Walsin MLCC Export vs Our Upload File: 7 Broken Cells Repaired, 27 Part-Number Contradictions and 651 Non-Tolerances Carried Through

The Walsin MLCC export and the upload file we built from it differ in three ways, and we counted each. 22,651 tolerance cells were rewritten, 7 broken capacitance cells were repaired, and 27 capacitance cells that contradict their own part number were left alone. Every part number reconciles: 23,360 uploaded plus 241 held back is 23,601.

Venatech Limited, an authorised Walsin distributor in Hong Kong, turns the parametric export from Walsin Technology (Walsin) into the upload files behind its parts pages. This post compares the two, row by row and cell by cell: 00_ALL_MLCC_parts_combined.xlsx against Venatech_MLCC_Upload_01.xlsx to Venatech_MLCC_Upload_07.xlsx and Venatech_MLCC_NEEDS_REVIEW.xlsx, then the chip resistor (Venatech_ChipR_Upload_01.xlsx onward and Venatech_ChipR_NEEDS_REVIEW.xlsx), inductor (Walsin_Inductor_ALL.xlsx), disc capacitor (Walsin_DiscCap_ALL.xlsx), RF, varistor (Walsin_MOV_ALL.xlsx) and antenna (Walsin_Antenna_ALL.xlsx) files at part-number level. Every part number below is real. The question is the one a buyer should ask of any distributor’s data: what did you change, and what did you leave wrong?

Does the upload file hold every Walsin part number from the export?

Yes, in all seven families, with no part number lost and none invented. We read the part-number column of each export and each upload file and compared the sets. The MLCC export has 24,026 rows but 23,601 distinct part numbers; the upload holds 23,360 and the held-back file 241, which sums to 23,601 exactly. Chip resistors reconcile the same way: 234,815 rows, 234,401 distinct, 234,378 uploaded and 23 held.

The 241 rows in Venatech_MLCC_NEEDS_REVIEW.xlsx were held back for two stated reasons: 231 are flagged as customised parts, and 10 have a termination field that disagrees with the tolerance field. The 23 rows in Venatech_ChipR_NEEDS_REVIEW.xlsx are attenuators whose resistance is in decibels. Thin film is not in this table; its 1,168,998-row upload set is the subject of a separate post on thin film coverage.

What did the upload step change in the MLCC data?

Two columns almost entirely, one column in 7 cells, and one in a single cell. We merged the 23,360 uploaded rows to the export on part number and compared ten shared fields, trimmed of blanks. Tolerance was rewritten on 22,651 rows, 97.0 percent, from the ASCII +- to the Unicode plus-minus sign. Dielectric was rewritten on 15,133 rows: 15,132 from NPO to C0G/NP0 and one, MT18N391J101, from 390. Series, rated voltage, length, width, thickness and symbol changed on none.

The rewrite of tolerance is the reason a filter on 10 percent works on the upload and fails on the export. We measured the export’s three spellings in a post on tolerance notation. The other half of that story is the 709 tolerance cells the step did not rewrite.

We found that 651 of those 709 are a bare negative number such as -0.1 or -0.05, on 651 MG-series rows, in the field that should state a tolerance. The upload step carried the known MG defect straight through. The other 58 are legitimate values already in the right notation, such as ±0.5pF and -20+80%. A numeric sort on the upload’s tolerance column still puts 651 MG rows where no tolerance exists. We have not seen this counted anywhere.

Which seven capacitance cells were repaired, and which 27 were left?

The seven are the cells where the export holds something that is not a capacitance. The 27 are rows where the cell is a valid capacitance and the part number says a different one. We decoded the capacitance code in the part number of all 21,229 uploaded rows outside the 01R5 and S prefixes, and compared it with the capacitance column.

The three MT31B474 parts in the file-path group are MT31B474J101, MT31B474K101 and MT31B474M101. All seven repairs agree with the code. The 27 that remain sit in three series: 18 in MT, 4 in RF and 2 in HH, plus 3 in general-purpose sizes. That is a repair rule that fixes cells which are not capacitances and does not test cells that are. The earlier count of rows whose code and column disagree is in 28 of 24,026 rows disagree; on the 21,229 rows we could decode with one rule, we measured 27, in the upload and in the export alike. Which side is right is on our request list for Walsin.

Do the columns the upload step added agree with their sources?

All of them, on every row. The upload adds an operating temperature, a numeric capacitance in picofarad, a numeric voltage and a text description, none of which the export holds. We checked each against its source. Capacitance in picofarad disagrees with the capacitance text on 0 of 23,360 rows, voltage in volts disagrees on 0, and the description disagrees with capacitance, size, dielectric or tolerance on 0.

The operating temperature needs a different test, because it comes from the dielectric alone: C0G/NP0, X7R, X7S and X7T map to 125 °C, X5R to 85 °C, X6S to 105 °C, X8G and X8R to 150 °C. We compared the 38 series-and-dielectric pairs in the file with the operating temperature printed in section 6 of each Walsin sheet. None of the 38 contradicts its sheet. The mapping is sound where a sheet prints a range. It is still an assignment by dielectric, not a measurement on the part.

Does removing the space from a chip resistor part number ever merge two parts?

No. The upload keeps the embedded space exactly as the export has it, on 39,162 of 234,378 chip resistor part numbers: 39,087 with one space and 75 with two. We removed every space and counted distinct values again. The count stays at 234,378. Stripping the space merges no two part numbers.

The practical rule is therefore safe in one direction: delete spaces before you join, and nothing collides. It is not safe to delete them from a code you intend to order, where the second space in the 75 can be a real suffix; that distinction is the subject of why Walsin part numbers do not paste into your BOM. Our free cross-reference tool takes up to 500 part numbers or an Excel or CSV file; compare what it returns against your BOM line, not against the string you typed.

What does the upload file leave empty on purpose?

Three columns on every row, and more on some families. The template has minimum order quantity, standard pack quantity and lead time in weeks. All three are empty on all 264,958 uploaded rows in all seven families. The export holds no values for them either, so nothing was dropped. We do not publish a lead time or a pack quantity that Walsin’s own file does not give us.

In Venatech_DiscCap_Upload_01.xlsx, three added physical columns, body diameter, body thickness and lead pitch, are empty on 929 of 929 rows, and the lead-spacing column of Walsin_DiscCap_ALL.xlsx is empty on all 929. On the inductor file, saturation current is filled on 147 of 5,411 rows. One of those 147 is not a measurement: the inductor file carries Walsin’s QA test row, described as a 1 nH, plus or minus 5 percent part, with the text “TEST Saturation current 123.88” in the saturation field. The real inductor count in Walsin_Inductor_ALL.xlsx is 5,410. We list it here so that anyone who has a copy of the file can discard that row.

Stock is a dated snapshot, not a feed: 1,125 part numbers in Hong Kong as at 23 September 2026 on ready stock. We hold 28 Walsin datasheets, 451 pages, covering 26 MLCC and 2 chip resistor series, and none for inductor, disc capacitor, RF, varistor or antenna, so the operating-temperature check above could not be repeated on those lines. The capacitor range sits on the MLCC page.

The short version

We compared the Walsin MLCC export with our upload file: 23,360 uploaded plus 241 held is 23,601 distinct part numbers, none lost. The step rewrote 22,651 tolerances and 15,133 dielectrics, repaired 7 broken capacitance cells, and left 27 capacitance cells that contradict their part number and 651 bare negative numbers in the tolerance field. The columns it added agree with their sources on every row. Removing the space from a chip resistor part number never merges two parts. We will not call a cleaned file correct. We will tell you what we changed. Confirm against both datasheets and qualify in your own circuit.

Send us the part numbers you filtered with a tolerance or a capacitance column and we will tell you which of them sit in the 678 cells we did not fix. And a question back: if your distributor cleaned the manufacturer’s file, can they tell you what they left wrong?

Frequently asked questions