Pls explain why C1 and C2 are needed?

Discussion in 'General Electronics Chat' started by JimG, May 29, 2010.

  1. JimG

    Thread Starter Member

    Dec 7, 2009
    53
    0
    This simple amplifier doesn't work without C1 and C2, and I don't understand why. No problems show up on my LTSpice simulations without them.

    But without C1 and C2 in the breadboarded version, the output from the LT1007 is unstable, and experiences large swings when I touch anything connected to the circuit (the sensor, the DMM, the trim pot, etc).

    Everything works as expected when C1 and C2 are added.

    Just FYI, the sensor outputs 0-100mV. The purpose of the circuit is to provide an adjustable gain of around 2 to enable use of a PM-128, 200mV panel meter.

    Thanks in advance.

    Jim
     
  2. R!f@@

    AAC Fanatic!

    Apr 2, 2009
    8,750
    759
    to prevent output oscillation from induced voltages
    for LT spice, u need to account for real world interference
     
    JimG likes this.
  3. Ghar

    Active Member

    Mar 8, 2010
    655
    72
    The LT1007 is only stable for closed loop gains of 5 or higher, it's in the datasheet.
     
  4. JimG

    Thread Starter Member

    Dec 7, 2009
    53
    0
    So the capacitors are filtering the oscillations before they get to the amp? Is that the idea?

    Jim
     
  5. JimG

    Thread Starter Member

    Dec 7, 2009
    53
    0
    I can't determine if that note applies to the LT1037 only, or to both the LT1007 and the LT1037? Both amps are covered by the same data sheet.

    Assuming that the LT1007 is only stable above 5, though, how does the addition of the capacitors apparently "fix" the problem at lower gains?

    Thanks for helping me.

    Jim
     
  6. Ghar

    Active Member

    Mar 8, 2010
    655
    72
    An opamp becomes unstable at a 'high' gain at 'high' frequency.
    Specifically you need a gain greater than 1 when your phase shift is more negative than -180 degrees for instability.

    Adding those capacitors adds a pole reducing your high frequency gain allowing stability.

    Here's a plot with (red curve) and without (green curve) the capacitors... I didn't have a model of the LT1007 so it's not totally accurate but the general idea is there.

    caps.png

    Edit:

    I should mention that this dependence only exists because of the non-zero output resistance of your voltage divider (Vbias). Make those resistors smaller or add a buffer and this no longer happens.
    The reason is that the voltage divider (with capacitors) appears in series with the gain setting resistors.
     
    Last edited: May 29, 2010
    JimG likes this.
  7. JimG

    Thread Starter Member

    Dec 7, 2009
    53
    0
    Ghar -

    Thanks for taking the time to explain. If you can bear with me a little longer, can you tell me if its possible to simulate that behavior in LTSpice?

    Jim
     
  8. SgtWookie

    Expert

    Jul 17, 2007
    22,182
    1,728
    Jim,
    LTSpice is a great tool, but you have to realize that it is only an approximation of what real-world components will do.

    You have to read the datasheets carefully when designing your circuit. If you attempt to operate a component outside of it's specified parameters, your results may very widely.

    SPICE simulations are just a starting point. You can develop a circuit using SPICE, but once you've slogged through a number of variations, you need to actually build the thing and test it.

    Any simulation is only as good as the sum of it's parts. For example, the wires in LTSpice are perfect; zero Ohms, no capacitance, no inductance. They are the unattainably perfect wires. A real wire or trace will have an inductance of about 15nH per 10mm at 10MHz, or about 450nH per foot - when the wire is straight. This adds up quickly, and can wreak havoc on circuit performance. Then there is parasitic capacitance, which is a whole 'nother can of worms.

    The better your schematic model is, the better it will approximate the performance of actual components.

    I suggest that you might have been better off starting with an LT1006 than an LT1007.

    I also suggest that your 100k resistors for the voltage divider are extremely high in value. Resistors create electrical noise, rather like when a garden hose nozzle is mostly closed and water is spraying out of it. Putting the nozzle in a bucket of water gets rid of the noise.

    That's what the caps do. They are the "water buckets" of the electronic world.

    Metal film (usually blue in color) resistors will have much lower noise than carbon or carbon film capacitors, and will be much more stable in their values over temperature.
    A simple 10nF metal poly or ceramic cap from the junction of the resistors to ground may be all that you need.

    I usually like to have around 0.1mA to 2mA current across a voltage divider. In your case, replacing R1 and R2 with 4.3k or 4.7k resistors and a 10nF cap to ground will probably effect a similar remedy, but achieve initial stability much more quickly.
     
    JimG likes this.
  9. JimG

    Thread Starter Member

    Dec 7, 2009
    53
    0
    I will try 4.7K resistors in the divider.

    Do you think the resistors in the negative feedback loop are also too large?

    The breadboarded circuit is actually working well as drawn. But I'm at a point where changes/improvements can be easily made.

    Jim
     
  10. SgtWookie

    Expert

    Jul 17, 2007
    22,182
    1,728
    Jim,
    Your values are reasonable. Try plugging the values I gave you into the simulation and see what you get.

    Upload your LTSpice .ASC file as an attachment if you would like more input.
     
  11. JimG

    Thread Starter Member

    Dec 7, 2009
    53
    0
    File attached.

    Jim
     
  12. SgtWookie

    Expert

    Jul 17, 2007
    22,182
    1,728
    No major changes.

    I simply moved the caps around a bit, and changed their values.

    You must always have at least one 0.1uF (100nF) bypass capacitor, either metal poly or ceramic, across the power pins of any IC. If you omit them, you will very likely have odd problems that won't show up in a simulation. Your C1 and C2 were providing the equivalent of an 0.5uF bypass cap, depending on how close they were to the opamp's supply pins.

    You could still use two caps for Vbias if you'd like; that will help to establish it more quickly when you first power up the circuit. I reduced the value of C1; just about any value between 1nF and .1uF should be sufficient to keep it quiet. The larger the value of capacitors, the larger they get physically.

    Metal film resistors generate less noise than carbon or carbon film resistors. They are also more accurate over temperature.

    I extended the Vbias trace a bit, to make it a tad more obvious that one needs to reference the output signal to it.
     
  13. JimG

    Thread Starter Member

    Dec 7, 2009
    53
    0
    Excellent.

    I have breadboarded the revised configuration and all is behaving well. Differences between the current version and the previous version aren't obvious on the output device, but I feel a lot better about my understanding of the circuit.

    Thanks again.

    Jim

    PS - I used 5.6K resistors because that's what I had handy, and in good supply.
     
Loading...