DEV Community

Yang Fei
Yang Fei

Posted on

Three Numbers in an RF Datasheet That Get Misread

When a switch, attenuator or termination is selected from a one-line datasheet summary, the part often works and the measurement still looks wrong. Three numbers tend to be read as absolutes when they are really conditional: insertion loss, VSWR, and the frequency range over which either of them applies.

1. Insertion loss is a curve, not a constant

A headline figure like "0.3 dB insertion loss" is meaningless without the frequency it was measured at. In a coaxial switch, conductor and dielectric loss rise roughly with the square root of frequency, contact geometry adds ripple, and the connector pairs at each end add another 0.05–0.15 dB that never appears on the component datasheet.

Two parts both quoted at 0.3 dB can differ by 0.4 dB at 18 GHz. That gap is larger than the contribution of a short cable run. For a measured path, ripple matters as much as the absolute number: if loss varies 0.2 dB across a sweep, a power measurement calibrated at one frequency inherits that variation as an error term.

A useful budget for a switched path, using worst-case rather than typical values:

Element Worst case Note
Input connector pair 0.05–0.15 dB Grows with frequency and mating cycles
SPDT or SP4T switch 0.2–0.8 dB State the frequency
Output connector pair 0.05–0.15 dB Same as input
Fixed attenuator pad nominal ±0.3 dB Accuracy is separate from flatness
Cable run 0.1–1.0 dB Dominates above 10 GHz

The practical consequence: above 10 GHz the interconnect is often the largest term, so a switch with a better mechanical layout can beat one with a better datasheet. Adding a fixed attenuator to improve match costs its nominal value plus its uncertainty — worth it for a poorly matched source, wasteful otherwise.

2. VSWR describes a match into a specified load

VSWR and return loss are the same reflection in different units:

VSWR Return loss Reflected power
1.10:1 26.4 dB 0.23 %
1.20:1 20.8 dB 0.83 %
1.50:1 14.0 dB 4.00 %
2.00:1 9.5 dB 11.1 %
3.00:1 6.0 dB 25.0 %

Return loss = −20 log₁₀(Γ), with Γ = (S−1)/(S+1). Going from 1.2:1 to 1.5:1 looks minor but multiplies reflected power almost fivefold.

What matters in a real assembly is the product of mismatches. Two 1.5:1 interfaces on either side of a device form a small resonator that produces roughly ±0.35 dB of amplitude ripple — completely unrelated to the insertion loss under test. That is the reason a 6 dB pad is commonly placed ahead of a detector: it improves the effective match by about 12 dB of return loss, at the cost of dynamic range.

It also explains why a datasheet VSWR measured into a matched load stops describing the installed system once the load itself is 1.5:1.

3. "DC to 18 GHz" and "18 to 26.5 GHz" are not comparable ranges

Coaxial line covers decades of bandwidth from DC; waveguide covers roughly one octave per physical size, with a hard cutoff below it. That single difference drives most of the decision:

  • Below 18 GHz — coaxial is nearly always the default. Precision connectors are available and one coaxial switch matrix can cover the whole band.
  • 18–40 GHz — the grey area. 2.92 mm and 2.4 mm interfaces still work, but loss and connector wear climb quickly. Short runs requiring continuous band stay coaxial; fixed-band, long, high-power runs move to waveguide switches.
  • Above 40 GHz — waveguide is normally lower loss and more repeatable, and flange tolerance becomes the dominant specification.

Switching speed is a separate axis. Electromechanical parts are robust but slow; microsecond switching needs a PIN diode switch, with a different set of loss, isolation and power tradeoffs. Comparing the two on insertion loss alone leads to the wrong conclusion.

The failure mode nobody writes into the RFQ

In an SP4T, the three inactive throws are not electrically invisible. Left open, each presents a reflection close to Γ = 1, and isolation is finite. Leakage returns through the junction and shows up as degraded isolation and a worse input match. Every unused throw needs a coaxial termination, rated for the power it may actually see — including transmitter leakage — not just for connector type. Where internal termination is required, specify an absorptive switch rather than a reflective one.

Similarly, hot switching changes contact resistance over life, so end-of-life insertion loss differs from the published curve, and a small load rated 2 W at 25 °C may be rated a fraction of that at 85 °C. Both belong in the purchasing specification.

A spec template that gets comparable quotes

State: operating band; maximum insertion loss at the top of that band; allowable ripple across band; VSWR and the load it is measured into; peak and average power with the ambient temperature; switching rate; whether switching occurs under RF power; mating cycle count; connector or flange type; and how unused ports will be terminated. Ten lines of text, and the quotes you get back are actually comparable.


These notes are part of a short reference series on reading RF component datasheets, written by DB Design / MeiXun (Wuxi) Communication Technology Co., Ltd. The full set, including the attenuator notes and the conversion tables, is at dbdesignmw.com. For specification review, contact Yang Fei at +86-510-85380190 / +86-153-5809-7083, G1-503/703/708, 200 Linghu Avenue, Xinwu District, Wuxi, Jiangsu 214028, China.

Top comments (0)