is not neccessarilly a bad thing.

    re- bypassing -12V - improving input decoupling to the regulator might
    be more efficient, and making sure the return reference is optimally
    located - it's a low current kelvin connection in most commodity
    regulators - whether it works better on one side of the current sensor
    or the other will be the question. I'd suggest RC decoupling the fan,
    in any eventm to keep it's ripple local or give it it's own
    independant, crude regulation.

    I'm pleased not to hear that the input is not from an SMPS ( I assumed
    you'd mention by now, if so), as this would complicate things
    somewhat. Suggest you snub the rectifiers and voltage change-over
    hardware iteratively, while examining output noise.

    RL
     
  8. Mike

    Mike Guest

    Oh yeah, I kinda forgot about checking that. I'll check it out for sure.
    I have an HP 54502A scope that should do nicely to capture the transition.
    That's exactly what I did. I used a 555 driving 3 IRF640s in parallel.
    I figured that would slow down the on/off transitions to avoid inductive
    ringing and that seems to work nicely. I have it running at 50Hz with a 50% DC.
    I let it run and have been playing with the output electronlytic as you suggested.
    I think a 100uf instead of the 39uf might be reasonable.
    I also found that the output was ringing for 2 cycles when the load turned on
    and it took 120us for that ringing to die out. I thought that was too long and
    found that reducing C8 from 5600pf to 2200pf shortened that to 40us.
    The aux supplies, relay drivers, and line filter are all on a seperate
    etched PCB. There is a 1000uf filter cap and a 2" run to the input pin
    of the 7912. There is also a 2.2uf cap right at the input pin. I just cut
    the trace between the 2 caps, inserted a 100 ohm resistor, and replaced
    the 2.2uf with a 10uf tant. I still have 17.5V input to the regulator.
    I haven't removed the the 2200uf bypasa on the regulator board, but
    the noise dropped to about 400uv P-P. I think that's plenty good enough.
    I played with the reference return earlier and found that it is optimum
    where it is, on the high side of the sense resistor where the neg output
    lead is grounded on the board.
    The case I have has a 120VAC fan already mounted, so I'll just use that and
    eliminate the 12V fan altogether. It's a little louder but it does moves more air.
    Oh no, No switcher! I thought about preregulating with one, but was afraid of the noise
    and decided to do it the easy way and just select one of 2 input voltages.

    I was hesitant to use the micro for the same reason, but it is not a problem at all.
    I'm glad I did use it because it allows me to easily do some things that wouldn't
    be too practical otherwise. I have a nice 20X2 LED backlit LCD that shows me the
    power output, the CV and CC setings, the heatsink temperature, and allows me to select
    the low or high input voltage and to shutdown in the event of a heatsink overtemp.
    I also added a standby switch that signals the micro to drop both input relays. I just
    have to remember to look and see what happens when it comes out of standby.

    Sometime I'll need add an overvoltage protection circuit. I spotted a nice simple
    and accurate crowbar using a triac and a TL431 in a TL431 datasheet.
    Will do. At least it only changes the input when I manually change the voltage
    so it's not quite as critical as it might otherwise be.

    Whew! I never thought building a decent power supply was so involved.
    I'm sure learning a lot doing it tho and it'll be a lot better than
    the cheapie I've been using!


    Mike


    "The scientist is possessed by the sense of universal
    causation...His religious feeling takes the form of
    rapturous amazement at the harmony of natural law,
    which reveals the intelligence of such superiority
    that, compared with it, systematic thinking and acting
    of human beings is an utterly insignificant reflection."
    Albert Einstein (theoretical physicist)
     
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.
Similar Threads
There are no similar threads yet.
Loading...
Electronics Point Logo
Continue to site
Quote of the day

-