AD8618 supply decoupling


I'm designing a PCB with four AD8618, single supply +3V and all channels as voltage followers (see attached figure, U3-U6). U1 is the voltage regulator in the circuit, and the output is decoupled with a 4.7uF tantalum cap (C2). Now my question is whether I actually need the decoupling caps on the op-amps (C5-C12). I'm using two decoupling caps per op-amp, one 0.1uF ceramic and one 10uF tantalum. The PCB traces between U1 and U4 (as well as U1 and U5) is approx. 4 cm. The traces between U1 and U3 (as well as U6) is approx. 10 cm.

If you please have some ideas on this I would be very grateful!

Thanks in advance,


  • Hi Harry,

    Thank you for your reply. Here I attach a schematic of the circuit.

    The purpose of this circuit is to interface a microphone array (whose outputs are TRRS) with an acquisition system (whose inputs are BNC). We plan to have 64 channels, hence we will actually need 4 of these boards. The choice of number of channels per board comes from manufacturing/assembly costs. The use of quads also lowers these costs, and their purpose here is to match the impedance from the array and drive the BNC cables. The long traces are difficult to avoid unless we modify the amount of channels per board. Do you have any particular insights or ideas about long traces?

    Many thanks in advance


  • Elias,

      Yes.  Even if you signals are low frequency, the AD8618 is 20 MHz gbw so you need to make sure that

    the 3V is really AC ground.   I am concerned about some of your long traces.  I'm not a big fan of quads,

    one reason being pc board layout problems.  See:

    Another factor is capacitive loading.  Can you supply a schematic and tell me a little more about what you are doing?


  • Elias,

      What does TRRS mean??

      What is your highest signal frequency?

      Depending on whether you need to go to DC or not and can AC couple, you could use op amps

    that have millivolts of offset, rather than <100 uV which drives up the price.

      Also, if your highest frequency is in the tens of kiloHertz range, you could use a lower gbw which

    will also reduce the possibility of oscillations.  Faster is not always better.

    Based on that, you could use what I call some "non-precision" op amps, such as the AD8592,

    AD8532, AD8602.  1k price less than $0.50 USD.

    Other possibilities would be AD8692, AD8646. AD8617, AD8604, etc.


  • Harry,

    Thank you again for your message. TRRS means the 4-conductor audio cable, stereo + mic (like the one used in smartphones).

    The highest freq. is around 10k, and the signals are DC coupled at 1.5V via a reference voltage.

    I believe I will go for the AD8604, since the footprint is the same and it is more than half the price of the AD8618.

    Please let me know if you have any additional comments on this.



  • Harry,

    Many thanks for your reply. At the moment we plan to build 4 of these boards. We might order one first and test the overshoot and oscillatory behaviour as you suggest. In any case, how could this behaviour be lessened?

    Regarding the GBP, should the overshoot be larger when the desired bandwidth is much smaller than the GBP? i.e. if I use the AD8618 I should expect more oscillations than if I use the AD8604?