Weird noise in Railsplitter circuit: Crossover distortion?

Discussion in 'The Projects Forum' started by Félix Thouin, Feb 27, 2016.

  1. Félix Thouin

    Thread Starter New Member

    Aug 14, 2015
    6
    1
    Hi all,

    I have recently built a railsplitter using an OPA551 from TI in follower mode after a suitable voltage divider. The circuit is powered by the following wall wart 24VDC power supply (see Eagle schematic below).
    https://www.digikey.ca/product-detail/en/triad-magnetics/WSU240-0500/237-1459-ND/3094985

    It provides a nice +-12VDC to loads down to 100 Ohms. However, when I use it to power my Wien bridge oscillator circuit (see other eagle schematic), weird spikes appears in the output waveform (see scope print screen). This bridge used to be powered by a linear rectified power supply and did not exhibit this weird behavior.

    It is a pain at low frequencies, since it adds spikes to the output waveform. At higher frequencies, theses spikes add ''phase jitter'' to the output, which makes it looks shaky on the oscilloscope.

    I am wondering if you guys had any idea what might be causing this. I read crossover distortion might produce similar spikes, but it doesn't look like the cases I've previously seen (some kind of cut in the waveform).

    Thanks in advance!
     
  2. #12

    Expert

    Nov 30, 2010
    16,257
    6,763
    There you have it. You're getting some high frequency crap out of your wall wart. Best to try to filter it out. Place a 10 uf aluminum electrolytic or tantalum capacitor plus a 100 nf ceramic cap right on the power leads of your IC. You might also try maybe 100 ohms in series with the power entry wire and add a the same pair of capacitors there.
     
  3. bertus

    Administrator

    Apr 5, 2008
    15,646
    2,344
    Hello,

    Also put a capacitor of 10 μF accross R2 to avoid noise from the powersupply on the virtual ground.

    In this thread there is more info on the decoupling capacitors #12 mentioned:
    Decoupling or Bypass Capacitors, Why?

    Bertus
     
    #12, GopherT and JWHassler like this.
  4. #12

    Expert

    Nov 30, 2010
    16,257
    6,763
    I missed R2.:oops:
    Your C1 is in the position I named second, but big electrolytic caps don't have good response at high frequencies. The bigger they are, the worser they are. That's why we add a small ceramic cap. They work up into the megahertz range.;)
     
    Félix Thouin likes this.
  5. Félix Thouin

    Thread Starter New Member

    Aug 14, 2015
    6
    1
    Thanks a lot for the answer guys. I thought adding a huge electrolytic capacitor would have gotten rid of all of the ripples: looks like it didn't!

    I have attached a revised schematic: does it comply with the advices you gave me? I also don't understand why a 100ohm resistor in series with the output of the wallwart would help. Is it a current limiter?
     
  6. #12

    Expert

    Nov 30, 2010
    16,257
    6,763
    Put another 100 nf next to C3...or not. If the problem is fixed without it, don't bother.
    Your C1 and C2 should be from pin7 to pin 4, not to a nearby ground trace on the circuit board.

    The suggestion for 100 ohms in the power line depends on a couple of things. If the chips only need a little current you can afford to waste some voltage in that resistor. The way it works is that the resistor and the capacitors make a filter for high frequencies. Let's pretend the wall wart is switching at 100 KHz and you need 6 milliamps to run the chips. The 100 ohm resistor uses up 0.6 volts of DC. The 10 uf capacitor uses up no DC current but it has an impedance of 0.159 ohms at 100 KHz. For the primary switching frequency, the noise is diminished according to Vout = Vin x 0.159/100.159 ohms. If you start with a volt of high frequency noise, it will come through the resistor diminished to 0.00158 volts while the DC voltage is diminished from 24 volts to 23.4 volts. That's 56 decibels of noise abatement for the primary switching frequency and even more for the harmonics. The problem with big aluminum electrolytic capacitors is that they don't work at megahertz frequencies. That's why we use a small electrolytic capacitor and the 100 nf capacitor. At 10 MHz the 100 nf capacitor has also accomplished 0.159 ohms of impedance. So you see, it's a frequency filter that is effective from DC to several megahertz.

    If I'm wrong about the current, you can lower the 100 ohm resistor to something you can afford, voltage-wize.
     
  7. Félix Thouin

    Thread Starter New Member

    Aug 14, 2015
    6
    1
    I see, I will make sure to have them close to the IC's pins.

    The chip might use up to 200 mA of current (worst case). A 2 Ohm resistor would give a similar effect as the one in your example, and dissipates moderates amounts of power (100 mW). However, the noise reduction would be a lot less drastic. Do you recommend still adding such a resistor? I suppose it is better than nothing.
     
  8. #12

    Expert

    Nov 30, 2010
    16,257
    6,763
    Any resistance at all is an improvement, but with 2 ohms you're getting down close to the resistance of circuit board traces.
    Most of the time nobody adds the resistor. I didn't see how that rail splitter could use 200 ma so I guessed wrong.
     
  9. AnalogKid

    Distinguished Member

    Aug 1, 2013
    4,518
    1,247
    What country are you in? The noise bursts are occurring approx. 8 ms apart, full-wave rectified power line frequency in the US.

    IF the wall wart is a small switcher
    AND IF it is very lightly loaded
    THEN it might be running in a discontinuous or hiccup mode

    Add a diode in series with the +DC lead coming in from the WW.

    ak
     
  10. bertus

    Administrator

    Apr 5, 2008
    15,646
    2,344
    Hello,

    The following PDF has information on decoupling on higher frequencies:
    rf_proto-1.pdf
    Look at the pages 8 and on.

    Bertus
     
Loading...