Weight - Stimulants' guide for mg level weighing equipment & methods

[Stimulants]

This guide will discuss the equipment and methods for weighing samples in the mg level.

The basic:
Accuracy - how close it is to actual weight.
Precision - how consistently it reads the result for the same weight.
Display resolution - the smallest change that can be seen on the screen.

Ideally, measurements should be accurate & precise as possible.

Equipment:
Specs are usually given like this:
210g x 0.0001g. This means that the instrument can handle up to 210g and read out to 0.1mg.

Testing methods: if any of these fails badly, the scale just sucks or it is damaged. Please read this part thoroughly. This comes ahead of "selection", because it explains the basics needed for proper selection.

Linearity:
+/- 0.4mg. This means how consistent the instrument holds the sensitivity across the entire range. For example, let's say the instrument reads a nickel at 4.9844g with nothing on it. A perfectly "linear" instrument will read 4.9844g when it is loaded with 50% the rated weight, then zero'd (tare) and a nickel placed on top of it. Although it is not very useful, it is a good test to perform on a used scale.

Repeatability/precision:
This is a measure of how precise the instrument is over a repeated measurements in one sitting. It is usually given as a standard deviation over ten measurements. It is fairly easy to test it for a 1mg resolution scale, however, operator skills can influence it on a 0.1mg or greater resolution.

Example: +/- 2mg. This spec is usually +/- 1 to 4 counts of least significant digit (the far right side number... )

Test: zero the scale first. Weigh a coin (the same coin must be used throughout the test) ten times. DO NOT zero the scale during the test.

Take the standard deviation. STDEV=(range of cells) on MS Excel. If it is significantly outside the range, the scale is likely damaged.

Note that this is not a guarantee that if a sample of 10.0000g is weighed, it will read 9.998 to 10.002 EVERY SINGLE TIME. It means that the reading will be within that range 68% of the time. Double the range (in this case, 9.996 to 10.004) for 95% certainty.

Corner load
How the reading varies when the weight is placed off center on a scale. This isn't usually explicitly given, however, using a weight that is approximately half the full capacity of the scale and the reading is off by more than about 10d (10 counts of lowest resolution) the scale is garbage to begin with or it is damaged.

Zero drift:
Digital scales have a tendency to drift a few counts. For this reason, many digital scales will quash out such a drift and return it to zero in the vicity of +/- few counts from zero. Some scales will let you disable this feature in the settings. It maybe called Auto-zero correct, ZeroTrack and the like. If there is no such option, it is possible to disable it by pre-loading on *SOME* scales. Pre-loading means: boot the scale, put a coin or two on it, then zero it. This is a user-set zero and some units' firmware see it different than a start-up zero.

Calibration drift:
How well the scale holds calibration over long term. (weeks, months)

Temperature drift:
How the sensitivity shifts as a funciton of temperature. Usually expressed in PPM.

Selecting a scale:
As shown in the repeatability section of testing, it is not wise to flirt with weighing a sample using the least significant digit even though it may let you disable the automatic zero reset feature.

#1) SWIM dislikes ACCULAB V series (V-xxx, VI-xxx, VIC-xxx). They're comparatively cheap and they comparatively suck! They don't explicitly include many specs, and they tend to suck at those specs. They suck at cornerload and calibration stability. SWIM doesn't have any experience with their high end stuff, like the L, ATL, but SWIM wouldn't buy it.

#2)
Minimum measurement weight needed to keep error within +/- 20% with 95% certainty:
+/- xx mg REPEATABILITY x 10
Equation is derived from
(2 x standard deviation /0.2 =minimum sample weight needed)

Example:
0.1mg resolution with +/- 0.1mg repeatability: 1.0mg
1mg res, +/- 1mg rep (a good 1mg res. scale): 10mg
1mg res, +/- 2mg rep(Acculab VIC-123): 20mg

Note that REPEATABILITY is not the smallest value the display can show. It is a separate spec. It is usually 1-4 times the smallest value the display can show. On a nice scale, this is usually 1. On a cheapo scale, it's a wild guess. It can be 20g x 0.0002g (+/- 0.001g) and still meet the claim that it reads to the nearest 2mg.

If the repeatability spec is NOT given, the scale likely sucks.

For the purpose of measuring RCs, 1mg is the MINIMUM acceptable resolution. 0.1mg is preferable. 0.01mg would be great, but requires greater operator skills and care as the smallest thing will affect the reading. 0.01mg units don't come by very easily and the pricing tends to be prohibitive, so they're outside the scope of this guide.

#3) Don't get a strain gauge scale if SWIY doesn't want a scale that sucks, such as ACCULAB VIC-123. That thing is good for weighing coins, accurate to +/- 2mg, as long as it is calibrated before each use.

#4) A digital scale must have an internal calibration weight or an external calibration weight must be purchased with it. The accuracy of the calibration weight is not that critical.

#5)Used mechanical scales and brand name analytical scales provide a good value if they're not damaged. Fortunately, it is easy to test them. Always buy from a supplier who will take it back if it fails the test unless it's so cheap SWIY is willing to risk it.

Brands:
Yes: Sartorius, Mettler-Toledo, Ohaus, Cahn, Denver Instruments, A&D, or Shimadzu. They're usually Made in Germany, Japan or Switzerland like fine optical and mechanical instruments.

No: Acculab V-xxx, MyWeigh Pro, generic Made-in-China garbage. As much as SWIM dislikes Acculab, it's his opinion that it's the lesser of the crappiness among crappy scales. He predicts their non-V model Acculab won't suck nearly as bad. They're usually Made in China, or assembled in (country name) using "foreign and domestic components".

Calibration weight
For the purpose of weighing out RC in the <50mg range using a balance having a capacity of 60 to 120g, the accuracy of calibration weight is not that important. If the scale calls for 100g weight and it is calibrated with 100.1g, a 10g weight will read 10.01g, 1g 1.001g and so forth. Since even eBay 100g weights are good to 0.01g or so and doesn't cost that much, it should be purchased if internal calibration is not present.

Good weighing practice/methods: (though not in particular order)
1.)Level the scale using the bubble balance on the scale. If there isn't one, well the scale probably sucks to begin with.

2.)Always keep it on. Allow to warm up for at least 15 minutes, or over an hour for 0.1mg balance before calibration or critical weighing.

3.)mg level scales are affected by convection current caused by temperature difference and it will affect the result. Don't use a weighing container that's not at the same tempreature as the scale and don't leave a hand inside the draft chamber forever.

4.) NEVER put chemicals directly on the weighing pan. Use a small weighing dish, a light weight shot glass or the like.

5.) Don't use a weighing paper with 0.1mg or 1mg resolution scales. A heavy grade tinfoil is acceptable for a 1mg scale.

6.) Always use the draft shield.

Reputation Comments on this post:

	Great, great resource for weighing methods
	Great piece- thank you!
	Useful information that is necessary for all research
	Thanks for writing this guide.
	Damn this is a great write-up. Very easy to understand great warnings against cheap imitations.
	Thanks, you supplied great information.