[quiff]

Disclaimer – I’m well aware this is both a first world problem, and highly unlikely to affect the quality of my coffee, I’m just interested in principle, particularly views from anyone in metrology / engineering / science fields. 

I was recently given some coffee scales, which are supposedly accurate to 0.1g. (For context, to make an espresso you use just e.g. 18g of beans, so in theory being able to tell the difference between 18g and 18.9g of beans could make a difference to flavour (see disclaimer). The scales are not hugely expensive, but are from a reputable kitchenware brand. 

I usually ‘pour’ beans in, then slow down as I close in on 18g. I noticed that if I poured to 17.Xg and then added beans slowly as I closed in on the magic number, it seemed impossible to get the scles to move in 0.1g increments – I could add say 3 beans before they changed, and then they would jump by 0.2 – 0.3g.

As a test, I then saw how many beans I could add to an empty scale before it registered anything. If I added the beans one, two, three etc at a time, I found it was possible to weigh (what turned out to be) 20+g of beans without the scales registering anything at all. I found that the scales wouldn’t register anything unless I added 7 or more beans at a time. I appreciate it’s variable, but 7 beans seems to weigh about 1g, putting each bean at more than 0.1g.

Are the scales faulty? Or am I misunderstanding what accurate to 0.1g means?      

 

           

Topic

[hawkinspeter]

Backladder wrote:

Its as clear as mud now, thanks 

You need a coffee spectrophotometer or refractometer for that: https://www.beanpoet.com/best-coffee-refractometer/

[mdavidford]

You forgot to include thaum

You forgot to include thaum variance.

[andystow]

We have some seven place

We have some seven place scales here, will weigh tens of grams to 0.0001 mg.

[chrisonabike]

I’m here for the balance (now
I’m here for the balance (now been decades since labs though…)

[ktache]

I love a good balance. We

I love a good balance. We have a couple of little battery ones in my lab, 2.2kg to 1g, and a 220g to 0.1g. there’s a 3 decimal place pan in the lab opposite, old, but good. A 4 figure down the corridor, with a sliding door, but placed on a solvent cabinet, so any movement near causes drift. When I was at Reading , the labs in the tower had weighing rooms between pairs of labs, with big granite blocks built on brick pillars to have incredible stability. Old school but they knew what they were doing back then.

I balance my ultra buckets on a 2 figure. And then check the zero, if I manage to kill an ultra it’s not going to be from bad balancing.

“Salvaged” a wonderful 4 figure from an old storeroom in Reading, motorised rotating door. Utterly lovely, but as with all nice things the mucky pups destroyed it, weighing sticky billowing media on a 4 figure, and with poor technique…

Had many digital scales for baking, currently a Heston salter set, that does 2kg to 1g and 200g to 0.1g,  I have not treated them well but they keep going. Big one does drift. Problem with the little one is finding something light enough in the kitchen to place the small amount of stuff in. Won’t tare high. No weigh boats in my kitchen…

[Backladder]

Its as clear as mud now,

Its as clear as mud now, thanks 

[ktache]

And I just checked they are

And I just checked they are us cup volume, rather than the larger British cup…

[ktache]

I have avoided any US recipe

I have avoided any US recipe that uses cups.

Though I did buy some really nice mason cash ceramic ones, just in case…

[andystow]

Ooh, did someone ask for a

Ooh, did someone ask for a measurement engineer/metrologist?

As others have mentioned, precision and accuracy are not the same. Going further down the rabbit hole, high quality measurement devices will specify at least:

Readability (or precision)
Zero offset
Non-linearity
Non-repeatability
Hysteresis
Temperature drift

The total uncertainty is generally calculated via the RMS (root mean square) of these.

Zero offset can often be ignored if you zero the instrument before use. That’s what “tare” does on a scale.

Non-linearity tells you how far it can wiggle around. For instance, if you calibrated the scale to read within 0.1 g at 10, 20, 30, 40, 50 g, that doesn’t mean it’s within 0.1 g at 25 g.

Non-repeatability would be how much the measurement would vary if you repeatedly made the same measurement in the same way.

Hysteresis is really related to what you’re seeing. The way it’s defined, though, is how much the reading can vary depending on the direction it’s approached from. If I tare the scale, then add a 20 g weight and see 19.8 g, then add another 20 g weight and immediately remove it, I may see 20.2 g, which means the hysteresis at that point is 0.4 g.

Temperature drift is just how much the reading could differ if the scale is not used at calibration conditions, usually 20 °C or 25 °C.

With scales, you also have errors due to buoyancy and local gravity, but those won’t matter for what you’re doing.

I believe what you’re seeing, since the scale has a mechanical component (something is moving a tiny bit) is stiction, but there’s a chance it’s purely electronic. You may be able to eliminate it by giving the scale a little shake before reading it. See how repeatable it is then. We have one model of dynamometer (engine torque measurement) that we calibrate where the procedure says to hit certain spots with a mallet before taking each point.

All this, and we also specify a coverage factor, generally 2σ which means two standard deviations. That gives us a 95% confidence interval, meaning even if I experiment with the scale and do all this math. and tell you the total uncertainty is ±0.35 g, it could still be expected to read outside of that interval 5% of the time.

I hope this brief, non-mathematical summary helps.

[quiff]

mdavidford wrote:

It does have 0.1g precision. It just doesn’t have 0.1g accuracy.

[Edit: should have finished reading through the thread – HP got there before me.]

Thank you both – a distinction I was unaware of. 

[quiff]

<Reported to moderators for offensive language>

[quiff]

Hang on, am I not supposed to

Hang on, am I not supposed to tamp the beans? 

[quiff]

Creakingcrank wrote:

BECAUSE THIS IS AN IMPORTANT ISSUE…

do some blind tasting with slightly different amounts to see what tolerances your palate can tolerate!

 as I say, well aware this is not earth shattering, just means that in practice the scales are not practically usable to 0.1g precision as advertised.

And it’s insignificant compared to the myriad other variables that I’m not measuring at all / as precisely (tamp force, milk volume etc).   

[hawkinspeter]

kinderje wrote:

Great fun reading the comments.

My solution is to put a Nespresso pod in the machine and then press espresso or lungo.

<<lobs grenade into discussion and retires very quickly>> 

If you really want to set off indignant coffee aficionados, my favourite is to keep referring to “expresso”. (There are good arguments for why “expresso” is as valid as “espresso”).

[Backladder]

Obviously it will be a

Obviously it will be a different number for each type of bean and you will do it because this is important to get right 

[Author]

Replies