I have the choice between 3 different Leaded Inductors and wonder if it makes a difference which
one i take for AVCC to 5V.
There are difference in SFR, resistance and quality
L-MICC 10µ
SMCC 10µ
L-HBCC 10µ
OK, the "Q" is much the same for each and you actually do not need a high "Q" for decoupling inductors.
Similarly, neither current rating nor resistance is of any concern for this application. Use the one with the highest resistance as a little resistance will not affect voltage measurement but will actually improve decoupling.
Really, they are virtually identical.
It is advised to use it for maximum accurate ADC samples. That is, when every sample has to be as accurate as possible.
However, with slow things (temperature, battery voltage, light intensity) it is not needed. Using the average of a number of samples will improve the result more.
Koepel:
It is advised to use it for maximum accurate ADC samples. That is, when every sample has to be as accurate as possible.
However, with slow things (temperature, battery voltage, light intensity) it is not needed. Using the average of a number of samples will improve the result more.
Well that's not actually true, the 328's ADC is good enough to return repeatably identical readings
from sample to sample (admittedly on battery power), so to improve the results you need to inject
analog noise into the system. Have a look at page 12 of this guide: http://www.microsemi.com/document-portal/doc_view/131569-improving-adc-results-white-paper
MarkT, that seems to be more theory than practical. As far as I know, I have some noise in every Arduino project. With or without the LC filter on AVCC. By averaging a lot, more than 10 bits is possible. Although those bits are not accurate (due to temperature influences and perhaps offset and linearity and so on), they can only be used to indicate that the signal is going up or down a little.
What about a sensor with a wheatstone bridge output ? Should the +5V of the sensor be connected to VCC or AVCC ?
I opened a schematic of one of my projects, and I have it connected to VCC. I think that is wrong.
Koepel:
MarkT, that seems to be more theory than practical.
Not just theory but practical as well, although the noise has to be truly random for it to give good results.
Russell.
Koepel:
MarkT, that seems to be more theory than practical
Standard technique in digital signal processing, not just theory.
Yes, this technique is in regular use. It requires either uncorrelated noise or a correlated ramp or triangle wave imposed on either the signal or the reference. Uncorrelated noise is often simpler to add as there is very often already noise present.
Usually called either oversampling or dithering.
http://www.atmel.com/images/doc8003.pdf
http://www.st.com/web/en/resource/technical/document/application_note/CD00177113.pdf
From Wikipedia on the meaning of "theory" in a scientific sense:
A scientific theory is a well-substantiated explanation of some aspect of the natural world, based on a body of facts that have been repeatedly confirmed through observation and experiment. Such fact-supported theories are not "guesses" but reliable accounts of the real world.
So the inductor on AVCC only has influence on ADC?
It does not help the ATmega overall? Less noise overall, more stable...
Paul__B:
Use the one with the highest resistance as a little resistance will not affect voltage measurement but will actually improve decoupling.
Sorry i guess my English is good but isn't that sentence a caveat?
Then why chose the high resistance if a LITTLE resistance will not affect voltage measurement but will actually improve decoupling?
A high Q (low resistance) inductor may ring, actually boosting certain frequencies. In fact, ferrite meant for things like ferrite beads are made of higher loss ferrite. In other words, the ferrite material itself is somewhat conductive and so causes losses through eddy current losses.
Yes, the inductor on AVCC only affects the ADC. You really don't want to use inductors for decoupling on the rest of the power.
Inductors work to reduce interference by slowing changes in current. That's fine for a circuit that generally draws the same amount of current all the time. But the rest of the circuitry in the Arduino and around it is going to draw wildly different amounts of current.
Typical ferrite beads and ferrite inductors meant for supply decoupling act more-or-less as resistances
across a wide range of frequencies (MHz -- GHz), because of the lossy nature. They used to be used all
over old-fashioned discrete VHF/UHF circuitry to prevent the risk of parasitic oscillation.
MrGlasspoole:
Then why chose the high resistance if a LITTLE resistance will not affect voltage measurement but will actually improve decoupling?
Because the resistance of any of these inductors, less than one ohm, is really little. 
As AVcc draws relatively little current, you could probably use ten ohms or more in series with no problem.
polymorph:
From Wikipedia on the meaning of "theory" in a scientific sense:
Quite. I think he was confusing theory with hypothesis. A frequent misconception.
Russell.