The data sheet of the ATmega48A/PA/88A/PA/168A/PA/328/P states that ADC input channel 5 and 4 use digital power, while ADC input channels 3-0 use analog power.

Does this mean that ADC 5 and 4 use Vcc and ADC 3 to 0 use AVcc?

Are ADC 3 to 0 then less subject to digital noise than ADC 5 and 4?

And if Vcc and AVcc share the same power source (which according to Atmel AVR126 should be done via an LC network), does it still matter which ADC input channel is used?

  • \$\begingroup\$ Any particular ATmega? \$\endgroup\$ – Ignacio Vazquez-Abrams Jun 6 '15 at 21:14
  • \$\begingroup\$ Sorry, the ATmega48A/PA/88A/PA/168A/PA/328/P, I updated my question \$\endgroup\$ – Torsten Römer Jun 6 '15 at 21:17

If the device specifies it uses DVCC (Digital) for a channel, that would mean that its initial stages are powered by the noisy digital power. So some noise may/will couple into your signals there, even more so if your signals are particularly weak/high-impedance.

Coupling the DVCC and AVCC through an LC network is exactly the right way to keep the Digital noise out of AVCC, so yes, it could still matter. The inductance the datasheet advises keeps high frequency switching noise out of AVCC and the capacitor helps that along, by shorting away those noisy peaks and also storing some "spare" power for the AVCC domain.

So, if you filter your AVCC domain with an LC network and then use the DVCC powered channels, the first thing you should check is if those signals are relatively strong. This will avoid a lot of the noise, since the ADC presumably still works on AVCC. And then try to avoid them fro signals that then would still be sensitive to 1 or 2 LSB of digital noise.

That is, if digital noise matters at all to you, there are some possible applications that would not care about digital noise all that much, but that's down to you, I suppose. It might even be awesome to have a little digital noise for free, if you want to use oversampling and decimation, but that's a WHOLE other chapter.

  • 1
    \$\begingroup\$ Actually, it's AVR121. \$\endgroup\$ – Ignacio Vazquez-Abrams Jun 7 '15 at 0:01
  • \$\begingroup\$ @IgnacioVazquez-Abrams :-D. Still, pretty much another chapter, though. \$\endgroup\$ – Asmyldof Jun 7 '15 at 0:19
  • \$\begingroup\$ @Asmyldof Nice explanation, they answer my questions, thanks! Actually, I am doing oversampling and I think there at least is coming quite some analog noise in through the long unbalanced sensor cables. \$\endgroup\$ – Torsten Römer Jun 7 '15 at 1:09
  • 1
    \$\begingroup\$ @TorstenRömer It's possible you might not really care in that case, but it is best to know about it and be aware of the things it can cause. If you're not sure, usually spending $0.25 on a cap and inductor is the better choice, just for certainty. \$\endgroup\$ – Asmyldof Jun 7 '15 at 1:30

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

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