Four Fields Leave 25 Candidates: What We Need From You Before We Can Quote a Rail

We measured how far four fields actually get you. Size, dielectric, capacitance and voltage — 0402, NPO, 10 pF, 50 Vdc — returns 25 Walsin part numbers across six series. Knowing how to specify an MLCC for quotation means knowing which field collapses that list. It is tolerance, and most requests leave it out.

---

Venatech Limited, an authorised Walsin distributor in Hong Kong, holds all 203 of Walsin Technology (Walsin) parametric export files and 28 Walsin datasheets. This post is measured from one of those files, 00_ALL_MLCC_parts_combined.xlsx — 24,026 rows and 23,601 distinct part numbers — and every part number below is real. Every count is a count of part numbers, not pieces.

One thing stated plainly before anything else: there is no price in this post and there will not be one. A price depends on quantity, date and grade, and a blog post cannot know any of those. What this post can do is show you exactly how much ambiguity each missing field leaves on our side of the conversation, measured rather than asserted.

How many Walsin part numbers does a four-field MLCC request return?

A median of two and a worst case of 25. We grouped all 24,026 rows by each successive field set and counted the distinct part numbers landing on each signature. Four fields — size, dielectric, capacitance, rated voltage — produce 7,608 distinct signatures, of which only 38.7% resolve to a single part number. Three fields are far worse: a median of four candidates and a worst case of 83.

Fields you give us	Distinct signatures	Median candidates	Worst case	Resolve to exactly one part
Size + dielectric + capacitance	2,225	4	83	580 (26.1%)
+ rated voltage	7,608	2	25	2,945 (38.7%)
+ tolerance	16,404	1	8	11,826 (72.1%)
+ maximum thickness	17,880	1	8	14,048 (78.6%)
+ series or termination	23,624	1	3	23,586 (99.8%)

We have not seen this counted anywhere. It is the arithmetic behind every “can you quote this?” email that arrives with four fields in it, and it is the reason our answer is sometimes a question rather than a part number.

Which single field collapses the candidate list fastest?

Tolerance, by a wide margin. Adding it to the four-field set moves the share of requests that resolve to exactly one part number from 38.7% to 72.1% — a 33.4-point jump, larger than any other single field contributes. Adding maximum thickness after it adds 6.5 points. Adding the series or termination preference adds the remaining 21.2 points but is really you choosing the part, not specifying it.

So the ranking is not intuitive. Most requests we receive are precise about capacitance and voltage, which are the two fields the buyer is already thinking about, and silent about tolerance, which is the field that actually decides which of 25 rows we are allowed to offer. If you send us four fields and a tolerance, three quarters of the time there is one answer and we can go straight to availability. If you send four fields, we are choosing on your behalf, and we would rather not.

That is worth saying as a position rather than a hint. We will not pick your tolerance for you and call the result a cross. Tolerance on an NPO part at 10 pF is the difference between a filter that centres and one that does not, and it is yours to set.

What do the 25 candidates for one request actually differ on?

Tolerance and series, and almost nothing else. We pulled every row matching 0402, NPO, 10 pF, 50 Vdc: 25 part numbers across six series, and 24 of the 25 share an identical thickness of 0.50 ± 0.05 mm. The twenty-fifth is the soft-terminated part, at 0.50 ± 0.20. So thickness separates one row out of twenty-five here, and tolerance separates the rest.

Part number	Series	Tolerance as exported	Thickness	Match / mismatch vs 0402 NPO 10 pF 50 Vdc
0402N100J500	General Purpose	±5%	0.50±0.05	Match on all four fields; the default commercial answer
0402N100F500	General Purpose	±1%	0.50±0.05	Match on all four; mismatch on tolerance if ±5% was assumed
HH15N100C500	High Q / low ESR	±0.25 pF	0.50±0.05	Match on all four; mismatch — absolute tolerance, not percentage
RF15N100B500	Microwave RF	±0.1 pF	0.50±0.05	Match on all four; tightest tolerance at this value
MT15N100F500	Automotive, AEC-Q200	±1%	0.50±0.05	Match on all four; mismatch — automotive grade, different qualification
MG15N100G500	Automotive without AEC-Q200	-0.02	0.50±0.05	Match on all four; mismatch — tolerance cell is unreadable, see below
SH15N100J500	Soft termination, Ag-poly	±5%	0.50±0.20	Match on all four; mismatch — only row here with a different thickness band

Six of the seven match on every field the request contained. We will not tell you a part is interchangeable. We will show you where it differs — and here the difference is a tolerance notation, a qualification status and a thickness tolerance band, none of which were in the request. Confirm against the Walsin datasheet and qualify in your own circuit.

Why can we not read the tolerance on 656 of Walsin’s MLCC rows?

Because the cell contains a bare negative number instead of a tolerance. We read the tolerance column across all 24,026 rows. Most cells hold a readable value — +-5%, +-1%, +-0.25pF — but 656 rows hold -0.01, -0.02, -0.05, -0.1 or -0.2, and every one of the 656 sits in the MG series, which has 788 rows in total. That is 83.2% of one automotive series whose tolerance is not machine-readable and not unambiguous to a human either.

The same 656 rows appear in the per-series file 02_Automotive_MG.xlsx, so this is not an artefact of combining the exports. The sizes affected are the ones that matter: 221 rows at 0603, 180 at 0805, 150 at 0402, 67 at 1206, 18 at 1210, 14 at 0201 and 6 at 1812. Our reading is that these are absolute picofarad tolerances with the sign convention stripped — -0.05 sitting where ±0.05 pF belongs — because on the same rows the termination column carries the identical string, and the banked finding on that column is that all 1,106 of its filled cells hold a tolerance rather than a finish. But that is a reading, not a fact we can show you.

Does maximum thickness matter enough to put in the request?

It buys 6.5 points of precision and occasionally saves the whole design. We measured it both ways, from the T column of 00_ALL_MLCC_parts_combined.xlsx. Across the catalogue, adding maximum thickness to the five-field set moves the resolve-to-one share from 72.1% to 78.6% — useful but not decisive. At the specific position above it separates exactly one row from twenty-five. The column holds 45 distinct thickness strings and is populated on 23,679 of 24,026 rows, so 347 rows publish no thickness at all.

The reason to send it anyway is that thickness is the field that fails late. A tolerance mismatch shows up in a measurement at bring-up; a thickness mismatch shows up when the part does not fit under a heatsink, a shield can or a connector overhang, which on a dense power-delivery board is the week before build. If there is a height restriction on your rail, one line in the enquiry is cheaper than finding out at the mechanical review.

So: send tolerance always, send thickness when you have a ceiling, and send neither as a guess. A guessed constraint narrows our search to the wrong 25 rows, which is worse than a wide search.

What do we need that is not in the export at all?

Four things, and we can be exact about why, because we checked which columns are empty. The MLCC export has sixteen columns and three of them are completely blank across all 24,026 rows: Packing, SPQ and Thickness. A fourth, Termination, is filled on 1,106 rows and holds a tolerance rather than a finish.

What we need	Why it is not in the export	What happens if you leave it out
Grade — commercial, automotive, or automotive AEC-Q200	No qualification column exists; grade is inferable only from the series name	We offer commercial by default; MT, MG and ST never surface
Reel size or minimum packing quantity	Packing and SPQ are blank on all 24,026 rows	We cannot confirm packing from the export and have to ask Walsin
Annual usage and the date you need parts	Not a catalogue field at all	We cannot tell you whether the dated Hong Kong shelf covers you
Whether a field is locked by an existing PPAP or AVL entry	Nothing in any export records your approval status	A cross that changes a locked field is a resubmission, not a substitution

That last one is the expensive one. If your incumbent part is on an approved vendor list, changing a field the submission covered turns a sourcing decision into a qualification cycle. Tell us which fields are locked and we will treat them as hard constraints instead of guessing which of your five fields is the flexible one.

And the honest limit on our side: we hold 28 Walsin datasheets totalling 451 pages, and across all 451 pages there is no country-of-origin statement, no change-notification procedure and no qualification report. The two chip resistor sheets we hold are Walsin_MR_Automotive_Thick_Film_Chip_Resistor.pdf and Walsin_WR_General_Purpose_Chip_Resistor.pdf, and only one document in the whole library — ASC_MR_V13 — names a RoHS exemption. We will tell you what we do not have. Nobody else will.

What should you send us to get candidates back in one pass?

Six lines per part, and a note on what is locked. The trade press reported MLCC lead times of 26 to 32 weeks on high-grade series through 2026, with AI server and electric-vehicle build-outs drawing on the same capacity, so the cost of a second round of questions is measured in weeks rather than days. One complete enquiry beats three partial ones.

1. Send size, dielectric, capacitance, rated voltage and tolerance. Those five fields resolve 72.1% of positions to exactly one Walsin part number. Add a maximum thickness if your rail has a height ceiling.
2. Say which grade you need and which fields are locked by a PPAP or an AVL entry. That turns our search from a guess into a filter.
3. Cross the whole BOM at once rather than one line at a time. Paste up to 500 part numbers or drop your Excel or CSV into the free cross-reference tool. It decodes 18 manufacturers’ part numbers and matches against 800,000+ crossable Walsin part numbers, returning candidates field by field with every difference from your original stated rather than hidden.
4. Check what is physically in Hong Kong on ready stock before you commit to a production slot. That page is a dated snapshot as at 23 September 2026, not a live feed.

A word on the application that is driving most of these enquiries. We are not going to imply an approval or a bill of materials we do not have: Venatech has no system vendor’s approval and no sight of any accelerator platform’s bill of materials, and nothing in this post should be read as one. What we can see from the catalogue is the second-order effect — the capacity pulled toward high-capacitance small-case parts is capacity that used to serve ordinary lines, which is why a general-purpose 0603 now quotes longer than it did. That is the squeeze worth planning for, and it is the one a second source fixes.

Check us while you are here. UL file E182369 is searchable in UL’s own public database in about thirty seconds, and the TUV certificates R50381780 and R50195920 are named and numbered for the same reason. Unit 207, 2/F, Sun Ling Plaza, 30 On Kui Street, Fanling, N.T., Hong Kong, and (852) 2628 4777 — a real address and a real phone number, which are trust signals precisely because you can check them.

For what the AI build-out did to ordinary lines, see What the AI build-out did to your general purpose MLCC. For the fields a cross may and may not change, see Reading a GPU VRM power stage as a shopping list. The full range is on the Walsin MLCC range page.

The short version

Knowing how to specify an MLCC for quotation is knowing which field does the work. Size, dielectric, capacitance and voltage leave a median of two Walsin candidates and a worst case of 25 — and at 0402, NPO, 10 pF, 50 Vdc that worst case is real, spanning six series with 24 of the 25 sharing one thickness. Adding tolerance takes the catalogue from 38.7% to 72.1% unique, a bigger gain than any other field. Maximum thickness adds 6.5 points and prevents the failure that arrives latest. Three of the sixteen export columns are blank on all 24,026 rows, so packing and reel size have to come from Walsin, and 656 MG rows carry a tolerance we cannot read.

Send us one line per part with all five electrical fields and a note on what is locked by an existing approval, and we will come back with Walsin candidates field by field, differences included — then tell you which of them was in Hong Kong on 23 September 2026. And if you read those 656 MG tolerance cells differently from us, tell us: that is the question we most want to put to Walsin, and your enquiry is what gets it asked.

---

Frequently asked questions

---