This is a major mistake: Even if you sample the data many many times a second, choose an actualisation rate of display of no more than about three actualisations per second. Faster actualisations cannot be read and interpreted by the human eye (brain), are totally useless and look heavily amateurish!

we discussed language in another thread, this "actualisations per second" had me stunned for a fraction of a second.

So, for the benefit of the less fluent in English, let me translate:

What Kai means is "do not update the display more than about 3 times a second.

Now for a 'trick' I have employed to get a REALLY smooth display:
If the new value is not at least 5 (your value may differ) different, do not update UNLESS 8 consequtive "update attpempts (1/3 second)" have shown that the value to be displayed is CONSTANTLY to the same side of the displayed value. If just one of the se readings is to "the other side" leave the display as is. In this method, you may have to add a "throw wild readings away" if you have spurious noise.

This method will lead to a display thet react fast to big changes and slow to very small changes which give a very pleasing effect.

EDITORIAL:
It seems that there is a general tendency to believe that an A/D is a "throw it in" component, nothing could be farther from the truth. For any analog, the game is layout, layout and LAYOUT. I once was involved with designing scales and the analog guys (some of whom had been doing this for decades) did, I believe, 28 layouts and tested them before they found one good enough and this was for economy scales selling for $100 -up not even high precision laboratory equipment.

Erik