Precision Resistor Divider Accuracy: Walsin’s ±3 ppm Grade Covers 4 of 8 Case Sizes

We parsed all 58,314 rows of Walsin’s WF thin film export. The ±3 ppm/°C grade reaches four of eight case sizes, and at 0603 it stops at 39.7 kΩ. Precision resistor divider accuracy is set by how two parts track each other, not by either one’s absolute tolerance, and tracking is in no column.

Venatech Limited, an authorised Walsin distributor in Hong Kong, holds all 203 of Walsin Technology (Walsin) parametric export files. This post is measured from the eight non-empty Thin_Film_Precision_Chip_Resistor_WF_Series_*.xlsx files and the three thick film array and network files beside them. Every part number below is real, every count is of part numbers rather than pieces, and the one part number that does not exist is marked as such.

Is a divider built from two ±0.01% resistors a ±0.01% divider?

No, and this is the error that costs a board respin. We counted the tolerance values in the WF export: 52,808 of the 58,314 rows carry a single figure of ±0.01%. That is each part’s absolute tolerance against its own nominal value. A divider’s output depends on the ratio of two such parts, and two independently trimmed ±0.01% resistors can sit at opposite ends of their windows, so the ratio error is the sum, not the figure on either datasheet line.

The same arithmetic applies to temperature. Two parts each specified at ≦±3 ppm/°C can drift in opposite directions, which is a ratio drift of up to 6 ppm/°C across the pair. Walsin’s WF export has no ratio-tolerance column, no TCR-tracking column and no matched-pair designator — we read all nine headers in every one of the eight files. If your divider needs tracking as a specification rather than as a hope, the export cannot sell it to you, and we would rather say so than let you infer it.

Which Walsin WF case sizes offer the ±3 ppm/°C grade?

Four of eight: 0402, 0603, 0805 and 1206. We counted 3,064 rows at ≦±3 ppm/°C across the whole WF export and every one of them sits at 1206 or smaller. At 1210, 2010 and 2512 the tightest grade available is ≦±5 ppm/°C; at 0201 the tightest is ≦±25 ppm/°C. We have not seen this counted anywhere, and it inverts the usual assumption that a bigger part is the more capable one.

Nine of the twenty files in Walsin’s thin film export directory return zero rows, including both professional AEC-Q200 WF files and the TTL high-power low-TCR current sensor file. We opened every one of them. If your requirement is an AEC-Q200 thin film precision part, the parametric data for it is not in our hands and we will not pretend otherwise.

Why does the tight TCR grade run out of resistance range first?

Because each grade is a shorter ladder than the one below it. We measured the resistance span of every grade at 0603: ≦±50 and ≦±25 both run 4.7 Ω to 1 MΩ over 1,058 values, ≦±15 and ≦±10 stop at 150 kΩ over 887 values, ≦±5 stops at 49.9 kΩ over 789, and ≦±3 stops at 39.7 kΩ over 769. The ladder gets both shorter and sparser as the grade tightens.

That last column is the one a divider designer actually needs, and we have not seen it published. A divider is two parts in the same grade, so the question is not how many values exist but how many pairs do. At 0603 and ≦±3 ppm/°C, 564 of the 769 values have an exact ten-times partner inside the same grade and 359 have a hundred-times partner. Both numbers are smaller than the value count, and the gap is the top of the ladder running out.

Can you build a matched 100:1 divider at 0603?

Not at ≦±3 ppm/°C, and yes at 1206. A 100 kΩ over 1 kΩ divider needs both values in one grade. At 0603 we found 100 kΩ in four grades — ≦±10, ≦±15, ≦±25 and ≦±50 — and in neither ≦±3 nor ≦±5, because both of those stop below 50 kΩ. At 1206 the ≦±3 grade reaches 120 kΩ, so the same divider is available there.

So the honest answer to a 100:1 divider at ≦±3 ppm/°C is a footprint change, not a grade change. We will not tell you a part is interchangeable. We will show you where it differs — and in this case the difference is that one of the two rows you need is not in the catalogue at all.

What does one 10:1 divider position look like across Walsin’s WF grades?

Eight real parts against one 1 kΩ value at 0603. We pulled every row at that value: eight part numbers, all stating ±0.01%, spanning six TCR grades and two power ratings. The selector will show you eight candidates that differ in the field you were not filtering on.

Does Walsin offer a thin film matched network at all?

No. We checked all three array and network files: Thick_Film_Chip_Resistor_Array__1of1.xlsx (7,250 rows), Thick_Film_Chip_Resistor_Array_Anti_Sulfuration__1of1.xlsx (2,772) and Thick_Film_Chip_Resistor_Network_Convex_Type__1of1.xlsx (98). All 10,120 rows are thick film. Not one of Walsin’s 93,914 thin film part numbers is in an array or network package.

That is the trade in one line. The best array tolerance in the export is ±1% and the best array TCR is ≦±100 ppm/°C, on 571 rows that are all 0603×4 convex type — WA06X10R0FTL is the lowest-value example. The network file is looser again at ±5% and ≦±200 ppm/°C across 97 non-jumper rows, such as WT04X100 JTL. So Walsin’s catalogue will sell you ±0.01% absolute accuracy in a discrete thin film part, or two to four elements on one substrate in thick film at ±1%, and it does not offer both in the same part number. Choosing between absolute and ratio accuracy is a real choice in this catalogue, not a marketing distinction.

Two notation defects sit in those array files and will break a script. The TCR column holds -300～+500 on 85 rows — a full-width tilde, not a hyphen — and a bare - on 12 jumper rows, alongside the ordinary ≦±200 form on 5,587. The array tolerance column carries the literal string Jumper on 12 rows where a percentage belongs.

What should a buyer in the United Kingdom or Germany ask for?

Ask for the ratio specification by name, because nobody will volunteer it. The trade press reported passive component lead times stretching through 2026 as AI server and EV build-outs drew on the same capacity, so a second source for a precision divider is worth identifying before you need it — and the field that decides whether the second source works is the one the catalogue does not print.

1. Specify the grade and the pair, not the part. “Two parts, same grade, same case size, ratio 10:1” is a buyable requirement. “±0.01%” is not, because it says nothing about tracking.
2. Cross the whole BOM at once. Paste up to 500 part numbers or drop your Excel or CSV into the free cross-reference tool. It is matched against Walsin’s published catalogue and returns candidates field by field, with every difference from your original stated rather than hidden. Filter by commercial or automotive grade, or for soft termination, high temperature, safety certified, high Q / low ESR or low profile.
3. Check what is physically in Hong Kong on ready stock before you commit. That page is a dated snapshot as at 23 September 2026, not a live feed, and 101 of its 1,125 codes have no row in any parametric export at all, so ask us directly about those.

For the material comparison behind the grade ladder, see Thin film vs thick film chip resistors. For the array packages, see Resistor arrays vs discrete. The full range is on the Walsin thin film range page.

The short version

Precision resistor divider accuracy is a ratio property, and Walsin’s WF export states only absolute properties. The ≦±3 ppm/°C grade reaches four of eight case sizes and none of 0201, 1210, 2010 or 2512. At 0603 it stops at 39.7 kΩ, so a 100:1 divider at that grade needs a 1206 footprint. Of the 769 values in that grade, 564 have an exact 10:1 partner and 359 have a 100:1 partner. All 10,120 array and network rows are thick film at best ±1% and ≦±100 ppm/°C, so the catalogue does not offer absolute and ratio accuracy in one part. Nine of twenty thin film files return zero rows, and we hold no thin film datasheet. Qualification stays with you.

Send us the divider — both legs, the ratio you need and the drift you can live with — and we will come back with Walsin candidates field by field, differences included, plus the pairs that exist in one grade and the ones that do not. Then tell us whether you would rather have ±0.01% absolute in two discrete parts or ±1% on one substrate, because this catalogue makes you pick, and we would like to know which way real boards go.

Frequently asked questions