13.8 V that isn't: cable drop, scope PEP and a lying wattmeter

Cable resistance. R = ρ·l/A = 0.0175 Ω·mm²/m × 10 m ÷ 2.5 mm² = 0.070 Ω.

Drop at 19 A. V = I·R = 19 × 0.070 = 1.33 V, so the rig should see 13.8 − 1.33 = 12.47 V. That matches the measured 12.3 V closely; the extra 0.15 V is fuse holder, crimp and connector resistance.

PSU verdict. 19 A from a 22 A continuous supply is within rating, so the supply is not the limiting item - though there is very little headroom left for an amplifier.

Fix. Move to 6 mm²: R = 0.0175 × 10 ÷ 6 = 0.029 Ω, drop = 0.55 V, giving about 13.25 V at the rig. Alternatively halve the cable length, which has the same effect. Re-terminate with proper crimped lugs, check the fuse holder for heat, and fuse at the PSU end so a short in the run is protected.

The judgement point is that a digital multimeter shows the steady drop but not the microsecond dips on key-down, which are what actually trigger the reset. Fixing the cable restores margin on both. Check the rig's specified supply tolerance before deciding 12.3 V was acceptable: many transceivers specify 13.8 V ±15 %, so the steady value was inside the window while the transient was not.

Calculation. 200 V peak-to-peak means 100 V peak, and 100/√2 = 70.7 V rms. P = V²/R = 70.7²/50 = 5000/50 = 100 W. The transmitter is delivering its rated power.

Conditions for validity.

• The load must genuinely be 50 Ω and non-inductive. Check the dummy load with a multimeter, and respect its rating and duty cycle; a 100 W load run continuously will drift in value.
• The waveform must be a clean sine. Flat-topping means the PA is clipping against the supply, and the true fundamental power is lower than V²/R suggests.
• The probe must be rated for 100 V peak at the operating frequency. Ten-to-one probes are typically rated a few hundred volts but derate with frequency; check the datasheet rather than assuming.
• The probe must be compensated and used with a short ground return. The long ground lead of a scope probe makes a loop that picks up RF and can show a misleading amplitude.
• Scope input impedance (1 MΩ in parallel with roughly 15 pF) loads a 50 Ω circuit very little at 14 MHz, but it is worth knowing it is there.

Because the reading matches the specification and the waveform is clean, this is a trustworthy 100 W measurement. It becomes the reference against which the wattmeter is judged.

First, rule out the obvious. Confirm the meter's mode: on SSB many meters read average and look low, but this is a continuous CW carrier, so a genuine average-reading meter should show close to 100 W. A 55 W reading is about 2.6 dB low, which is too much to explain by meter ballistics.

Likely causes, in order.

• The meter is sensing through the ATU or through a mismatched section, so the forward tap is not seeing 50 Ω; directivity errors of 1–2 dB are common in low-cost cross-needle meters.
• The meter is out of calibration or has a damaged detector diode; insert it directly at the rig output into the dummy load, with nothing else in line, and compare.
• Frequency response: a meter calibrated at 1.8 MHz reads low at 14 MHz and worse at 28 MHz.
• User error at the scope: if a 10:1 probe were read with the scope set to 1:1, the computed power would be wrong by 100×. Here the number came out at exactly 100 W, which is consistent with correct scaling, but always sanity-check the probe setting.

Resolution. Trust the clean-sine oscilloscope measurement as the reference, then check the meter against it at low power into the dummy load, noting its reading at 5, 20 and 50 W. If it is consistently 2.6 dB low, label it and apply a correction, or replace it. A station whose only power indication is an uncalibrated meter cannot tell a 100 W signal from a 55 W one, which matters both for licence compliance and for EMC diagnostics.