VFB as a voltage reference

Hi -

I am using an ADA4505 VFB as a voltage reference to obtain Vdd/2 (Vdd = 3.3V) as a midpoint reference in an active filter circuit, using a pair of 1% 10K resistors at the NI input. the 4505 is being supplied by an LDO, which is preceded by a switcher supply.

My question is, how stable would this VFB output be for a comparator? The comparator requires a threshold of around 200mV, so I would need to take the VFB output and divide it down to feed Vref of the non-inverting comparator.

My thinking is, perhaps the instability of the 3 different devices (switcher, LDO, VFB) plus the resistor divider may contribute a lot of instability so my Vref at comparator could vary by some conjugate of those numbers that could end up too low or high. I can tolerate around 50mV on the threshold.

So...bad idea? Should I use a dedicated voltage reference?

Thanks for your help!

  • Groger,

      Stability is really a function of the output impedance and the capacitive load.

    In general, the lower the power/supply current, the wimpier the output, so 50 pF

    can cause oscillation.  Unfortunately, there is no standardized parameter and test

    conditions, so there is nothing in the search table for you to choose.  But a lot of

    data sheets have a graph of overshoot vs. CLoad, so that gives you a clue.  

    You don't need a RRI/O op amp, and I would suggest using more supply current.

    This will give you lower output impedance.

    So also look at:  AD8603, AD8657, and LT6000. 

    What is the max supply current you can allocate? 

    Is this battery powered?

    What LDO??

    Note that ripple or noise on the 3.3V will be divided in half by the 10k/10k, and go straight

    through the voltage follower, so you need a cap at the midpoint of the two resistors to ground.

    Say 1u-47uF.

    Depending on the input impedance of the comparator, and whether or not it changes, you might

    need a cap on it's input. (inverting or non-inverting config??;  what part number?)

    I hope you are not using an op amp as a comparator;  bad!  See:

    The Maximum Supply Current That Wasn't | Analog Devices 

    and the references at the end of the article.

    If you need some capacitance on the "output" of the op amp, don't put it on directly;

    see:   Op Amps Driving Capacitive Loads | Analog Devices 


  • hi,

    Thanks for the detailed reply. To be specific:

    No, not using the 4505 as a comparator.

    It's a battery powered app, so I cannot/do not want to increase current. Total circuit consumption will be in the 4 to 5mA range. So, I understand the part about wimpy. 8)

    I think you gave me enough to realize I need to use a proper voltage reference for the comparator to insure correct threshold triggering.

    However, your good reply gave rise to another question. Essentially, I am using the prescribed active filter supply reference as seen in the AD filter wizard. This takes a VFB and uses it in the way I described to get Vo at midpoint for the ADC. (using single ended input on uController)

    Does your response mean I am better off to use a proper voltage reference as opposed to the one described above? The entire active filter circuit (using ADA4096) may consume under a few mA, not including the Vref bias currents - which I assume are pretty low.


  • Gary,

      Which comparator?

    The Filter Wizard is really good;  I know the guy that wrote it.

    I would have to see the full design to comment further, but I would guess they use an AD8031 which is a high current

    bipolar that will drive a Tesla motor......

    What is the frequency of your filter?  You might save some current with something other than the 4096.

    Also, the industry makes a lot more types of duals than quads, so don't use the 4096-4.  Two duals is

    a lot easier to lay out, especially with active filters that have lots of R's and C's around them.  The additional

    space and money is minimal and the performance is better.  I  would look at the LT6078 and the LT6020.

    What threw me off is your dual use of the midpoint.  For the AF, and also for the comp ref.

    If the comparator input is well behaved, you could just divide down again from the 3.3V/2 and again, put

    a small cap on the 200 mV node.


  • harry -

    Filters are centered around 100Hz - an audio application. For comparators, I've been looking at the ADCMP370, as it's got a very low Iq. (no bad pun here) For opamps, your suggestion of duals vs quads makes a lot of sense, in terms of space efficiency.

    So, please clarify - does it make sense to use a voltage reference for the active filter (at Vdd/2), and also use that same reference and divide that down for the comparator? Or, should I use the wizard's recommended VFB (substituted with a 4505 for power saving) and divide that down for the comparator? It's a 4 amp filter design. The ADA4505 has 40mA drive, that should be more than adequate. It does have a better Iq than the 8031.

    also - great read on the link, re: VFB and current drive and stability.

    thanks again!

  • If you divide down the Vdd/2, then the comparator is isolated, so shouldn't be a problem.

    The ADA4505 does NOT have 40 mA drive;  that is short circuit current.  See fig 22.