Measuring loss two ways
Insertion loss (S21) β the direct measurement. Connect the coax between the VNAβs two ports and read S21 across the band. That number is the loss, in dB.
Return loss with a short (S11) β for when you only have one port, or the far end is up a tower:
- Put a short on the far end of the coax.
- Measure return loss at the near end.
- The signal made two trips, so one-way loss is half the measured value.
A short at the far end is a perfect reflector, so anything missing from the returned signal was absorbed by the cable. If you measure 4 dB of return loss against a short, the cable has 2 dB of one-way loss.
This second method is how you check a feed line that is already installed, without taking anything down.
What the numbers should be
Manufacturer figures, per 100 feet, evaluated at three amateur frequencies from each makerβs own published data (Belden 8259 and 9258; Times Microwave LMR-400):
| Cable | 14 MHz | 146 MHz | 440 MHz | Velocity factor |
|---|---|---|---|---|
| RG-58 (Belden 8259) | 1.8 dB | 6.7 dB | 13.2 dB | 0.66 |
| RG-8X (Belden 9258) | 1.1 dB | 3.8 dB | 7.0 dB | 0.82 |
| LMR-400 | 0.46 dB | 1.5 dB | 2.7 dB | 0.85 |
Read the practical consequence, not the table. At HF almost any coax is fine. At 440 MHz, 100 feet of RG-58 throws away almost everything β 13.2 dB leaves under 5% of your power, so about 95% is gone as heat before it reaches the antenna. Put 50 W in, radiate 2.4 W.
These are figures for new cable at 25 Β°C with a matched load. Yours will be worse. A run that reads 3 dB worse than the datasheet has water in it, a bad connector, or both.
Why loss has exactly two terms β and what each one is made of
Times Microwave publishes not just a table but the model behind it. For LMR-400, loss in dB per 100 feet is
Two terms, because there are exactly two places the power goes.
The term is conductor loss. At RF, current flows in a surface layer of thickness
so skin depth falls as . The effective cross-section is a shell of that thickness, so the resistance per unit length rises as β and loss rises with it. This is why a fatter cable wins at HF: the term scales with conductor circumference.
The linear term is dielectric loss. The insulation is a lossy capacitor; the power it dissipates goes as , linear in frequency. This is why foam dielectric wins at UHF β you are replacing polyethylene with mostly air, and airβs loss tangent is essentially zero.
The crossover is the useful part. Set the two terms equal:
\sqrt{f} = \frac{0.122290}{0.000260} \approx 470 \quad\Longrightarrow\quad f \approx 221{,}000 \text{ MHz}$$ For LMR-400 the crossover is at 221 GHz β far above anything you will use, so **conductor loss dominates across the entire amateur spectrum**. That is a genuinely useful conclusion: at every frequency you care about, loss goes as $\sqrt{f}$ and the fix is a bigger conductor, not a better dielectric. The dielectric term only matters for the last few percent. Check it against the table: quadrupling frequency should roughly double loss. 150 MHz β 1.5 dB; 600 MHz should be β 3.0, and interpolating the published 450 MHz (2.7) and 900 MHz (3.9) points gives β 3.1. It holds. *The model above reproduces every point in Times' published table to within 0.05 dB/100 ft β verified in `scripts/tests/test_rf_numbers.py`.* </details> ## Locating a fault without cutting A VNA in **time-domain (TDR) mode** turns a frequency sweep into a picture of the cable's length, showing where impedance changes. A fault reflects, and the reflection's delay gives its distance. The arithmetic is the **velocity factor** from the feed-lines lesson: $$d = \frac{v_f \times c \times t}{2}$$ The factor of two is there because the pulse travels out and back. **Velocity factors worth knowing:** solid polyethylene coax (RG-58, RG-8) is **0.66**; foam dielectric (LMR-400, RG-8X) is about **0.78β0.85**. Use the wrong one and your distance is off by 20%. **What a fault looks like:** | Reflection | Cause | |---|---| | **Open** (positive spike, high impedance) | broken centre conductor, disconnected connector | | **Short** (negative spike, low impedance) | crushed cable, braid touching centre | | **Small bump partway along** | connector, adapter, or a kink | | **Bump at the far end that should not be there** | the antenna is not matched β which is the same measurement as lesson 1 | The everyday use is unglamorous and valuable: sweep a 100 foot run, see a bump at 40 feet, and know the fault is at the barrel connector you buried 40 feet out β without pulling anything down. ## What a bad connector looks like This is the failure you will actually meet, so it is worth being able to recognise. - **A small, sharp discontinuity** at the connector's position in TDR mode. - **Ripple in the SWR sweep** β a periodic wobble across frequency rather than a smooth curve. Two mismatches separated by a length of line interfere constructively and destructively as frequency changes, and the ripple's period tells you their spacing. - **Loss that changes when you wiggle it.** Sweep continuously and flex the connector. A trace that jumps has found your fault. **Water is the usual culprit.** It wicks into braid by capillary action and can travel many feet from the entry point, so the damage is often nowhere near the visible problem. Once inside, loss rises and keeps rising. ## The exercise 1. Measure a **known-good patch cable**. Record loss at 14, 146, and 440 MHz. 2. Measure your **actual station feed line** the same way. Compare with the datasheet. 3. Put a **short** on the far end and confirm the two-way method gives twice the one-way loss. 4. Run **TDR** and identify every discontinuity. Account for each one β every connector should appear, and anything you cannot explain deserves investigation. Do this once and record the numbers. When performance degrades a year from now, you will have a baseline, and comparing against it is far quicker than diagnosing from scratch. ## Check yourself 1. You short the far end of a run and measure 5 dB of return loss. What is the one-way loss? 2. TDR shows a reflection at 30 ns on RG-58. How far away is the fault? 3. Your SWR sweep ripples up and down across the band instead of curving smoothly. What does that suggest? <details> <summary>Answers</summary> 1. **2.5 dB.** The signal traversed the cable twice. 2. $d = \frac{0.66 \times 3 \times 10^{8} \times 30 \times 10^{-9}}{2} \approx 3.0$ m β about **10 feet**. The velocity factor and the factor of two both matter. 3. **Two mismatches separated along the line** β typically a bad connector plus the antenna. Their interference alternates with frequency, and the ripple's period indicates the spacing. </details>