How to Increase the Resolution of analogRead()! (from 10-bits up to 21-bits)

hello,

i wrote library 21-bits of precision (user-specified 10 21 bits) when reading in analog voltages arduino's built-in adc (analog digital converter).  this useful when need precision reading in analog sensors, thought i'd post link library here.  

realistically, based on limitations of oversampling, & fact single 21-bit reading takes ~500 seconds (8.33 minutes--refer table @ link below), i'd stick 16-bit resolution or less (where single 16-bit reading takes ~0.5 sec).  however, specify "up 21-bit precision" because library accepts precision value of 21-bit without breaking.  i'm not saying adc give reliable 21-bit result, i'm saying won't break code.  having said that, think 16-bit precision results returned library valid, , have watched , compared them other commanded precisions while seeing them read potentiometer.   doing in example codes can run come library when download it.  i recommend run examples see yourself.

additionally, library permits data smoothing sampling many times (a user-specified amount), returning average of samples, @ user-specified sampling precision.  

here is:  http://electricrcaircraftguy.blogspot.com/2014/05/using-arduino-unos-built-in-16-bit-adc.html

happy coding!  i hope find useful.  
i love arduino!

sincerely,
gabriel


reference:
if click on link above, see explanation how doing this.  you see reference, namely, that:  the basis doing atmel application note "avr121 enhancing adc resolution oversampling" (found here: http://www.atmel.com/images/doc8003.pdf).    please don't respond or make comments until have informed concept reading paper, , visited first link above, having read article on website, , downloaded & tested library.

your theory misconceived.

the a/d sampling on arduino chip can detect 0-5 volts in 1023 steps.   in round numbers 5 mv per step.

if feed 5 mv chip,  a/d answer   1
if feed 10 mv chip,  a/d answer 2      , on.

if feed 7.5 mv chip,   you maybe 1,  , 2.   and if bunch of times,  , 1 , 2 bunch of times,    can averaging , average outcome of 1.5,    , maybe can conclude voltage somewhere between 5 mv , 10 mv,   so let's call 7.5 mv.

so,  maybe,   you can 1 bit of implied precision.

but here theory falls down in twisted wreck.

if input signal 6.25 millivolts,    theory depend on getting a/d result 1,   75% of time,   and a/d result 2,   25% of time.   you then,   supposedly,  calculate "average" a/d outcome of 1.25,   and interpolate result of 6.25 mv  in between 5 mv , 10 mv.

the problem is,   the a/d conversion not work way.  response of device input voltage,  looks staircase.   for input voltage range,  between 3 , 7 mv,   you answer 1.   for input voltage range between 8 , 12 mv,   you answer 2.  between 13 , 17 mv,  3.  , on.

these flat parts of staircase.    in small region -   which might between 7 , 8 mv,  or might not be,   you uncertain answer,  either 1 or 2.   that's concept might work,   for 1 bit.   but place work.

for constant input in flat step regions,   you same reason.    if not close vertical part of step ( , don't know ),    won't different measurement outcomes.   if take bunch of measurements 5.1 mv input,  you'll outcome 1.   if repeat measurement 5.2 mv,   you'll outcome 1.    , @ 5.3 mv.   and @ 6.3 mv.      , anywhere else,   up transition region of unknown location , extent,   where device outcome jump 2.     

you cannot distinguish a/d readings 21 bits of precision,   as ludicrously claim.

Arduino Forum > Using Arduino > Sensors > How to Increase the Resolution of analogRead()! (from 10-bits up to 21-bits)

arduino