Measurement geek, here

jpender wrote:A scale error of 0.5g at 500g is 0.1%. If you're weighing out 20g of beans that's an error of 0.02g or about 1/7th of a typical coffee bean.

That's assuming the error is linear. Not necessarily a valid assumption

jpender wrote: That's probably less than the scale precision.

ITYM "resolution?"

Resolution is how finely a thing can measure. E.g.: ±1g, ±0.1g, etc.

Precision is how repeatably a thing can measure. I.e.: Does placing the same weight upon the scale produce the same result every time?

Accuracy is how absolutely correct it is. It can have a resolution of 0.1g and precision enough to repeatably produce the same result within that resolution, but is the value it's presenting really what the thing weighs?

As for the OP's question: That's up to you. Me: I'm kind of OCD with this kind of thing, so I bought a precision calibration weight and tested my new scale. (It was off 0.6g at 200g.) Did it really matter, from a practical standpoint? Not really. But I'm happier knowing the scale's calibrated.

(Btw: Despite the fact it was initially off 0.6g at 200g, it was dead nuts, w/in its 0.1g resolution, at 0.5g and 1.0g. [My 1.0g weight is NIST traceable.])

SEMIJim wrote:Despite the fact it was initially off 0.6g at 200g, it was dead nuts, w/in its 0.1g resolution, at 0.5g and 1.0g.

This is not surprising to me. The linearity of the load cells and compensating electronics I am hoping/assuming are generally pretty reasonable. It is my unproven belief that, with non-lab-grade scales, "calibration" amounts to setting the gain.

0.6 g / 200 g is about 0.003 g/g or about 0.05 g at 18 g

Assuming the linearity is good over the range, measuring an additional 18.0 g at "0" or at 200 g should have the same incremental error (you generally don't care that your basket weights 200.0 or 200.6 g, as you tare it out).

Unless you're working with determining EY% or the like, repeatability is the key. "Nobody cares" if your 18.0 is really 17.9 or 18.1. What matters more is that every time the scale reads "18.0" it is the same weight, within reason.

SEMIJim wrote:That's assuming the error is linear. Not necessarily a valid assumption ;)

What do you think the calibration routine is doing?

SEMIJim wrote:ITYM "resolution?"

I meant precision.

SEMIJim wrote:Btw: Despite the fact it was initially off 0.6g at 200g, it was dead nuts, w/in its 0.1g resolution, at 0.5g and 1.0g. [My 1.0g weight is NIST traceable.])

That's because the error is, to first approximation, a linear scaling error.

Research the weight of coins and get some which are "fresh." Most banks will be accommodating in that regard. Or go to ebay and get some calibration weights.

Marcelnl wrote:most scales do not even have the option to calibrate....

Really? I must be lucky. All of the 8 or so digital scales that I have ever used -- most of them cheap -- had a calibration function. If you don't have the printed instructions, you can probably find it online.

perhaps I'm just buying even cheaper digital scales I have a kitchen scale, no calibration/adjustment possible, a drugs dealer scale for coffee, no calibration/adjustment possible. The kitchen scale needs replacing as it has wonky readouts, every now and then I find myself wondering why dry end takes so long and then I notice post roast that I must have roasted something close to 500g.