Op amp to drive 10uF capacitor

The datasheet for the AD5541A DAC (http://www.analog.com/static/imported-files/data_sheets/AD5541A.pdf) recommends using an AD8628 op amp (http://www.analog.com/static/imported-files/data_sheets/AD8628_8629_8630.pdf) to buffer the reference and drive a 10uF capacitor to ground. However, this op amp is very noisy for my application (frequencies up to around 100kHz are important). Does anyone know of a better alternative that is still capable of driving such a large capacitor?

  • 0
    •  Analog Employees 
    on Sep 30, 2014 9:58 AM

    Hi James,

    I am the applications engineer supporting the AD5541A. I can't recommend a suitable amplifier, (though someone on this forum should be able to) but I can give some pointers.

    The AD5541A is an R-2R DAC architecture, which means that when you change codes the amount of current taken in through the VREF pin varies. The 10uF capacitor is there as a charge reservoir so that the dynamic Vref currents don't disturb the reference voltage. If your low noise amplifier has sufficient drive capability you can probably reduce the capacitance value needed.

  • OK, thanks - I was planning to use the ADA4004 amplifier (http://www.analog.com/static/imported-files/data_sheets/ADA4004-1_4004-2_4004-4.pdf), which I guess could conveniently produce an output current of up to around 1mA. What capacitance would be needed for this?

  • A common solution is to place a series resistor between the opamp output and the capacitor, typically around 100 ohms. You then take two feedback paths back to the inverting opamp input, a capacitor direct from the opamp output and a resistor from the 10uF load capacitor. This approach lets you use pretty much any opamp you like, with the basic trade off that the output slew rate is reduced (unlikely to be a concern for a DAC reference voltage)

  • Thanks for the tip - what values would you recommend for the resistor and the capacitor on the two feedback paths?

  • 0
    •  Analog Employees 
    on Oct 7, 2014 10:44 PM

    Hi James,

    The op-amp isn't really "driving" the capacitor in the linear sense, i.e., it will not have a suitable linear response to an input such as a step, but is simply providing a DC voltage source with low source resistance.  Just because an op-amp has a large capacitor on its output does not mean that it is actually driving a heavy cap load.  Normally, load C is problematic due to the lagging phase shift produced by the pole that is introduced by the amplifier's output resistance and the load C.  The lagging phase shift diminishes phase margin and can cause an otherwise stable amplifier to become unstable.  This is true, as long as the op-amp is behaving like a linear op-amp

    Putting 10 uF on the output of many op-amps will cause the open-loop magnitude response to drop considerably (though in a nonlinear fashion), moving the 0 dB crossover frequency back to where the open-loop phase shift is much less.  This is similar to what happens when one makes an amplifier with very high gain, but this nonlinear effect gives you the benefit of high gain without having to have high gain.  You can run the amplifier at G = 1 and still be stable.  The AD8031 is a popular amplifier for this application, and is seen used this way in many data sheets.  I've analyzed open-loop simulation plots of the AD8031 with 10 uF || 0.1 uF Load C, and the results show close to 80 degrees of "phase margin," when one includes the parasitic elements in the capacitors.  The capacitor ESL actually helps, since the 10 uF capacitor becomes inductive above its SRF, providing phase lead.  I used quotes around "phase margin" because the amplifier is not operating linearly anymore, but it is still very stable.

    I recommend that you contact the applications engineer that supports the op-amp you want to use this way.  They can then check with the designer to be sure that the part will not be damaged in any way with a 10 uF load.  You can also test the op-amp you want to use by placing a 10 uF capacitor on its output, set up with your desired gain, and apply a fast step to the input.  The response should be very sluggish, somewhat exponential-looking, and may have regions that look nonlinear.  As long as you see no oscillatory behavior, you should be okay from a stability standpoint.

    This approach is quite useful for providing a DC voltage source with broadband low source resistance, all the way down to DC.  Once you have the 10 uF in place, you can add smaller capacitors in parallel to keep the source impedance low as the larger-valued capacitors pass through self-resonance.

    Best regards.