How this instrument works
Insertion loss answers one narrow, practical question: how much weaker does the signal get because this particular device is sitting in the line? You measure the power arriving at a load, insert the filter, connector, cable, or splice you actually care about, measure again, and take ten times the base-ten logarithm of the ratio: IL = 10·log₁₀(Pin ⁄ Pout). That framing is what separates it from a generic loss figure — it is always a comparison against a specific reference state, not an absolute property of a material.
The logarithm exists because a two-connector chain should cost twice the decibels of one connector, and only a log turns a multiplied ratio into an added one. Ten times log rather than twenty is a power convention: insertion loss compares watts in to watts out directly, so no squaring is involved the way it is for a voltage or current ratio. A device that lets every watt through scores 0 dB; one that loses half its power scores 3.01 dB, regardless of whether that power was 1 milliwatt or 100 watts, since only the ratio survives the division.
The number is deliberately agnostic about where the missing power went. A connector can dissipate power as resistive heating in its contacts, or it can reflect power straight back toward the source because its impedance does not match the line — insertion loss folds both mechanisms into one figure and reports only what failed to arrive at the far end. That is a real limitation: a component with excellent insertion loss can still be a poor match, throwing power backward instead of forward, which is precisely why datasheets list return loss as a separate specification alongside it.
- Enter Input power — the power delivered to the device under test, in watts or milliwatts, at the near-side reference point.
- Enter Output power — the power that actually arrives at the load on the far side of the device, measured in the same conditions.
- Read Insertion loss, dB — the instrument applies 10·log₁₀(Pin ⁄ Pout) and returns the figure directly.
- Keep both readings in consistent units; only their ratio matters, so watts and milliwatts may each be entered on their own field.
Worked example — a connector that drops 10 W to 8 W
A coaxial jumper and its two mated connectors sit between a transmitter rated at 10 W and a power meter at the antenna feedpoint. With the meter reading 8 W, the ratio Pin ⁄ Pout is 10 ⁄ 8 = 1.25, and the formula gives IL = 10·log₁₀(1.25) = 0.969100130081 dB, which the instrument displays as roughly 0.97 dB for these exact inputs.
That looks trivial on its own, and for a single connector pair it is. The reason RF technicians still log every jumper, adapter, and splice individually is that these figures add along the chain: four such joints at roughly a dB apiece can quietly cost a full watt or more out of a ten-watt budget before the signal ever reaches free space, and a link margin calculated from one generous connector spec instead of the measured chain routinely comes up short in the field.
Questions
How is insertion loss different from attenuation?
Insertion loss is defined relative to a specific reference measurement — the power that would arrive with the device absent or bypassed, compared to the power that arrives with it switched into the line. Attenuation more broadly names any power loss over a path or medium, often quoted per unit length independent of any single component. Every insertion-loss figure is an attenuation figure, but it is always tied to one device and one reference state, which is why manufacturers publish it as a per-part spec rather than a per-metre rate.
Does insertion loss tell me whether power was absorbed or reflected?
No. It only reports the shortfall between what went in and what came out, whether that power turned into heat inside the device or bounced back toward the source from an impedance mismatch. Distinguishing the two needs a separate reflected-power measurement, reported as return loss. A component can have low insertion loss and still have poor return loss, which is why RF and fiber datasheets always list both figures.
How is insertion loss actually measured in a lab?
With a reference reading first. A vector network analyzer is calibrated with a direct 'through' connection standing in for the device, recording the baseline power transfer, then the device under test replaces that through connection and the drop is logged as its insertion loss. Fiber technicians follow the same logic with a light source and power meter: reference power first, device inserted second.
Can the result be negative?
Yes, if Output power exceeds Input power. That happens with an active device such as an inline amplifier or repeater rather than a passive connector, and the formula returns a negative number because the log of a ratio below one is negative. Engineers usually just call that condition gain and quote it as a positive number in the opposite direction.
What insertion loss counts as good for a connector or filter?
A single mated RF connector pair in good condition typically runs 0.1 to 0.3 dB; a fusion-spliced fiber joint is often near 0.1 dB, while a mechanical fiber connector pair commonly sits closer to 0.3 dB. A passband filter is a different case entirely, since it is designed to pass most power at some frequencies and none at others — 1 to 3 dB of passband insertion loss is common even in a well-built design.
Why does the instrument only need power in and power out?
Because the ratio of those two readings is the entire definition — no cable length, frequency, or impedance value changes the arithmetic once both powers are measured in the same unit at the same reference points. Those other quantities explain why the loss has the value it does, but the decibel figure itself only depends on the ratio you actually measured.