# Can an instrument ever have an accuracy better than its repeatably?

1. Oct 12, 2013

### rollingstein

I have a dosing pump spec. sheet from Milton Roy (a fairly reputable manufacturer) which says:

Repeatably: +/- 3%

Is this possible? Can an instrument be more accurate than it is repeatable? Intuitively, this didn't make sense to me.

The spec. sheet doesn't say, but I'm assuming that both are expressed as percent of reading or full scale deflection but not a mix.

http://www.miltonroyindia.com/M-Series.pdf (Look at Page 2)

2. Oct 13, 2013

### Simon Bridge

They are different measures.
Check what they mean by "steady state accuracy".

3. Oct 13, 2013

### rollingstein

I think I now understand it (perhaps).

You mean the "accurate but not precise" case?

I cannot find a definition for "Steady State Accuracy" on their site. I took it to mean the offset measured against some better gold standard callibrating instrument without changing flowrates.

I'm still confused. I was under the impression that repeatablity puts a hard upper bound on how accurate one can be.

4. Oct 13, 2013

### Simon Bridge

accurate vs precise - that's right.
but without knowing how the manufacturer is arriving at the values - it's anyone's guess.

5. Oct 13, 2013

### rollingstein

Funnily, I've never seen a spec. sheet of an instrument that defined either of these terms.

So I guess there's a conventionally agreed upon definition?

6. Oct 13, 2013

### Simon Bridge

You have presumably googled them to see?

7. Oct 13, 2013

### Averagesupernova

Well I would assume that steady state means that if you set the instrument to do X then it will continue to do that with an accuracy +/-1%.
-
If the controls are moved and returned to the original position I would assume you can expect it to be within +/- 3% of where it was before you moved it. In other words, with calibration it can always get results of +/-1% repeatable. I look at as a way the manufacturer made the accuracy look better than it is and still avoid liability issues. I used to work for a test equipment manufacturer and I know the games they like to play when it comes to specs.

8. Oct 13, 2013

### rollingstein

Hmm...I interpreted it to mean for a actual flow of say 100 if many readings are taken it may read 101, 102, 103, 99, 98, 97.

The problem with such an interpretation is that unless they specify how many readings they average over, it's meaningless.

9. Oct 13, 2013

### Averagesupernova

Specs are always meaningless unless they are defined.

10. Oct 13, 2013

### DaveGun

I love the Target example. I'm going to apply my answer in the regard of taking measurements (and not the operation of this pump). It explains the relationship between precision (inherent in the sensor) and accuracy (the true ability to give a correct reading of the condition from the sensor). Accuracy is normally achieved using calibration for the sensor for a known condition.

For static conditions the Target example is correct. However under actual conditions where the measured medium is moving - you have to think of the Target as moving too (and your measurements as the bullets fired at the target).

Taking a measurement is essentially taking a snapshot of a condition in a given time frame. A single reading cannot accurately portray the dynamic condition that may exist at a site. The instrument can be precise, the instrument can be accurate - but without multiple measurements you cannot actually determine the true nature of the condition you are measuring.

Now, on to this Pump. As stated, it is more accurate under static (stable) conditions. Under varying conditions it may have inaccuracies due to hysteresis. That would be my guess for this situation.

My first entry here. Hope it was helpful.

Dave