channel change bug in ADC_vect

[HamishB]

Hi,

a patch by trevorshannon exists in the sparkfun comments, after years of being known the bug is still present (but unreported :) so it would be nice to have it finally fixed.

https://www.sparkfun.com/products/10537

quote:

i didn’t see enough noise to make this ‘unusable’ but i did see some strange jumps every once in a while. for example, for one sample, x would increase a lot while z would simultaneously decrease a lot. suspicious…

taking into account what others have already mentioned, i made the following change to the firmware (sorry for pseudo code) and i seem to have eliminated those random jumps.

ISR(ADC_vect){
 cli();
 if (!channelChanged){
  do the normal ADC_vect
  channelChanged = true;
 }
 else{
  store ADCL and ADCH in a temporary variable
  channelChanged = false;
 }
 sei();
}

someone else added:

I managed to do something similar to this, but I don’t understand why the ADCL and ADCH are stored in a temporary variable in the “else” block. Are you using that temporary variable somewhere in the code? Shouldn’t the values be discarded? That’s what I did and the strange jumps are gone.