8

Does it ever make sense to have more precision than accuracy? E.g. if you had a variable resistor which you could control with a resolution of 0.01 ohms but your multimeter only had 1% accuracy, then that extra resolution would be pretty useless wouldn't it?

Roh
  • 4,598
  • 6
  • 41
  • 86
jrandj
  • 105
  • 3

3 Answers3

10

Absolutely. There are many applications in which you're much more interested in the relative changes in a signal than it's absolute magnitude.

Take digital audio, for example. It is the precision of the converters that gives you the high signal-to-noise ratio you're interested in. The absolute accuracy (i.e., whether full-scale is 1.0 or 1.1 V) is not generally all that important — it just means the signal overall sounds a little softer or louder, which you can compensate for by adjusting the volume control.

Dave Tweed
  • 168,369
  • 17
  • 228
  • 393
  • measuring parameter variance of high-volume prodution hardware also comes to mind – scld Dec 16 '13 at 13:47
  • There is also a distinction between *measurement* precision and *value* precision. A 12-bit ADC provides 12 bits of value precision by definition but might provide less measurement precision. Deviation of a value from reality may be more or less systematic (with constant offset being a simple case) with the system being more or less easily modeled (randomness and limited measurement precision related to quantum mechanics being an extreme). –  Dec 17 '13 at 23:07
  • 1
    @PaulA.Clayton: I think what you're calling "value precision", I would simply call "resolution". But we're really getting into hair-splitting territory here. In terms of the actual performance of the application, it's the "measurement precision" that's relevant. – Dave Tweed Dec 17 '13 at 23:20
1

Yes it does, because when calibrating it allows you to measure the error with accuracy equal to the precision.

geometrikal
  • 4,921
  • 2
  • 23
  • 41
  • Exactly. Accuracy is typically just how a component is specified... with characterization and calibration you can do sometimes *much* better. – darron Dec 17 '13 at 14:42
0

Errors add up. If I measure a circuit in production at 1%, and it tells me that I need 1R, I'd like to be able to set 1R, and be within 1% of correct value. If I set 1R +/- 1%, I'm within 2% of correct value.

david
  • 4,364
  • 15
  • 30