Decoupling at the Package: Why Case Size Beats Capacitance

Close to a large package, the capacitor’s job is set by its loop inductance, and loop inductance is set by geometry — case size, height and placement — not by capacitance. Walsin Technology (Walsin) data makes the trade visible: at 1206, 100 nF X7R ships at four different thickness nominals from 0.80 mm to 1.60 mm, and the field that decides is rated voltage.

---

Venatech Limited, an authorised Walsin distributor in Hong Kong, holds Walsin’s full MLCC parametric export and 26 Walsin MLCC datasheets. This post is written from the 24,025 real rows of that export, reading the Size, Capacitance, Dielectric, Rated Voltage and T columns together. Every part number is real and every count is of part numbers, not pieces.

Why does case size matter more than capacitance close to the package?

Because the impedance the die sees at high frequency is dominated by the inductance of the loop from the capacitor through the vias and planes to the package, not by the capacitor’s value. Above the part’s self-resonance it behaves as an inductor, and no amount of extra capacitance moves that. What moves it is a shorter, lower loop: a smaller body you can place closer, a thinner body you can put on the opposite side of a thin board, more parts in parallel rather than one big one.

So the selection question is not “how much capacitance can I get in this footprint”. It is “how small and how thin can I go and still have the capacitance I need”. The export answers the second question well, and it answers it in a column most people never filter on.

What happens to thickness when you buy capacitance in a fixed footprint?

You lose the tight thickness band. Measured across all 3,990 0402 rows in Walsin’s MLCC export, grouping on the T column:

0402 capacitance band	T = 0.50±0.05 mm	T = 0.50±0.20 mm	No T published
Under 100 nF	3,655	74	41
100 nF to under 1 µF	128	28	1
1 µF to under 10 µF	23	25	0
10 µF and above	0	9	0

Every one of the nine Walsin 0402 part numbers at 10 µF or above carries a thickness tolerance of ±0.20 mm on a 0.50 mm nominal — a 40 percent band — while 3,655 of the 3,729 dimensioned 0402 rows below 100 nF hold ±0.05 mm. We have not seen this measured anywhere. Two parts in the low profile TT series and one 1 µF soft termination part are the only 0402 exceptions above 100 nF worth noting, and the TT parts buy their 0.30±0.03 mm profile by giving up capacitance range.

That matters in exactly the place this post is about. A part going into a 0.4 mm cavity under a package, or on the back of a board that has to fit a gap pad, cannot absorb a ±0.20 mm band on a 0.50 mm part. Chasing capacitance in the same footprint hands you the loosest mechanical grade Walsin publishes at that size.

Does going up a case size give you one taller part or a choice of heights?

A choice, and it widens as the case grows. Taking one fixed electrical specification — 100 nF, X7R — and counting distinct thickness nominals:

Case size	100 nF X7R rows	Distinct thickness nominals	Range
0402	39	1	0.50 mm
0603	47	1, in three tolerance bands	0.80 mm
0805	50	2	0.80 and 1.25 mm
1206	59	4	0.80 to 1.60 mm
1210	36	3	0.95 to 2.00 mm
1812	29	4	1.25 to 2.50 mm

At 0402 there is one answer. At 1206 there are four, spanning a 2x height range for identical capacitance, identical dielectric and identical footprint. A designer who specifies “100 nF X7R 1206” and lets purchasing pick has not specified the height of the part, and the mechanical drawing may assume they did.

What decides the height of a 100 nF X7R, if not the capacitance?

Rated voltage. Higher voltage needs more dielectric between electrodes, so the body grows in the only direction the footprint does not constrain. Five real 1206 parts, all 100 nF, all X7R:

Part number	Series	Rated voltage	Thickness (mm)	Match / mismatch against a 0.80 mm keep-out
MT31B104K500	Automotive MT, AEC-Q200 compliant	50 Vdc	0.80±0.10	Match — 0.90 mm worst case
1206B104K101	General Purpose	100 Vdc	0.95±0.10	Mismatch — 1.05 mm worst case
MT31B104K101	Automotive MT, AEC-Q200 compliant	100 Vdc	1.25±0.10	Mismatch — 1.35 mm worst case, and 0.30 mm taller than the same series at 50 V
MG31B104K201	Automotive MG, no AEC-Q200	200 Vdc	1.60±0.20	Mismatch — 1.80 mm worst case, double the keep-out
SH31B104J201	Soft Termination SH, Ag-poly	200 Vdc	not published	Mismatch — no thickness in the export at all

Note the third and fourth rows against the first. MT31B104K500 and MT31B104K101 are the same series, the same capacitance, the same dielectric and the same case size. The only electrical difference is the voltage rating, and it costs 0.45 mm of height. Nothing in the part number tells you that; the last three characters encode the voltage, and the height is a consequence you have to look up.

The honest version of the same comparison at 0402 is shorter and makes the point better:

Part number	Series	Capacitance	Rated voltage	Thickness (mm)	Match / mismatch
0402B104K160	General Purpose	100 nF	16 Vdc	0.50±0.05	Match — tightest band Walsin publishes at 0402
0402B104K500	General Purpose	100 nF	50 Vdc	0.50±0.20	Mismatch — same value, four times the thickness band
SH15B104K160	Soft Termination SH	100 nF	16 Vdc	0.50±0.20	Mismatch — soft termination widens the band
ST15B104K160	Automotive ST, AEC-Q200 compliant	100 nF	16 Vdc	0.50±0.20	Mismatch — the only 0402 part in the ST series
0402X106M6R3	General Purpose	10 µF	6.3 Vdc	0.50±0.20	Mismatch — 100 times the capacitance, loosest band, lowest voltage

Confirm all of these against the Walsin datasheet for the series and qualify them in your own circuit. We will not tell you a part fits; we will show you where it differs.

How should I actually pick a package decoupling capacitor?

Fix the geometry, then the voltage rating, then take the capacitance the geometry allows. At 1206 that order matters: 100 nF X7R spans 0.80 mm to 1.60 mm across 59 rows, and rated voltage decides which you get. Four steps, in this order, and capacitance is not the first.

1. Fix the geometry first. Decide the case size and the maximum height including tolerance, from the mechanical constraint, not from the capacitance you would like.
2. Fix the voltage rating next, because that is what drives height at a given case size, as the 1206 table above shows. Specify it in the design, do not leave it to the purchase order.
3. Then take the capacitance the geometry allows, and get the rest by placing more parts in parallel rather than one larger part. Parallel parts lower loop inductance; a bigger single part usually raises it.
4. Check the dielectric and derating last, because that is the step that is reversible on paper. X5R, X6S, X7R and X7S all appear at 0402 above 1 µF; they differ in temperature characteristic and in how much capacitance survives DC bias.

For the DC bias half of step 4, see MLCC DC bias and voltage derating. For the dielectric codes, see NP0 vs X7R vs X5R. The full range is on the Walsin MLCC range page.

Which Walsin parts have no published thickness at all?

At 0402, 41 rows below 100 nF carry no T value, as does one row between 100 nF and 1 µF. At 1206, two of the 59 100 nF X7R rows have an empty T field — both are soft termination SH parts, SH31B104J201 and SH31B104J251. For those the export cannot answer a mechanical question, and a selector that filters on thickness will silently drop them.

What can Venatech add that a parametric selector cannot?

We read the mechanical column, which most cross tools do not surface at all, and we tell you when it is empty rather than returning nothing. We hold the 26 MLCC datasheets and will send the one covering your series, which matters here because the export and the datasheet can disagree — the S2 safety series’ Rated Voltage field, for instance, carries the impulse voltage rather than the working voltage on all 173 of its rows, a thing you would only catch by holding both.

And we will say what we do not hold. Six Walsin MLCC series have a full approval sheet on file and zero rows in the parametric export, so no catalogue selector — ours included — can offer them, and we would rather tell you that than return an empty result.

The short version

Near a package, geometry beats capacitance. In Walsin’s export, every one of the nine 0402 parts at 10 µF or above carries a ±0.20 mm thickness band while 3,655 of the 3,729 dimensioned 0402 rows below 100 nF hold ±0.05 mm — so buying capacitance in a fixed footprint costs you mechanical predictability. Going up a case size does not give you one taller part either: at 1206, 100 nF X7R spans 0.80 mm to 1.60 mm, and rated voltage is what decides. Fix the case size and the height first, fix the voltage rating second, take the capacitance the geometry allows, and parallel rather than enlarge. Qualification stays with you.

Paste your decoupling lines into the free cross-reference tool and we will come back with Walsin candidates field by field — including the thickness column, and including which part numbers have no thickness published. Use the cross-reference tool for the list, and ready stock to see which of them was physically in Hong Kong on 23 September 2026.

---

Frequently asked questions

---