How we compare IEMs: sources, units and editorial method

GK AudioLab is owned and operated by the GK / GK HiFi brand. This is GK’s product-information and shopping guide. Comparisons involving GK products are first-party editorial content; they are not independent tests. Other brands are included for sourced comparison, without implying their endorsement or authorization.

A comparison of two named models is useful when it answers a hardware question: whether the source connectors differ, whether one configuration uses multiple driver types, whether the stated electrical figures share a reference, or whether a specification changes with a nozzle or cable option. Those questions are narrower than asking which model sounds better, but they are still practical. A missing microphone, the wrong connector or an unsupported USB-C implementation can matter more to a purchase than a category label.

We keep product specifications, editorial interpretation and seller terms separate. The database is a reference to published fields, not a listening ranking. A source-based conclusion can identify a real difference and explain the next thing to check; it cannot create evidence for bass quality, imaging, isolation or long-term comfort that has never been measured or observed.

Each model record identifies a publisher, its URL, the type of source and the date on which that source was reviewed. A manufacturer sheet can establish what the manufacturer publishes. A retailer page can establish what a seller lists for a particular version. Neither automatically becomes an independent laboratory report, and a retailer repeating a manufacturer claim is not a second independent measurement.

GK catalog material is labeled as catalog information. An outside review remains attributed to its original author or channel. We do not convert a linked video into our own listening experience, infer sponsorship terms we have not checked, or turn an unverified product graphic into a measured result. If the source changes, the historical review date remains historical until the underlying claim is checked again.

Impedance in ohms describes an electrical load, but a single nominal value does not describe its full variation with frequency. A published tolerance or test frequency belongs with the value. The source output impedance, available voltage, noise floor and the earphone load can all affect the practical result. Lower impedance alone does not mean louder, better, or easier to drive.

For example, a comparison between 23.5 ohms and 32 ohms establishes a nominal load difference. It does not establish the volume setting needed for equal loudness if the sensitivity reference is missing. Before buying an amplifier, identify the existing source and the actual problem: insufficient level, audible hiss, distortion or connector compatibility. A new specification number is not a diagnosis.

A sensitivity number needs an input reference. dB SPL/mW describes the sound-pressure figure at a power input; dB SPL/Vrms describes it at a voltage input. Two bare numbers cannot be ranked when those references differ. A source that says only dB has not supplied enough information for conversion. We preserve that omission rather than guess that a familiar market convention was intended.

With a known resistance R, one volt corresponds to 1000/R milliwatts. The idealized conversion is S per volt = S per milliwatt + 10 log10(1000/R). At 32 ohms the reference difference is about 14.95 dB; at 43 ohms it is about 13.67 dB. These examples explain the arithmetic, not the measured performance of a particular model. Frequency, tolerance, fixture, version and impedance assumptions must still match. We do not apply this formula to unknown references or use it to infer a tuning winner.

A nozzle option can change the relevant published sensitivity or tuning statement. A Type-C cable can add a digital interface whose compatibility cannot be inferred from the passive earpiece impedance. A 0.78 mm two-pin description does not establish shell recess geometry, connector keying, microphone support or compatibility with every replacement cable. Read the exact selected version, not just the family name.

The directory price bands are historical categories. They are useful for understanding the catalog arrangement but are not current offers. Compare two delivered prices in the same destination and currency, including the selected microphone or cable, shipping, taxes and return conditions. Record the seller and the date when you check them. We do not fill in local prices, warranty promises or availability from an old band.

A frequency-response or distortion claim needs a linked result with enough method information to understand it: the sample and version, coupler or fixture, calibration, insertion procedure, input level, smoothing and repeatability. A matched comparison also needs comparable conditions for both models. A manufacturer frequency-range label such as 20 Hz to 40 kHz is not a response curve and cannot establish flatness or audible extension.

Listening impressions require an identified listener, source, ear tips, fit conditions and a clear distinction between preference and a controlled observation. We have not performed the listening or measurement work implied by the published specification tables. Until such evidence is documented, we will not write invented test histories, comfort scores, failure rates, soundstage rankings or claims that one architecture must sound superior.

Existing comparison URLs come from a maintained list rather than an automatic expansion every time a model is added. Each page must preserve exact source fields and explain a useful difference or an important evidence limit. We prioritize richer analysis where a reader has a real choice to make, including ambiguous sensitivity references, edition changes and compatibility questions. More pages are not treated as proof of a better reference library.

Search history and inbound links are reviewed before removing an established page. Absence from a capped analytics export is not proof that a URL has no users or links. Where two pages answer the same intent, consolidation needs a real equivalent destination. Indexing decisions and permission to show advertising are separate: a useful public reference can remain accessible while advertising is paused.

A page assembled from two model profiles and a category prompt is a specification lookup, not a separately authored pair article. We keep that lookup accessible while excluding it from search indexing. A pair article needs its own conditional decision, version and compatibility analysis, dated sources and explicit limits in each published language. Search history alone does not satisfy this standard, and no sound measurement or personal audition is invented to meet it.

If a number or statement is wrong, send the page URL, the disputed field, the source URL and the version or date to the contact address. We can then compare the same claim instead of replacing one unsupported number with another. Missing or conflicting specifications remain marked as unknown until a source supports the correction. A substantive review date records that work; it is not refreshed automatically because the site was rebuilt.

Product links can lead to AliExpress and may generate an affiliate commission. The seller and platform control the transaction and current terms. A commission is not measurement evidence and does not justify selecting a winner. Use the source links, the uncertainty notes and the purchase checklist together, and stop the decision when an important fact is still missing.