Dropping bits will not prevent flickering, even not when integrating and decimating! And adding noise is of no help either:

Assume an 8bit-ADC with a static signal superimposed by 1LSB noise at its input. This is a typical situation when using a microcontroller containing a built-in ADC. Now assume that the input signal is in that "window" that gives an output signal of "128". But due to the 1LSB noise the output data stream isn't all "128", but contains a few "127" bytes. Assume two such byte streams now, one containing 247 * "128" + 9 * "127" and the other 248 * "128" + 8 * "127". This yields:

First byte stream: 247 * "128" + 9 * "127"

Summed up over 256 samples and divided by 256:

0111 1111. 1111 0111

Rounded to 12bits:

0111 1111. 1111

Dropping the lowest two bits:

0111 1111. 11
=============


Second byte stream: 248 * "128" + 8 * "127"

Summed up over 256 samples and divided by 256:

0111 1111. 1111 1000

Rounded to 12bits:

1000 0000. 0000

Dropping the lowest two bits:

1000 0000. 00
=============

Again flickering! And, no, this has nothing to do with rounding. I can always fabricate an example, where dropping any bits finally results in flickering.


Kai


PS.: By the way, by oversampling you can only increase the resolution, but not the accuracy and precision!