Cascaded Op Amps Going Unstable

Discussion in 'The Projects Forum' started by kyle7119, Apr 20, 2014.

  1. kyle7119

    Thread Starter Member

    Feb 11, 2011
    71
    17
    My project is a 10W audio amplifier. I designed the power amplifier stage using a generic push-pull configuration. The amplifier must be able to work with a signal input as small as 50mV peak. I have stacked a couple of op amps on the front end in order to amplify the 50mV to ~20V peak. I can get about 4V peak out of one stage, but as soon as I stack additional stages, the output goes unstable past 2V peak.

    I don't understand why this is going unstable or how to fix it.

    I have attached a circuit diagram below.
     
    Last edited: Apr 21, 2014
  2. bertus

    Administrator

    Apr 5, 2008
    15,648
    2,347
    Hello,

    Are you using decoupling capacitors on the powerlines?

    This is given in the datasheet:

    Bertus
     
  3. kyle7119

    Thread Starter Member

    Feb 11, 2011
    71
    17
    Yeah, I have some 100nF caps on the lines. Forgot to throw those in the diagram.
     
  4. bertus

    Administrator

    Apr 5, 2008
    15,648
    2,347
    Hello,

    Why do you use 3 cascaded opamps, as one can give you the same amplification?
    Also does the signal generator have an offset?
    You have made a DC coupled circuit, this will amplify the offset.
    Put a capacitor at the output of the signal generator and see if the offset is creating the distortion.

    Bertus
     
  5. kyle7119

    Thread Starter Member

    Feb 11, 2011
    71
    17
    I need the bandwidth. The amp needs constant gain from 500Hz to 10Khz.

    The capacitor does not help.
     
  6. bertus

    Administrator

    Apr 5, 2008
    15,648
    2,347
    Hello,

    Do you actualy build the amplifier or is it only in simulation?

    The GBW of the opamp is 4 MHz.
    With a gain of 100 the bandwith will still be 40 kHz.
    Also the offsetts of the invidual opamps will be amplified and added to the total signal.

    Bertus
     
  7. kyle7119

    Thread Starter Member

    Feb 11, 2011
    71
    17
    It works perfectly in simulation. I actually built the circuit and this is where I am seeing the problems.

    The instability only manifests itself when the op amps are loaded (attached to the push pull).
     
  8. bertus

    Administrator

    Apr 5, 2008
    15,648
    2,347
    Hello,

    Looking at the schematic, I think that C7 is the wrong way around.
    It can also be that the values of C7 and C8 are to big for the opamp to drive.

    Bertus
     
  9. ericgibbs

    AAC Fanatic!

    Jan 29, 2010
    2,503
    380
    Check the OPA output drive specification into a 25uF load, this could be causing the instability.

    E
     
  10. AnalogKid

    Distinguished Member

    Aug 1, 2013
    4,542
    1,251
    In most circuits, and high-gain audio circuits in particular, ground path impedance is critical for stable operation. Can you post a photo of your build? Is this on a pc board or perf board or a proto board or what?

    A very small capacitor from the collector to base of either your driver or output transistors might calm things down. This is called a Miller capacitor, usually in the 1 to 22 pF range.

    On all of their IC audio power amps, National Semiconductor (now TI) shows an RC network on the output with a corner freq of around 100 KHz. See data sheets for the LM380 and LM386.

    Was there a typo in your original post? 500 Hz seems like a pretty high lower freq for audio.

    Your overal circuit gain is less than 300. This is not enough to raise 50 mV to 20 V.

    ak
     
    Last edited: Apr 20, 2014
  11. kyle7119

    Thread Starter Member

    Feb 11, 2011
    71
    17
    Tried removing that but still no luck.

    This really looks like a phase/gain margin problem from to much phase shift, but I can't figure out where it's happening.
     
  12. kyle7119

    Thread Starter Member

    Feb 11, 2011
    71
    17
    Here are the photos.
     
  13. kyle7119

    Thread Starter Member

    Feb 11, 2011
    71
    17
    Yeah, there was a typo in my post about gain.
     
  14. ericgibbs

    AAC Fanatic!

    Jan 29, 2010
    2,503
    380
    On that photo of the PCB with the MC33079 OPA, I dont see any HF decoupling, ie 100nF etc,.??
     
  15. DickCappels

    Moderator

    Aug 21, 2008
    2,655
    632
    Its good to know that the Spice model is stable.

    You have oscillation, it does not appear to be RF, and oscillation requires feedback. The only intentional feedback in within the discreet power stage. I suspect that Bertus hit the nail on the head when he mentioned decoupling.

    At about this point in debugging, I would drag out some substantial electrolytic capacitors and add them to ±15. It might be beneficial to isolate the output stage with a low value resistor and bypass caps of its own.

    One other thing to check is whether your output stage itself is oscillating. I see that you have no collector resistors on the 2N2222A and PN2907, so at that point in the circuit, there might be a pole and an accompanying substantial phase shift at a low frequency.
     
  16. kyle7119

    Thread Starter Member

    Feb 11, 2011
    71
    17
    Yeah, those pictures were from earlier. Here are some with the correct op amp and decoupling caps.
     
  17. kyle7119

    Thread Starter Member

    Feb 11, 2011
    71
    17
    I have some 4700's sitting on the power supply rails (see photos).


    I'm a little confused what you mean by this...
     
  18. ericgibbs

    AAC Fanatic!

    Jan 29, 2010
    2,503
    380
    That method of adding decoupling will make matters worse... mini antenna's
     
  19. kyle7119

    Thread Starter Member

    Feb 11, 2011
    71
    17
    I have verified both stages independently. The power stage works perfectly when driven by the function generator. The amplifier stage also works when disconnected. The oscillation only appears when you hook them together.
     
  20. kyle7119

    Thread Starter Member

    Feb 11, 2011
    71
    17
    Okay, I'll shorten those.
     
Loading...