RF Filter Selection Data Defect: One Double Space Hides 2 of Walsin’s 235 Filters

We found an RF filter selection data defect in Walsin’s 235-row filter export: the Filter Type column holds five spellings of four categories, and one extra space in the majority spelling silently deletes the only 2 filters in the whole RF line that clear 6 GHz. A reader filtering on the column’s own most common value gets the wrong answer and no error.

---

Venatech Limited, an authorised Walsin distributor in Hong Kong, holds every parametric export file published by Walsin Technology (Walsin), including FRESH_rf_filter.xlsx, the 235-row file this post is measured from. Every part number below is real and came out of that file. We hold no RF datasheet of any kind, so every figure here is the export’s own value, stated as such.

How many Filter Type spellings exist in Walsin’s 235-row RF filter export?

Five, covering four categories. We counted every cell in the Filter Type column of all 235 rows: Band Pass Filter with a double space on 110 rows, Low Pass filter with a lowercase “f” on 94, Balanced Filter on 26, High Pass Filter on 3, and Band Pass Filter with a single space on 2. That is one category — band pass — written two different ways by a single character.

Filter Type value, as exported	Rows	What differs from the “normal” spelling
Band Pass Filter	110	Two spaces between “Band” and “Pass”
Low Pass filter	94	Lowercase “f” in “filter”
Balanced Filter	26	—
High Pass Filter	3	—
Band Pass Filter	2	One space — the grammatically normal form

We have not seen this counted anywhere, including by us until today. It is not a cosmetic curiosity: it decides which part numbers a script can see.

Which naive filter query returns all 112 band pass rows, and which returns only 2?

Only a whitespace-normalising query gets all 112. We ran four queries a reasonably careful engineer might reach for first, against the real file, and recorded exactly what each one returned.

Query against Filter Type	Rows matched	Rows missed	Match / mismatch vs “give me every band pass filter”
== 'Band Pass Filter' (single space, the form that reads correctly)	2	110	Mismatch — misses every double-space row
== 'Band Pass Filter' (double space, the majority form)	110	2	Mismatch — misses both parts above 6 GHz
.str.contains('Band Pass', case=False)	2	110	Mismatch — a plain substring search still needs the single space to be present
.str.split().str.join(' '), then == 'Band Pass Filter'	112	0	Match — collapsing repeated whitespace first is the only method of the four that returns the whole category

That substring result is the one worth sitting with. .str.contains('Band Pass', case=False) looks defect-proof — it is not an exact match, it ignores case — and it still returns the same 2 rows as the strict equality check, because the literal text “Band Pass” with one space simply does not occur inside “Band Pass Filter” with two. A fuzzier query without whitespace normalisation does not protect you here.

Which two part numbers does a double-space filter silently delete?

RFBPF20077G2W0T and RFBPF25207G3W1T. We cross-checked these against the frequency columns across all seven RF files in a separate pass: they are not only the two single-space rows in this file, they are the only two part numbers in Walsin’s entire 457-row RF catalogue whose stated passband clears 6,000 MHz. The defect and the ceiling of the whole product line point at the identical two rows.

Part number	Filter Type as exported	Size	Passband	Insertion loss	Match / mismatch vs the majority-spelling query
RFBPF20077G2W0T	Band Pass Filter (1 space)	2012	6,240 – 8,240 MHz	2 dB	Mismatch — invisible to a query written on the 110-row majority spelling
RFBPF25207G3W1T	Band Pass Filter (1 space)	2520	6,230 – 8,500 MHz	3.9 dB max @ 6,230 – 7,238 MHz	Mismatch — same
RFBPF1608060K9T	Band Pass Filter (2 spaces)	1608	5,150 – 5,950 MHz	0.6 dB	Match — visible to the same query, wrong decade
KFBPF2012100C67B1U	Band Pass Filter (2 spaces)	2012	1,125 – 1,675 MHz	2.5 dB	Match — visible, lowest band in the file

We will not tell you a part is equivalent to another. We will show you, field by field, which query sees which row.

Does the same whitespace defect reach the Frequency column too?

Yes, and it is worse there. We parsed the separator between the two band-edge numbers in all 235 Frequency cells and found seven distinct spacing and character styles for what is, in every case, the same tilde-range notation: ~ on 71 rows, ~ on 69, - on 52, - on 33, ~ on 5, - on 2, and ~ on 1. A script written to split on ~ alone silently fails on the 87 rows that use a hyphen instead — more than one row in three.

That is a second, independent instance of the same class of defect in the same 235-row file: a column that looks like a clean delimiter and is not. We have not seen either one counted before this post.

Does a case-sensitive filter on Low Pass Filter even return a result?

No — it returns zero. Every one of the 94 low-pass rows spells the category Low Pass filter, with a lowercase “f”, not Low Pass Filter with a capital one. A case-sensitive equality check written the way the category name is normally capitalised in English returns nothing at all, with no error raised, against 94 real part numbers. Band Pass, High Pass and Balanced all capitalise “Filter” correctly; only the low-pass rows do not. We checked all 235 rows for this pattern and found it nowhere else in the column.

What RF documentation does Venatech hold to check any of this against?

Zero. We hold 28 Walsin datasheets, 451 pages — 26 MLCC and 2 chip resistor — and not one of them covers a filter, balun, coupler, multiplexer, antenna switch, SAW or BAW part. Our own _Walsin_Datasheets folder has an MLCC subfolder and a Resistor subfolder and nothing else; we checked it again while writing this post.

What should an engineer filtering this export do instead?

Normalise whitespace and case before you filter on any text column in this file, and do not trust a substring search to save you. Collapsing repeated spaces and lower-casing both sides of the comparison turned 2 visible band-pass rows into all 112, and it is a one-line fix once you know the column needs it.

1. Do not filter Filter Type on a literal string without normalising it first. .str.strip().str.lower() with internal whitespace collapsed is the minimum; we verified it recovers all 112 band pass rows and the 94 low-pass rows in one pass.
2. Treat the Frequency column the same way. Split on both ~ and -, with or without surrounding spaces, or you will drop roughly a third of the file.
3. Cross your RF-adjacent MLCC and passive BOM lines through the free cross-reference tool rather than hand-filtering a raw export. Paste up to 500 part numbers or drop your Excel or CSV; it is matched against Walsin’s published catalogue and states every difference from your original rather than hiding it.
4. Check what is physically in Hong Kong on ready stock, a dated snapshot as at 23 September 2026, before you commit a filter choice to a production slot.
5. Ask us for the RF documentation. We hold none today; a specific request is what puts a series on our list to Walsin.

For how the frequency strings themselves are structured, see reading an RF filter part number by band and choosing an RF chip filter by band. The same offset-column pattern appears in the multiplexer export offset by one column. The full range sits on the Walsin RF product page.

The short version

An RF filter selection data defect does not announce itself. We counted five Filter Type spellings across four real categories in 235 rows, and the only difference between two of them is a single repeated space character — the difference that hides the two highest-frequency filters Walsin publishes. The same Frequency column uses seven separator styles across the same 235 rows, and the low-pass category is spelled with a lowercase “f” on every one of its 94 rows. We hold no RF datasheet to check any of it against, so we are publishing the measurement rather than the specification.

Send us the RF line off your BOM with the raw Filter Type text still attached, and we will tell you honestly which rows a whitespace-naive query of yours would have silently dropped — then tell you which candidates were in Hong Kong on 23 September 2026. If you have found a sixth spelling we have not, tell us: that is exactly the kind of thing we would rather hear from a reader than find ourselves.

---

Frequently asked questions

<!– wp:rank-math/faq-block {"questions":[{"id":"faq-question-01","title":"What is an RF filter selection data defect?","content":"A data quality problem inside a manufacturer's own parametric export that changes which parts a selection script can see, without raising an error. In Walsin's 235-row RF filter file, the Filter Type column spells the same \”band pass\” category two ways — one with a single space, one with a double space — and a naive filter on either spelling misses rows that belong to the other.”,”visible”:true},{“id”:”faq-question-02″,”title”:”How many filter type spellings are in Walsin’s RF filter export?”,”content”:”Five, across four real categories: Band Pass Filter with a double space (110 rows), Low Pass filter with a lowercase f (94 rows), Balanced Filter (26 rows), High Pass Filter (3 rows), and Band Pass Filter with a single space (2 rows). We counted every cell in the 235-row file.”,”visible”:true},{“id”:”faq-question-03″,”title”:”Which Walsin RF filter part numbers get silently excluded by the common spelling?”,”content”:”RFBPF20077G2W0T and RFBPF25207G3W1T. Both are spelled Band Pass Filter with a single space, against 110 rows spelled with two, so a filter written against the majority spelling misses exactly these two part numbers. They also happen to be the only two parts in Walsin’s whole 457-row RF catalogue above 6,000 MHz.”,”visible”:true},{“id”:”faq-question-04″,”title”:”Does a substring search fix the Filter Type defect?”,”content”:”No. A case-insensitive substring search for \”Band Pass\” still returns only 2 rows, because the literal text \”Band Pass\” with a single space does not appear inside \”Band Pass Filter\” with two spaces. Only a query that first collapses repeated whitespace returns all 112 band-pass rows.”,”visible”:true},{“id”:”faq-question-05″,”title”:”Is the Frequency column affected by the same kind of defect?”,”content”:”Yes. We found seven different separator styles between the two band-edge numbers across the same 235 rows — a plain tilde, a spaced tilde, a plain hyphen, a spaced hyphen, and three rarer spacing variants. A parser that splits only on the tilde silently drops the 87 rows that use a hyphen instead.”,”visible”:true},{“id”:”faq-question-06″,”title”:”Does Venatech hold a datasheet to verify any Walsin RF filter specification?”,”content”:”No. We hold 28 Walsin datasheets, 451 pages, covering MLCC and chip resistor families only. Nothing covers filters, baluns, couplers, multiplexers, antenna switches, SAW or BAW parts, so every figure in this post is the parametric export’s own value, unconfirmed against any datasheet.”,”visible”:true},{“id”:”faq-question-07″,”title”:”What is the fastest way to avoid this defect when building a BOM?”,”content”:”Normalise whitespace and case on every text column before filtering, and cross the whole list through a tool that does not rely on an exact string match. The free cross-reference tool takes a pasted list of up to 500 part numbers or an Excel or CSV file, matches against Walsin’s published catalogue, and states every difference from your original rather than hiding it.”,”visible”:true}],”listStyle”:””,”titleWrapper”:”h3″,”titleCssClasses”:””,”contentCssClasses”:””,”sizeSlug”:””} –>

---