We have designed a custom prototype board based on the Arduino Mega (ATmega2560). Our knowledge on electronics are limited so we assigned someone else to design the board.

On his design he used the following circuit for noise filtering and decoupling the Voltage on the AVCC and AREF pins of the 2560:


My problem is that on the Arduino Mega R3 reference design the decoupling is done this way:

Atmega R3 decoupling

So there is a difference that on the AVCC there is no coil filtering (L4) and that they are using 3 100nF decoupling capacitors, while hs is using 1 100nF and 1 10nF Also on the AREF decoupling is done with 1 100nF on the Reference design, while he is using 100nF and 1nF

When asked he said that his approach is better, but since my knowledge is limited on the field, I wanted to make sure we are definitely doing this correctly.

Here is the Reset circuit

enter image description here

  • 3
    \$\begingroup\$ Arduino devices are an example of... adequate... engineering. One can always do better than them. \$\endgroup\$ Commented Oct 17, 2017 at 9:49
  • 1
    \$\begingroup\$ Theoretically desifn with coil filters more noise. You can also use ferrite beans as well. \$\endgroup\$ Commented Oct 17, 2017 at 9:49

1 Answer 1


Well in theory, with decoupling, more is usually better (although at some point you get diminishing returns). This actually involves two different issues, Decoupling and Bypassing. You can find out more about them here, but basically Bypassing is about getting rid of noise from the supply, and Decoupling is about being able to provide short bursts of current. Technically L4, C10 and C13 are more dealing with Bypassing, while C12 and C11 are Decoupling.

In the reference design, those capacitors are more meant for Decoupling, although of course it has an effect on Bypassing as well. The type of capacitor isn't specified, but it should typically be ceramic.

The decoupling for the two designs is basically the same, so the difference is the Bypassing. Using a 10uF capacitor instead of a 100nF capacitor will definitely be a bit better, but using an inductor as well will be much better. It turns the RC filter formed by the trace and the capacitor into an LC filter, which is a second order filter and has much higher attenuation at higher frequencies.

I don't think theres a question of whether your design is higher performance (it is higher performance), but there is a question of whether it's worth the extra cost and board space, and only you can really answer this as it depends on how precise you need your ADC readings, and what environment your board is in etc.

  • \$\begingroup\$ I don't mind have a better decoupling/bypassing design (cost isn't an issue here). My concern is if this is actually working as it should as we had a few resets on the Atmega when testing this (might be from noise on the digital VCC though, since we also have a Sim800 on the pcb) \$\endgroup\$
    – Lefteris
    Commented Oct 17, 2017 at 9:59
  • \$\begingroup\$ Would it be better to replace the Coil with a Ferrite as @PeterJ_01 said? \$\endgroup\$
    – Lefteris
    Commented Oct 17, 2017 at 10:00
  • \$\begingroup\$ @Lefteris I would be surprised if LC bypass on the analog supply pin was causing a reset, if it was on the digital power pins perhaps. You can always short out the L and see if it helps though. \$\endgroup\$
    – Trevor_G
    Commented Oct 17, 2017 at 10:04
  • 2
    \$\begingroup\$ I would be very surprised if you were getting resets due to this, have you checked your power supply on a scope? What is your reset pin connected to, and are you in an extremely high emi environment (driving motors for example)? Technically noise on your Analog supply wouldn't (I think) cause resets, but of course they are internally connected. I don't know too much about the benefits of beads vs coils, so you might want to research that, but I do know that if choosing a bead fixes your problem, it's going to be an extremely marginal fix and you need to find the real cause. \$\endgroup\$
    – BeB00
    Commented Oct 17, 2017 at 10:05

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Not the answer you're looking for? Browse other questions tagged or ask your own question.