Why a 0.001 g Reading Is Not 1 mg Accuracy

Most people who buy a precision balance are pulled in by one number: 0.001 g. Marketplace listings print it at the top, and the first line of every review says "reads to a thousandth of a gram." But a reading and accuracy are two different things. A balance that displays 0.001 g only proves its screen shows three decimals. It does not prove those decimals belong in your lab notebook.

I copied the official specifications of four balances that get compared a lot, ran the numbers, and fetched each Amazon product page three times to check stock. The results were less flattering than the listings suggest: for three of the four, the last digit is essentially a display digit, not a trustworthy one.

Three words you must separate: division, repeatability, linearity

The U.S. National Institute of Standards and Technology (NIST), in its weighing-and-scales FAQ, draws a clear line between two easily confused quantities. The scale division d is the smallest change you can read on the display. The verification scale division e is the value used to decide whether the balance is within tolerance. When a balance needs legal-for-trade use, what decides compliance is e, not d. The OIML R 76 recommendation treats "reading" and "verification" as separate registers as well.

Once those two are separated, the advertised 0.001 g drops to a display parameter. What actually decides whether you can trust the number is three things:

• Repeatability: how much a single object scatters when you weigh it ten times in a row. It sets your minimum trustworthy change.
• Linearity: whether the error stays consistent across different points of the range. It decides whether you can trust the number after switching weights.
• Temperature drift: how far the reading moves when the room changes by one degree. It decides how long the balance holds up outside a climate-controlled room.

A balance's effective resolution should be its capacity divided by its repeatability, not its capacity divided by its division. That is the baseline for every judgment below.

TL;DR: On a $40-ish budget for samples under 20 g, pick the AWS Gemini-20 and verify it yourself with weights, because AWS publishes no repeatability or linearity. For routine weighing to 200 g that settles in one second, pick the Ohaus SPX222, but note Ohaus itself marks it Not Applicable for legal trade. For a 1 mg reading with a draft break, pick the A&D EJ-303 and budget for its ±20 ppm/°C drift. For 0.1 mg with internal calibration, pick the Ohaus PX224, at the cost of a 4.5 kg body and a higher price.

Four candidates, and which tier they sit in

Compute effective resolution from the official specs and the tiers become obvious:

The first column makes Gemini-20 and EJ-303 look like the same 0.001 g tier. The fifth column shows their effective resolution differs by more than 5x, because AWS never publishes Gemini-20's repeatability. The easiest column to miss is the last: only PX224 lists a glass draft shield as standard. The other three either omit it or sell it separately.

The bill nobody adds up: temperature drift outranks the reading

The A&D EJ manual lists sensitivity drift at ±20 ppm/°C. That sounds tiny, but multiply by capacity and it is not: 310 g × 20 ppm ≈ 6.2 mg/°C. In other words, a 1 °C room change can move the reading by 6.2 mg, while the display resolution is 1 mg.

Push it further. If a bench sees a 5 °C swing across a day—ordinary in any room without active temperature control—the drift is 31 mg. That is 31 display steps, and ten times the balance's own repeatability of 0.003 g.

This does not make EJ-303 bad. It means its 0.001 g only holds under a stable temperature. The manual is explicit: operating range 10 °C to 30 °C, below 85% RH, non-condensing. Put it by a north window or under an AC vent and the 1 mg reading stops meaning anything.

I ran the same calculation on the other three. Only PX224's official page omits a drift coefficient, so this metric cannot be quantified for it—which is not an advantage, it is missing information.

Each model, with its defects written in

AWS Gemini-20: the cheap entry-level weight scale

The listing title reads 20 g × 0.001 g, with calibration weights, tweezers, and a tray included. At fetch time the price was $37.99, stock showed In Stock, rating 4.3. Its role is clear: powders, precious metals, gems—small samples.

The defects must be stated. First, the capacity is only 20 g; anything slightly larger will not fit. That is physical, not precision. Second, the official spec has no repeatability and no linearity—you cannot tell whether the last 0.001 g digit is signal or noise. Third, there is no draft shield; at the 1 mg level a single breath moves the air enough to jump the reading. Fourth, the included weight only does single-point calibration, which guarantees accuracy near that one point, not across the range.

Ohaus SPX222: fast, but Ohaus itself says not for legal trade

Ohaus spec page: 220 g capacity, 0.01 g read, linearity ±0.01 g, typical repeatability 0.01 g, stabilization 1 s, pan 120 mm, four AA cells for about 120 hours, item 30268906.

Three defects worth writing down. First, the "Legal for Trade" field reads Not Applicable—this balance cannot be used where verification or trade settlement is required, which rules it out if you weigh outgoing goods and issue weight records. Second, linearity ±0.01 g matches its 0.01 g reading exactly, so the last digit is a display digit, not a trustworthy one. Third, the line "ABS draftshield (1mg models only)" means the shield ships standard only on the 1 mg model; this 0.01 g model has none.

On price: at fetch time the buy box was priced in CNY at ¥2,776.35, with other offers on the same page between ¥2,566.78 and ¥2,880.50, and stock noted "Only 4 left in stock." The same model's list price varies a lot by region, so verify before ordering rather than treating this as a global price.

A&D EJ-303: the most honest spec sheet in the 0.001 g tier

The EJ manual: 310 g capacity, 0.001 g read, repeatability (std dev) 0.003 g, linearity ±0.003 g, sensitivity drift ±20 ppm/°C, pan Ø110 mm, about 80 hours on four AA cells, factory calibration weight 300 g.

Its strength is honesty: repeatability 0.003 g is stated as larger than the 0.001 g reading, which is the maker admitting the last digit is a statistical display. The weakness is the same place—the 1 mg you paid for is really trustworthy to 3 mg. It also has no internal calibration; you load the 300 g weight by hand, the EJ-11 draft shield is optional, and the operating ceiling is only 30 °C.

On stock I must be straight: I ran three fetches of the same product page. Two returned In Stock, one returned Currently unavailable. That inconsistency usually comes from page caching and seller switching, but it also means you have to confirm before ordering rather than trusting my three results.

Ohaus PX224: the 0.1 mg tier, buying away a whole class of trouble

The Ohaus China page: 220 g capacity, 0.1 mg read, linearity ±0.0002 g, typical repeatability 0.0001 g, pan 90 mm, internal calibration, glass draft shield standard, certification field "No", item 30372587.

Its value is making draft shield and internal calibration standard—exactly the two items most often cut at the 0.1 mg level and most able to wreck the data. The defects are also real: first, 4.5 kg plus a glass shield makes it a fixed instrument, not portable; second, the pan is only 90 mm, smaller than the SPX222's 120 mm; third, the certification field is again "No", so it still fails where a verification certificate is needed; fourth, the stabilization time differs between Ohaus and its resellers, ranging from 2 s to 4 s, so trust the random manual over any listing.

On stock and price: at fetch time the buy box showed no usable price, stock read "Only 4 left", rating 4.1. That state usually means a third-party seller supplies the ASIN with a floating price, so check the "Ships from" and "Sold by" lines before buying.

Decision table: work backward from use, not from the reading

• Under 20 g powders, gold, gems, tight budget: Gemini-20, and run your own ten-weigh repeatability test first.
• Routine weighing to 200 g, one-second settle, no verification needed: SPX222.
• 1 mg reading, stable bench temperature, OK with manual calibration: EJ-303, and get a draft shield.
• 0.1 mg reading, internal calibration and standard shield, fixed location: PX224.

If your use involves issuing weight records to customers or needs a verification certificate, none of these four is the answer—their official pages say Not Applicable or "No." Look at models that explicitly carry the matching certification class and ask for a calibration certificate.

FAQ

Does a 0.001 g reading mean I can weigh to 1 mg?

No. 0.001 g is only the division d, the smallest step the display shows. NIST's weighing FAQ defines d and the verification division e separately, and compliance is judged on e. In daily use the more useful yardstick is repeatability: a balance with 0.003 g repeatability is trustworthy to 3 mg, not 1 mg.

My balance sits near an AC vent and the digits jump. Is it broken?

Usually not; it is airflow. Balances at the 1 mg level are extremely sensitive to moving air. EJ-303's manual lists 10–30 °C and below 85% RH non-condensing, but says nothing about wind resistance. Add a draft shield, move away from the vent, put it on a solid bench, and let it warm up before weighing.

I need to weigh outgoing goods and issue weight records. What do I pick?

Read the "Legal for Trade" or "Certification" field first. In this article, SPX222 says Not Applicable and PX224 says "No," so neither fits a verified scenario. Choose a model that explicitly carries the matching certification class and request a calibration certificate.

Do I have to buy calibration weights?

Yes, and buy them for the capacity you use. EJ-303's factory weight is 300 g, PX224 uses internal calibration, SPX222 ships a calibration mass for capacities under 620 g. Only weights let you answer "is this balance still accurate today," and at the 1 mg level that question needs asking every day.

Why do prices for the same balance differ so much across pages?

Currency and seller both vary. In this article SPX222's buy box was in CNY (¥2,776.35) while Gemini-20 was in USD ($37.99); that is localization from where the network sits. The same ASIN can also show several third-party offers side by side. Treat every price here as a magnitude reference and read the live page at checkout.

On price, stock, and links

All prices, stock levels, and ratings were captured before publication and will change with seller and region. Official spec data comes from the Ohaus and A&D product pages and manuals; the definitions of d and e are quoted from the NIST weighing-and-scales FAQ. This article contains Amazon affiliate links; buying through them may earn a commission, which does not change what you pay and does not change the defect judgments above.

👉 Join Xiaomi MiMo Platform: Leading AI model platform with cost-effective inference

👉 Join Aliyun AI: Top AI products with exclusive coupons for business innovation

📌 This article was AI-assisted generated and human-reviewed | TechPassive — An AI-driven content testing site focused on real tool reviews