Antennas · Volume 26

RF Power, SWR & Field-Strength Measurement

Through-line wattmeters (Bird 43, MFJ-849, Daiwa CN-801), SWR meters, dummy loads (50 Ω 5 W → 1.5 kW), RF probes, signal generators (HP/Agilent, SDR-as-source, NanoVNA port-1, ADF4351 cheap synth boards) — the test bench that complements the VNA

Contents

SectionTopic
1About this volume
2Through-line wattmeters — Bird 43 and its descendants
3Modern cross-needle SWR/power meters
4Dummy loads — the load that doesn’t radiate
5RF probes — high-impedance voltage measurement
6Signal generators — HP/Agilent, R&S, the lab gold standard
7SDR as signal source — HackRF, BladeRF, USRP
8NanoVNA port-1 as a CW source
9ADF4351 / ADF4355 cheap PLL synthesizer boards
10Field-strength meters and EIRP estimation
11DIY build — a 50 Ω 100 W RF dummy load
12DIY build — a Tayloe RF probe for VHF/UHF
13Commercial buys
14Common gotchas and myths
15Resources

1. About this volume

The NanoVNA covered in Vol 24 (NanoVNA Deep Dive) measures ratios — S-parameters of a network referenced to a calibrated 50 Ω port. It tells you everything about impedance match, insertion loss, and return loss, but it carefully avoids telling you anything about absolute power. The bench gear in this volume answers the complementary question: how many watts is actually flowing here, and how much of it is going where you wanted it to go?

That distinction matters because antenna characterization without absolute-power instruments is incomplete. A NanoVNA can confirm an EFHW presents 50 Ω on 40 m, but it cannot tell you whether your 100 W transmitter is actually delivering 100 W into it, or whether a connector is dissipating 15 W as heat because the center pin is intermittent, or whether your linear amplifier is producing the rated PEP without splatter on adjacent channels. Those questions need wattmeters, dummy loads, RF probes, and signal generators — the legacy bench instruments that have been around since the 1950s and remain in production because nothing has replaced them.

The volume is organized around the two complementary measurement tasks. Sections 2–5 cover measuring what your radio is putting out: through-line wattmeters (Bird 43 and its descendants in §2, the modern cross-needle ham-shack meters in §3), the dummy loads that absorb the test signal (§4), and the RF probes that read voltage at points where a 50 Ω instrument can’t reach (§5). Sections 6–9 cover generating known signals for receiver and antenna testing: bench signal generators (§6), SDRs pressed into service as sig-gens (§7), the NanoVNA’s own port-1 used as a fixed-frequency CW source (§8), and the cheap ADF4351-based synthesizer boards that have democratized signal generation up to 4.4 GHz (§9). Section 10 covers field-strength measurement and EIRP estimation — the bridge to regulatory compliance in Vol 31 (Regulatory & RF safety). Sections 11 and 12 are DIY builds (a 100 W dummy load, a Tayloe RF probe). Section 13 is the commercial-buy ladder, §14 the standard gotcha collection, §15 the resource list.

Cross-references run heavily into Vol 3 (dB and dBm) for the power math, Vol 4 (Antenna theory) for SWR and reflection-coefficient definitions, Vol 5 (Transmission lines) for the coax-loss accounting that converts indicated power to delivered power, Vol 16 (BALUNs and UNUNs) and Vol 17 (Antenna tuners) for matching-network measurement, and forward into Vol 27 (Spectrum analyzers) for the harmonic and spurious measurements that an absolute-power meter cannot make. Read this volume after Vol 24 so the VNA + wattmeter split is concrete in your head.

2. Through-line wattmeters — Bird 43 and its descendants

The Bird Electronic Corporation Model 43 was introduced in 1953 and remains in production today, seventy-three years later. That longevity isn’t sentiment — it’s because the architecture is essentially unimprovable for its purpose. The Model 43 is a passive through-line meter: signal goes in one N-female connector, comes out the other, and a directional-coupler slug installed in a rotating turret samples either the forward or reflected wave and drives a 30 µA D’Arsonval meter movement calibrated in watts. Insertion loss is below 0.05 dB; the meter draws no power from the line; there are no batteries; there is nothing to update. A Model 43 from 1965 with a fresh slug reads as accurately today as a new one.

The architecture lives or dies on the slugs — the plug-in directional coupler cartridges that determine the meter’s frequency range and full-scale power for any given measurement. Each slug is a hand-wound coupler tuned and calibrated by Bird for one frequency band and one full-scale wattage. Rotate the turret 180° to reverse the directivity (forward becomes reflected); swap slugs to change band or power range. A typical ham-shack Bird 43 collection might hold five or six slugs covering HF through UHF at 25 W, 100 W, 250 W, and 1 kW levels. The slug naming convention uses a number for full-scale watts and a letter for frequency band:

Slug suffixFrequency rangeTypical full-scale options
C2 – 30 MHz100C / 250C / 1000C — HF/60 m and below
D25 – 60 MHz25D / 100D / 250D / 1000D — 6 m and low VHF
E50 – 125 MHz25E / 100E / 250E / 1000E — 2 m and aviation
H100 – 250 MHz25H / 100H / 250H — high VHF
B200 – 500 MHz25B / 100B / 250B — 1.25 m and 70 cm
A450 – 940 MHz25A / 100A / 250A — 70 cm and ISM 900
J900 – 1800 MHz10J / 25J / 50J — 23 cm and DECT/cellular
K1.7 – 4 GHz10K / 25K — S-band and Wi-Fi 2.4

The full slug catalog runs to thousands of variants (Bird publishes a folded chart of all of them); for amateur work a 100C, 100D, 100H, 250C, 250H, and a 1000C is a reasonable starting set that covers 160 m through 70 cm at QRP, 100 W, and high-power HF station levels.

The accuracy specification is 5% of full scale, which is the single most-misunderstood number on the meter. A 250 W slug indicates within ±12.5 W anywhere on its scale — meaning that at 25 W indicated, the actual power is somewhere between 12.5 W and 37.5 W. People look at the meter, see the needle pointing at 25, and confidently report “25 watts” when the meter is actually saying “somewhere between 12 and 38.” The practical rule is use a slug whose full scale is within 4× of the measurement — for a 50 W measurement use a 100 W or smaller slug, not a 1000 W slug. The Bird 43 reads accurately around the middle two-thirds of its scale; below ~10% of full scale it’s a relative indicator only.

Bird has produced two notable variants over the decades. The Bird 43P adds a peak-reading circuit (a held-peak indicator) that is essential for SSB and digital-modes PEP measurement — a standard 43 reads average power, which on an SSB signal is roughly half the PEP. The Bird 4391A is the modern digital version: same slug interface, but the analog meter is replaced by a microprocessor that reads forward and reflected simultaneously, computes SWR and return loss in real time, and displays everything on an LCD. It also auto-ranges within a slug’s range. Used 4391As trade at $700-1500 in mid-2026.

The used market is the realistic acquisition path for the classic 43. A clean Model 43 body in mid-2026 trades $200-300 on eBay, with individual slugs at $30-80 each depending on rarity (HF slugs are cheap; the rare microwave J and K slugs can run $150+). A complete five-slug working bundle should cost $300-450 — meaningfully less than a single new MFJ cross-needle meter of equivalent capability. The slug-swap ergonomics are sometimes criticized as inconvenient compared to a fixed cross-needle meter; in practice you set up the right slug for the session and leave it for the day. Bird also sells the Model 4421 which is essentially a 43 in a desktop case with two slug positions so you can flip between bands without re-installing a slug.

3. Modern cross-needle SWR/power meters

The dominant ham-shack instrument has not been the Bird 43 since the 1980s — it’s the cross-needle SWR/power meter, a class of instrument whose mechanical innovation made measurement so much more convenient that it displaced both the dedicated SWR bridge and the single-needle wattmeter in most shacks. The cross-needle design uses two mechanically independent D’Arsonval movements whose needles are arranged at right angles, one driven by the rectified forward-wave sample and the other by the reflected-wave sample. The needles cross at a marked grid of SWR contours printed on the meter face. You read forward power on one scale, reflected power on the other, and SWR at the crossing point — all simultaneously, with no mode switch, no nulling adjustment, and no calculation. For a typical SSB transmission you see PEP forward, average reflected, and the resulting SWR contour in one glance while the operator is talking. This is genuinely brilliant industrial design.

The market splits along two axes: HF-only versus HF-plus-VHF/UHF, and power range (sub-100 W QRP-suitable versus full-legal-limit 2 kW). The major vendors are Diamond Antenna (Japan), Daiwa (Japan), MFJ (USA), and a long tail of Chinese rebrands. Comparison of current mid-2026 options:

ModelFrequencyPower rangesPeak/avgBacklightUSD (mid-2026)
MFJ-8491.8 – 525 MHz30 / 300 WYesLED$130
MFJ-862144 – 525 MHz30 / 300 WYesLED$130
Daiwa CN-801V1.8 – 150 MHz20 / 200 / 2000 WYesLamp$250
Daiwa CN-801HP140 – 525 MHz20 / 200 / 2000 WYesLamp$260
Diamond SX-2001.8 – 200 MHz5 / 20 / 200 WNoLamp$150
Diamond SX-400140 – 525 MHz5 / 20 / 200 WNoLamp$170
Diamond SX-11001.8 – 1300 MHz30 / 200 / 1000 WYesLamp$310
Comet CMX-23001.8 – 525 MHz30 / 300 / 3000 WYesLED$290

For an HF-only station the Daiwa CN-801V is the workhorse choice — broad power coverage, peak/average switching, sensible 200 W mid-range that suits 100 W transceivers, and the meter face is large enough to read across the room. For HF-plus-VHF/UHF coverage in a single instrument, the Diamond SX-1100 or the Comet CMX-2300 are the two contenders; the SX-1100 wins on price and frequency range (it actually works at 23 cm where the Comet falls off above 525 MHz), the Comet wins on the 3 kW high range that suits amplifier work.

Insertion loss in a properly designed cross-needle meter is below 0.1 dB through 500 MHz; the connector and PCB traces dominate above that. The Diamond SX-1100’s specified loss at 1.3 GHz is 0.4 dB, which is meaningful for amplifier work but invisible for ordinary transmit power monitoring. None of these meters is suitable for measurement above its specified upper frequency — the directional coupler’s directivity collapses outside its design band, and a 1 dB error in coupler directivity becomes a 30% error in SWR indication.

The cross-needle architecture is not without limits. The mechanical movement has finite response speed — voice peaks faster than ~50 ms are averaged together by the meter ballistics, which is what the peak-hold circuit is for. Below about 5 W on most meters’ lowest scale the needles barely deflect, so QRP operators routinely run a separate 5 W cross-needle or use a low-power slug on a Bird 43. And the SWR scale assumes a clean sinusoidal signal — heavily modulated digital signals (FT8, RTTY) drive the indicated SWR slightly higher than the actual matched-condition SWR because the meter is responding to the spectral envelope rather than the carrier. For digital-modes operation, calibrate your station’s “good SWR” reading in CW mode and treat the digital-modes reading as a relative indicator.

4. Dummy loads — the load that doesn’t radiate

A dummy load is a resistive termination — ideally a pure 50 Ω with no reactance from DC to your highest frequency of interest — that converts RF power to heat without radiating. You use it whenever you want to test a transmitter without putting a signal on the air: tune-up procedures, SWR-meter calibration, amplifier conditioning, key-down testing during repairs. The whole point is thermal management — the load has to dissipate the transmitter’s full output continuously without changing impedance, melting, or starting a fire.

PowerTypeCoolingBandwidthApprox USD (mid-2026)
5 WSMD chip-resistor on PCBConvection (still air)DC – 6 GHz$5 – $15
50 WThick-film resistor + heatsinkConvection or small fanDC – 1 GHz$25 – $50
100 WCantenna (oil-filled)Conduction to oilDC – 500 MHz$50 – $100
500 WOil-cooled large cantennaConduction to oil + ambientDC – 500 MHz$200 – $400
1.5 kWOil-cooled (paint-can class)Conduction + forced convectionDC – 500 MHz$400 – $800
2 kWAir-cooled with finned heatsink + blowerForced airDC – 2.5 GHz$1,200 – $2,500
5 kWWater-cooledLiquidDC – 1 GHz$5,000+

The thermal math is non-negotiable. A resistor dissipating P watts in steady state reaches a temperature above ambient of ΔT = P × R_θ, where R_θ is the thermal resistance (°C/W) of the path from the resistor’s hottest junction to the surrounding environment. A TO-220 thick-film resistor with no heatsink has R_θ of roughly 60 °C/W — at 5 W dissipation the junction sits 300 °C above ambient, well past the failure point. Bolt that same resistor to a finned heatsink with 5 °C/W (a typical 50-mm-cube heatsink), and the junction sits 25 °C above ambient at 5 W, or 250 °C above ambient at 50 W (a marginal but workable design). Push past about 50 W continuous and convection cooling stops being enough; the design moves to forced air or oil immersion.

Oil immersion is the classical solution for amateur QRO. The “cantenna” — a coffee can or paint can full of mineral oil with a resistor and a coax connector — was invented by Heathkit in the 1960s and remains the cheapest path to 1 kW dissipation. Oil’s specific heat (~1.7 J/g·°C) and the can’s surface area let you absorb a kilowatt for several minutes of continuous key-down before the oil temperature rises into the “uncomfortable to touch” range. The trade-offs: oil takes time to cool between transmissions (a five-minute key-down is followed by twenty minutes of cooling); the can must be sealed to prevent oil from leaking; and the oil eventually degrades and needs replacement every few years.

Mineral oil is the right choice today. Original Heathkit cantennas were filled with transformer oil that contained polychlorinated biphenyls (PCBs) — anyone restoring a vintage cantenna must assume the original fill is PCB-contaminated and dispose of it as hazardous waste. Food-grade mineral oil (the same product sold as USP-grade light mineral oil for ~$15/quart at any pharmacy) works well, is non-toxic, and is safe to handle. Modern commercial transformer oils (Shell Diala, Nynas Nytro) are also acceptable but cost more. Do not use motor oil, vegetable oil (which polymerizes and gels), or silicone oil (which can have problematic dielectric breakdown behavior).

The non-inductive resistor requirement is absolute. A wirewound resistor is a coil — it has substantial series inductance that turns it into a resonant circuit at RF. A “non-inductive” wirewound (Ayrton-Perry-wound) is better but still problematic above 30 MHz. The right choices are thick-film resistors (e.g., the Vishay LTO050 series), thin-film, or carbon-composition (which has the broadest RF bandwidth but is electrically noisy and has lower power density). For dummy-load duty, thick-film terminated with the appropriate heatsink is the modern standard.

The specific products worth knowing in mid-2026:

  • Bird Termaline series — the professional gold standard. The 8201 is a 50 W convection-cooled brick; the 8401 is a 500 W oil-filled; the 8930 is a 1 kW air-cooled. New they are eye-wateringly expensive ($1,500-$5,000 range); used Termaline 50 W and 500 W units regularly trade for $200-$500 on eBay.
  • MFJ-260C — a 1.5 kW oil-filled cantenna at $130 new, the budget standard for amateur high-power testing. Comes empty; you add the oil. SWR <1.2:1 to about 30 MHz, degrading to ~1.5:1 at 150 MHz.
  • Mini-Circuits TERM-50W-N — a $30 N-male terminated 50 Ω load, 8 GHz bandwidth, 1 W continuous. The default lab-grade SWR-reference termination; every bench has at least one.
  • Bird 8890 — the high-end air-cooled solution: 2.5 kW, DC to 2.5 GHz, with a built-in temperature sensor and forced-air blower. $1,500 new.
  • Heathkit HN-31 “cantenna” — historical reference, the original. If you have one, treat the original oil as hazardous and refill with food-grade mineral oil.

A common DIY approach is exactly what Heath did: put a non-inductive resistor in a paint can full of mineral oil. It works, the bandwidth is limited by the resistor’s own self-inductance (typically usable to 250 MHz with a TO-220 part), and the price is right ($30 in parts). See §11 for a full DIY 100 W build.

5. RF probes — high-impedance voltage measurement

An RF probe is a small enclosure containing a diode rectifier circuit, a high-input-impedance DC connection on one end, and a probe tip on the other. You touch the tip to an RF voltage source — typically a high-impedance node where a 50 Ω instrument can’t be connected without disturbing the circuit — and the DC output goes to a high-input-impedance DC voltmeter (any modern DMM with 10 MΩ input). The probe measures RF voltage amplitude, from which RF voltage and (with knowledge of impedance) RF power can be computed.

The classic single-diode design is a series germanium diode (1N34A or 1N270, chosen for low forward drop ~0.2 V) feeding an RC low-pass filter consisting of a 0.01 µF ceramic capacitor and a 100 kΩ resistor. The diode rectifies the RF; the cap charges to the peak voltage; the 100 kΩ resistor sets the discharge time constant. The DC output approximates V_DC ≈ V_peak − V_diode_drop, where V_peak is the peak voltage of the RF signal and V_diode_drop is the small (and frequency-dependent) forward voltage of the diode. A typical implementation might read 0.7 V_DC when probing a 1 V_peak (about 0 dBm into 50 Ω) signal at HF.

The Tayloe probe is the improved twin-diode full-wave version, originally described by Dan Tayloe N7VE in QST and widely copied since. It uses two Schottky diodes (1N5711 is the standard part — fast, low forward drop at ~0.3 V, low junction capacitance) wired as a full-wave rectifier. The full-wave topology rectifies both halves of the RF cycle, doubling the DC output for a given RF input and improving the linearity at low signal levels because both diodes are conducting on alternate half-cycles and self-bias each other. Performance is roughly 2× better in measurement sensitivity than the single-diode design, with usable response from 1 MHz to 500 MHz.

Calibration is the operator’s responsibility. Build the probe, then drive a known signal level from a calibrated source (a bench sig-gen at a known frequency and amplitude) and record the DMM reading. Repeat at several signal levels (-20 dBm, -10 dBm, 0 dBm, +10 dBm) and several frequencies (1 MHz, 14 MHz, 50 MHz, 144 MHz, 432 MHz). Plot or table the results. You’ll find that at low frequencies (HF) the response is linear and roughly equal to V_peak; at higher frequencies the input capacitance of the probe progressively loads down the measurement and the reading drops. Above 100-200 MHz typical perfboard construction self-resonates because the probe’s input lead inductance combined with its body capacitance forms an LC circuit — your DC reading goes wild at the self-resonant frequency. The Tayloe topology, built compactly with short leads (~5 mm), can push usable response to 500 MHz; sloppy construction limits it to 50 MHz.

The right tool selection question: when do you use an RF probe instead of a scope or a wattmeter? The probe shines at high-impedance points where a 50 Ω instrument would short out the signal — the high-voltage end of a half-wave dipole, the output of an antenna tuner before the SWR meter, the gate of an unmatched FET amplifier. It’s measuring amplitude, not waveform — if you need to see harmonic distortion or amplitude modulation, use a scope (HF) or spectrum analyzer (VHF and up, see Vol 27). And it’s a relative instrument unless you’ve calibrated it; the diode’s forward drop varies with temperature and current, so an uncalibrated probe is good for relative measurements (is this trace at higher or lower amplitude than that one?) but not absolute power.

A probe is also useful for trimming antennas in the field. With the transmitter delivering ~1-5 W into the antenna, probe the feedpoint or the tip of a wire and see the relative voltage — a higher reading means closer to the resonant peak. Trim the wire until the reading maximizes.

6. Signal generators — HP/Agilent, R&S, the lab gold standard

A signal generator is a frequency-agile RF source with controllable output amplitude, calibrated to laboratory accuracy. The bench instruments in this section are the gold standard for receiver testing, filter measurement, mixer characterization, and any antenna or RF system test that requires a known input level with very low phase noise and harmonic content. The market falls into four tiers by both era and capability.

The classic HP/Agilent workhorses (used market, $300-$3,000). The HP 8657A (100 kHz – 1 GHz) and HP 8657B (100 kHz – 2.06 GHz) are the most-deployed bench sig-gens in amateur and small-lab service worldwide. Released in the early 1980s, they featured a calibrated step attenuator (140 dB range in 0.1 dB steps), AM/FM/phase modulation, GPIB control, and stability sufficient for almost any non-cellular measurement. Phase noise is around -130 dBc/Hz at 20 kHz offset at 500 MHz — fully adequate for filter testing, receiver sensitivity measurement, and antenna sig-gen duty. Used 8657As in mid-2026 trade at $300-600; 8657Bs at $400-800. The principal failure modes are the reed-relay step attenuator (typical life is 100,000 cycles, after which relays stick open or closed — replacement attenuators are scarce and expensive), the GPIB interface card (often failed on units not used in years), and the NiCd battery for non-volatile state memory, which leaks and damages the PCB if not removed (always pull the battery on any used HP instrument before powering it).

The HP 8648C (100 kHz – 3.2 GHz) and HP 8648D (100 kHz – 4.0 GHz) are the next-tier replacements introduced in the mid-1990s, with VFD or color displays, lower phase noise (-145 dBc/Hz at 20 kHz on the 8648D at 1 GHz), and a different attenuator topology that’s more reliable. Used 8648C/D in mid-2026 run $1,000-$2,500 depending on options.

Mid-tier vector and modulation-capable instruments ($2,000-$10,000 used). The Agilent E4423B (250 kHz – 2 GHz), E4438C (250 kHz – 6 GHz), and N5181A (100 kHz – 6 GHz) MXG add I/Q modulation, arbitrary-waveform generation, and the ability to play complex modulated signals — essential for testing modern digital receivers (LTE, Wi-Fi, etc.). Used E4438Cs are $3,000-$5,000; N5181As are $5,000-$10,000. For antenna and filter work these are overkill; they earn their cost on the digital-receiver bench.

The premium R&S tier ($30,000+ new). The R&S SMA100B (8 kHz – 67 GHz with options) defines the current state of the art for phase noise, with -135 dBc/Hz at 1 kHz offset at 1 GHz carrier — roughly 30 dB better than the HP 8657 series. Phase noise at this level matters for two specific applications: very narrow-band filter selectivity measurement (where the sig-gen’s skirts must be steeper than the filter under test), and receiver sensitivity testing where reciprocal mixing in the receiver’s first LO is the dominant noise source. For most antenna work, the 100x lower phase noise of an SMA100B versus an 8657A is invisible.

Phase noise in practical antenna work: it almost never matters for SWR sweeps, return-loss measurements, or basic gain comparison — these are essentially CW tests where the signal-to-noise ratio inside the measurement bandwidth is set by other factors (insertion loss, cable loss, dynamic range of the receiver). Where it does matter is selective measurement — narrowband filter characterization, mixer LO leakage, and any test where you’re trying to measure a small signal close in frequency to a large one. The HP 8657A is the right instrument for amateur and most professional antenna work; the SMA100B is the right instrument for cellular-base-station development.

Buying advice for the used HP/Agilent tier: always test the attenuator at the seller’s bench if possible. Take a small signal level and step the attenuator from minimum to maximum in 10 dB increments; the output should drop cleanly with each step. Skipped or stuck steps indicate a reed-relay failure that may or may not be repairable. Test the modulation modes (AM and FM). Confirm the GPIB interface works if you intend to use it. And budget for at least one of: a 10 MHz frequency reference (rubidium or GPSDO) for calibration drift correction, a calibrated power meter (Bird 43 or an HP 8481A sensor) for verifying the attenuator settings, and a spectrum analyzer for harmonic content verification.

7. SDR as signal source — HackRF, BladeRF, USRP

A modern SDR is, structurally, a programmable signal generator. The transmit chain has a DAC, an upconverter, a programmable LO, and an output amplifier — exactly the architecture of a bench sig-gen, packaged into a $300-$2,000 USB device that costs an order of magnitude less than an equivalent benchtop instrument. For antenna and RF system testing where the demanding phase-noise and harmonic-content requirements of receiver-sensitivity work don’t apply, an SDR makes an excellent and convenient signal source.

HackRF One (1 MHz – 6 GHz, half-duplex). The HackRF’s TX path uses the Maxim MAX2837 transmitter chip, delivering 0 to +15 dBm depending on band and the IF/RF gain settings (-15 dBm to +15 dBm software-controllable, with the upper limit dropping off above 3 GHz where the chain runs out of headroom). For amateur and antenna testing this is plenty. The simplest way to generate an unmodulated carrier is:

hackrf_transfer -t /dev/zero -f 144000000 -s 8000000 -x 47 -a 1 -l 0

This commands the HackRF to TX from /dev/zero (a zero-valued stream → unmodulated carrier at the center frequency), at 144 MHz, with 8 MHz sample rate, TX VGA gain 47 (near max), the amp on (-a 1), and the LNA disabled. For host-side waveform generation (FM modulation, AM modulation, OFDM stimulus) use GNU Radio or SDR++ in TX mode.

The HackRF’s known limitations as a sig-gen: the MAX2837 produces second and third harmonics typically 30-40 dB below the carrier — for clean output you need an external lowpass filter for the band of interest. The broadband noise floor is roughly -100 dBm/Hz — orders of magnitude higher than a bench sig-gen’s -160 dBm/Hz. Phase noise is around -85 dBc/Hz at 10 kHz offset — adequate for SWR sweeps and antenna trim, but not for selective filter measurement. Frequency accuracy is determined by the onboard crystal oscillator, which is typically within ±10 ppm out of the box; for precise frequency work feed it an external 10 MHz reference through the CLKIN connector.

BladeRF 2.0 micro xA9 (47 MHz – 6 GHz, full-duplex). The BladeRF uses an Analog Devices AD9361 transceiver — the same chip in the USRP B200 series and many software-defined-radio products. Output is up to +6 dBm. The TX-path phase noise is roughly -105 dBc/Hz at 100 kHz offset, an order of magnitude cleaner than the HackRF, suitable for moderately selective filter testing. The integer-N PLL produces a cleaner spectrum than the HackRF’s fractional-N. New units in mid-2026 are $530 (xA4) to $720 (xA9).

USRP B205mini and B210 (70 MHz – 6 GHz). The Ettus B-series uses the AD9361 (B210) or AD9364 (B205mini), with phase noise around -120 dBc/Hz at 100 kHz offset — within 10-15 dB of bench-instrument quality. Output is up to +10 dBm on B210, +6 dBm on B205mini. Pricing in mid-2026 is $1,200 (B205mini) to $2,000 (B210). For prototyping cellular-class receivers or doing serious RF system tests where the SDR replaces a sig-gen, this is the right tier.

The decision matrix for when SDR-as-sig-gen is the right call: when convenience, programmability, and modulation capability are decisive (you can play arbitrary waveforms, sweep frequencies under software control, integrate with measurement scripts); when the antenna or filter under test doesn’t care about phase noise (most SWR, return-loss, and gain measurements); when you already own the SDR for receive duty. When it’s wrong: when you need calibrated absolute power accuracy (the SDR’s output level is not calibrated and varies with frequency and software gain settings — characterize against a power meter), when you need very clean spectral output for compliance-class harmonic measurements, when you need frequency reference accuracy without an external 10 MHz lock, when you need phase noise below -120 dBc/Hz.

In practice: for antenna tuning, BALUN measurement, and filter sanity-checks, the HackRF you already own is fine. For receiver-sensitivity testing at MDS levels (-130 dBm), use a bench sig-gen.

8. NanoVNA port-1 as a CW source

The NanoVNA’s port 1 generates a stimulus signal whenever the instrument is sweeping — that stimulus is just a CW source. By setting the NanoVNA into single-point sweep mode (sweep span = 0 Hz, center = your test frequency), port 1 outputs a continuous-wave signal at approximately 0 dBm (typically -3 to +0 dBm depending on hardware version and frequency).

This makes a NanoVNA an adequate emergency or low-cost sig-gen for filter through-loss measurement at a single frequency, antenna SWR check at a single point, or sanity-testing a receiver. The procedure on a typical NanoVNA-H or LiteVNA is: enter the menu, set sweep span to 0, set center frequency to the desired test point, ensure the calibration is appropriate for the frequency (a TDR or DC-coupled cal is wrong for high-frequency CW use — use a normal CAL.). Port 1 now outputs CW; connect your device under test to port 1 and your measurement instrument (DMM with RF probe, power meter, or scope with high-bandwidth probe) downstream.

The limitations are stark. The output level is not flat across frequency — variations of ±2-3 dB are typical across the 50 kHz to 1.5 GHz range, with bigger drops above 1 GHz on stock NanoVNA-H V3 hardware. There is no modulation capability. Phase noise is not specified and is generally not characterized — it’s adequate for filter measurement at moderate Q (Q < 1000) but not for narrow-band selective measurement. Frequency accuracy is set by the TCXO (typically ±0.5 ppm — better than most HackRFs). And there’s no high-resolution amplitude control — port 1 is essentially fixed-output, with no calibrated step attenuator.

The right use case is opportunistic. You’re already at the bench with the NanoVNA out for an SWR measurement; you need to check whether a filter has dropped its passband; flip the NanoVNA to span-0 mode at the filter’s center frequency, connect, and read. For dedicated sig-gen work, use a real bench instrument or an SDR.

9. ADF4351 / ADF4355 cheap PLL synthesizer boards

The Analog Devices ADF4351 is a fractional-N PLL synthesizer chip — wideband internal VCO covering 2.2 to 4.4 GHz with onboard frequency dividers extending the output range down to 35 MHz. Programmed via a 3-wire SPI interface, it outputs a CW signal at a user-selectable frequency in 1 kHz steps (or finer, depending on PFD frequency and word-length choices) at -7 to +5 dBm output level. The chip alone costs about $10 at distributors; the Chinese AliExpress eval boards that put the chip on a PCB with SMA connectors and a USB-to-SPI bridge cost $10-$25. Genuine Analog Devices eval boards (EVAL-ADF4351 / EV-ADF4355SD1Z) cost $200-$400.

The chip’s appeal: a $20 board gives you 35 MHz to 4.4 GHz signal-generation coverage, programmable from any microcontroller or Raspberry Pi. The use cases in the amateur and embedded-RF world have been everything from cheap local oscillators for homebrew transverters (e.g., for 23 cm and 13 cm operation), to bench sig-gens for filter characterization, to PCB-level clock sources for FPGA designs, to GPS reference distribution networks.

The trade-offs are real and need to be understood:

Spur content — the ADF4351 produces charge-pump spurs at offsets equal to multiples of the PFD (phase-frequency-detector) reference frequency, typically -20 to -30 dBc. With the default 10 MHz PFD you get spurs at ±10 MHz, ±20 MHz, ±30 MHz, etc. — visible on a spectrum analyzer and potentially problematic for selective filter work. Cleaner operation is possible with a lower PFD frequency (1 MHz) but at the cost of slower lock and finer step resolution.

Phase noise — measured phase noise of an ADF4351 board with the default external 10 MHz reference is roughly -90 dBc/Hz at 10 kHz offset, -100 dBc/Hz at 100 kHz offset. This is 20-30 dB worse than a bench sig-gen, more than adequate for most antenna and filter work, well short of what you want for receiver-sensitivity testing.

Output filtering — the cheap AliExpress boards do not implement output lowpass filtering. The chip’s output spectrum includes the desired frequency, plus the second and third harmonics, plus the VCO and divider outputs all leaking. For clean output you need an external lowpass filter for the band of interest. The genuine Analog Devices eval boards include switched-bank lowpass filters; the AliExpress clones do not.

PLL lock and reference — the chip needs a 10 MHz (or other-frequency) crystal-oscillator reference. The AliExpress boards include a 25 MHz or 10 MHz onboard crystal of unknown quality; for precision work you feed it an external GPSDO or rubidium reference. Always confirm PLL lock by reading the LD (lock detect) pin or measuring the LE-asserted state in the board’s status register.

The ADF4355 extends the upper frequency to 6.8 GHz with a similar internal architecture and similar AliExpress eval-board market (~$30-50 boards). For 5-6 GHz antenna and Wi-Fi work this is the right cheap choice.

A particularly useful variant is the F4HWN board (also seen as “ADF4351 with OLED” or “RF Signal Generator ADF4351 with display”), which adds an STM32 or similar microcontroller, an OLED display, and a rotary encoder so the board operates as a standalone bench sig-gen without needing a host computer. You set frequency with the encoder, the OLED shows you the current setting, and the chip outputs CW. Pricing is $30-$50 on AliExpress. For the “I need a CW source right now without booting a laptop” use case, this is the practical answer.

10. Field-strength meters and EIRP estimation

A field-strength meter (FSM) measures the electric-field intensity at a point in space, expressed in either V/m or dBμV/m. It does this by combining a calibrated antenna (typically a half-wave dipole, a known-gain horn, or a small loop) with a calibrated RF voltmeter. Knowing the antenna’s effective height or aperture and the receiver’s sensitivity, the instrument translates received voltage to incident field strength.

The professional commercial-broadcast FSMs are the high end. The Potomac Instruments FIM-71 (now Holaday HI-3604) is a long-standing broadcast-survey instrument covering 0.15 to 30 MHz at $1,000-$3,000 used. The Narda SRM-3006 is the EMC-test workhorse, a selective FSM/spectrum-analyzer hybrid covering 9 kHz to 6 GHz with field-strength probes calibrated for both magnetic and electric field components — $15,000-$30,000 new. The Wandel & Goltermann EFA-3 (now Narda EFA-300) is the classic for low-frequency power-line EMF surveys.

For amateur and informal work, the homebrew approach is entirely practical: a calibrated dipole at the test frequency, feeding a NanoVNA or SDR receiver of known calibration. You measure received voltage, look up the dipole’s effective height for that frequency (h_e = λ/π for a half-wave dipole), and compute E = V_rec / h_e. The accuracy is limited by your knowledge of the calibration chain, but ±2 dB is achievable with care.

The more common amateur use case is EIRP estimation by reciprocity. You want to know your station’s actual EIRP (Effective Isotropic Radiated Power, expressed in dBm or watts), which is the transmitter output power plus the antenna gain minus the feedline losses. Direct measurement requires a calibrated receiver at known distance in the far field. The free-space path loss equation, in dB, distance in meters, frequency in MHz, is:

FSPL (dB) = 20·log(d) + 20·log(f) − 27.55

The factor of -27.55 is the speed-of-light constant; equivalent constants exist for d in km and f in GHz. The Friis equation relates transmit and receive powers:

P_RX (dBm) = P_TX (dBm) + G_TX (dBi) + G_RX (dBi) − FSPL (dB)

Solve for EIRP = P_TX + G_TX:

EIRP (dBm) = P_RX − G_RX + FSPL

Worked example. You’re at 446 MHz (UK PMR / US ISM band). You set up a 6 dBi reference antenna (e.g., a calibrated Yagi) at 10 m distance from the antenna under test, with both antennas at the same height and polarization-aligned. Your receiver reads -40 dBm at the reference antenna terminals.

FSPL = 20·log(10) + 20·log(446) − 27.55 = 20 + 53.0 − 27.55 = 45.4 dB

EIRP = -40 − 6 + 45.4 = -0.6 dBm, or about 0.87 mW.

That’s the EIRP of whatever was transmitting at the antenna-under-test — the transmitter power plus antenna gain minus feedline losses, expressed at the antenna’s terminals. For a 5 W (37 dBm) transmitter into a feedline with 1 dB loss, this measurement suggests the antenna under test has a gain of -0.6 − (37 − 1) = -36.6 dBi — clearly a malfunctioning or very-low-gain antenna. Far more typical: a 5 W transmitter through a 6 dBi antenna with 1 dB cable loss gives EIRP ≈ 42 dBm (16 W), measured at the reference antenna at 10 m as P_RX = 42 + 6 − 45.4 = 2.6 dBm — a strong, easily measurable signal.

Practical caveats. The far-field requirement: r > 2D²/λ where D is the larger antenna’s largest dimension. At 446 MHz (λ=0.67 m), a 6 dBi Yagi might be 1 m long, requiring r > 3 m. The 10 m distance in the example is fine. Ground reflections: at any practical outdoor measurement, the direct path and the ground-reflected path interfere. The Friis equation assumes free-space (no reflections), so measurements over typical ground show ±6 dB pattern variations with distance — choose multiple distances and average, or use a “two-ray model” correction. Polarization alignment is mandatory — a 90° mismatch costs 20-30 dB. Use the same polarization on both sides and verify.

EIRP estimation is the crucial measurement for regulatory compliance — see Vol 31 (Regulatory & RF safety) for MPE (Maximum Permissible Exposure) calculations and OET-65 workflow.

11. DIY build — a 50 Ω 100 W RF dummy load

A complete BOM for a continuous-50-W (intermittent-100-W) dummy load with usable response from DC to about 500 MHz, built around twelve parallel non-inductive thick-film resistors:

ItemPart # / sourceQtyUnit USDTotal
Non-inductive thick-film resistor 600 Ω 5 W TO-220Vishay LTO050F6000JE3 (Mouser 588-LTO050F6000JE3)12$1.50$18.00
Aluminum enclosure 110×60×30 mmHammond 1590B (Mouser 546-1590B)1$15.00$15.00
Thermal pad TO-220 0.15 mm siliconeBergquist SP400 (Mouser 851-SP400-0.007-00-00)12$0.30$3.60
Mounting screws M3x12 with washersMcMaster 91290A11512$0.15$1.80
SO-239 (UHF-female) chassis connectorAmphenol RFX-12181$4.00$4.00
50 mm × 50 mm × 10 mm 12V fan, ball bearingSunon ME50101V1 (Mouser 645-ME50101V1)1$5.50$5.50
Hookup wire 18 AWG silicone insulation 1 mGeneric1$2.00$2.00
Misc. (heatsink compound, solder, hardware)——$5.00$5.00
Total———$54.90

The topology: twelve 600 Ω resistors in parallel give 600/12 = 50.0 Ω; total power rating is 12 × 5 W = 60 W continuous. Derate to 50 W continuous for thermal margin (per-resistor dissipation must stay below 80% of rating for life). The Hammond enclosure is the heatsink — clamp the resistors to one face with thermal pads, using the enclosure walls as the thermal mass and the 50 mm fan providing forced airflow across them. At 50 W continuous, expect the enclosure to stabilize at 50-60 °C above ambient — warm but well within the resistors’ 155 °C maximum junction temperature.

The construction sequence:

  1. Lay out the twelve resistors on one of the long faces of the Hammond enclosure, in two rows of six with 15 mm pitch.
  2. Drill twelve M3 mounting holes for the resistor tabs.
  3. Drill an SO-239 hole on one end face and a fan-mounting hole + airflow opening on the opposite end.
  4. Apply thermal pads, mount resistors with M3 screws and washers, finger-tight then snug.
  5. Wire all twelve resistors’ “input” leads (the ground-referenced lead, depending on the package — confirm from datasheet) to a single common bus connected to the SO-239 shell. Wire all twelve “output” leads to the SO-239 center pin. Use star-grounding at the SO-239 to minimize lead inductance.
  6. Mount the fan with M3 hardware. Wire the fan to a 12V DC input — a barrel jack or a USB-C breakout works.
  7. Close the enclosure.

Tuning. Connect the dummy load to a NanoVNA (see Vol 24 (NanoVNA)) on the SO-239 port. Run an S11 sweep from 1 MHz to 1.5 GHz. Expect:

  • 1 MHz to 30 MHz: SWR < 1.05:1 (essentially perfect)
  • 30 MHz to 200 MHz: SWR < 1.2:1
  • 200 MHz to 500 MHz: SWR < 1.5:1
  • Above 500 MHz: SWR climbs progressively because the lead inductance of the TO-220 packages becomes a significant fraction of 50 Ω

If your SWR is dramatically wrong (e.g., 2:1 at 14 MHz), check for cold solder joints at the SO-239 connector, an open resistor (one out of twelve open changes 50 Ω → 54.5 Ω, which is SWR 1.09 — small but detectable), or a short across one resistor (one shorted gives 50 Ω → 45.8 Ω, SWR 1.09).

Critical: do not substitute wirewound resistors. A wirewound resistor is a coil. Even “non-inductive” Ayrton-Perry-wound types have residual inductance that ruins the RF response. The Vishay LTO series is thick-film and is designed for RF service. Mouser also stocks Ohmite TGHHV (thick-film, similar specs) as an alternate.

Critical: do not put fewer resistors in parallel. Eight parallel 400 Ω 5 W resistors would also give 50 Ω at 40 W, but per-resistor dissipation at 40 W is 100% of rating — they will fail in months. The 12-in-parallel design keeps per-resistor dissipation at 67% of rating at 50 W input, well within the derating curve.

Going higher in power. For 200-500 W continuous, scale up: 24 × 1.2 kΩ 25 W resistors in parallel (Vishay LTO100 or RCD HSL050) give 50 Ω at 600 W rating, mounted to a finned aluminum heatsink with forced air. The bandwidth degrades slightly (the larger packages have more lead inductance) but you can still hit 1.5:1 SWR through 200 MHz with care. Above 500 W continuous, mineral-oil immersion is more practical than air cooling — at that point a paint-can cantenna with one large slab thick-film resistor (Florida RF Labs 800-50, $80) and 1 gallon of food-grade mineral oil is the right approach.

12. DIY build — a Tayloe RF probe for VHF/UHF

A complete BOM for a Tayloe twin-Schottky RF probe usable from 1 MHz to 500 MHz:

ItemPart # / sourceQtyUnit USDTotal
Schottky diode 1N5711 SOT-23 or DO-35NXP 1N5711 (Mouser 771-1N5711)2$0.40$0.80
Ceramic capacitor 0.001 µF X7R 100VMurata GRM21A5C2A102JA01L (Mouser 81-GRM21A5C2A102JA01L)1$0.10$0.10
Ceramic capacitor 0.01 µF X7R 50VMurata GRM188R71H103KA01D1$0.10$0.10
Resistor 100 kΩ 1/8 W carbon filmVishay CFR-12JT-52-100K1$0.05$0.05
BNC chassis jack panel-mountAmphenol B6252B1-ND or equivalent1$3.00$3.00
Knurled aluminum tube enclosure 12 mm OD × 80 mmMcMaster 1610T11 or 3D-printed PLA1$8.00$8.00
Probe tip — copper nail, sharpened or telescoping antenna tipHardware store1$1.00$1.00
Test lead BNC-to-banana 1 mGeneric1$5.00$5.00
Perfboard 15×30 mmGeneric1$1.00$1.00
Total———$19.05

Topology. The Tayloe configuration is a full-wave rectifier built from two Schottky diodes in opposing orientations:

Probe tip ---||--- D1 ---+---||--- output (to DMM via 100k)
            (C_in)       |
                         +--- D2 ---+
                                    |
                                   gnd

Specifically: input capacitor (0.001 µF) blocks DC and AC-couples the probe tip. The diodes are arranged so that on the positive half-cycle, D1 conducts and charges the output cap (0.01 µF) positive; on the negative half-cycle, D2 conducts and the output cap continues to charge, with the ground side of the output cap pulled negative. The net DC across the output cap is approximately 2 × V_peak − 2 × V_diode_drop, doubled compared to the single-diode design. The 100 kΩ resistor sets the discharge time constant (τ = 100 kΩ × 0.01 µF = 1 ms — fast enough to follow voice modulation) and presents a high impedance to the DMM.

Construction. Lay the components out on the perfboard as compactly as possible — total component-to-component lead length under 10 mm. This is critical for high-frequency performance; any extra lead inductance becomes a series impedance that progressively shorts out the high-frequency response. Place the probe tip at one end of the perfboard, the BNC jack at the other end (or vice versa, with the BNC jack at the back so the test lead comes out the back of the enclosure). Solder the components directly to the perfboard with minimum lead trim. Mount the perfboard inside the knurled aluminum tube (the tube acts as both enclosure and RF shield); for the 3D-printed shell variant, add a layer of self-adhesive copper tape inside the shell for shielding.

The probe tip is whatever you’ve got — a sharpened copper nail epoxied into the end of the perfboard is fine, a telescoping antenna tip from a junk-box rubber duck antenna works well for the “extendable probe” feel.

Calibration. Connect the BNC to a DMM in DC volts mode (10 MΩ input impedance — standard for any modern DMM). Drive a calibrated sig-gen (bench sig-gen or a calibrated SDR with an external power meter) at 14 MHz to a known output level: 0 dBm = 0.224 V_RMS = 0.316 V_peak. Touch the probe tip to the sig-gen output and read the DMM. With a 0 dBm input at 14 MHz, expect a reading of around 0.2-0.25 V_DC. The multiplier (V_peak / V_DC_reading) at 14 MHz is the calibration constant for HF. Record it.

Repeat at 50 MHz, 144 MHz, 220 MHz, and 432 MHz. The multiplier rises gradually as frequency increases because the input cap progressively loads down the source and the diode forward drop’s contribution becomes more significant relative to the smaller peak voltages. Typical multipliers: 1.0 at 14 MHz, 1.1 at 50 MHz, 1.3 at 144 MHz, 1.7 at 432 MHz. Print or laminate the table and attach to the probe handle.

Usable range and accuracy. Calibrated, the Tayloe probe reads V_peak within ±1-2 dB from 1 MHz to about 500 MHz. Above 500 MHz the input cap (0.001 µF) becomes inductive (self-resonant series inductance dominates) and the probe response collapses. For >500 MHz work, use a smaller input cap (0.0001 µF) and smaller diodes (BAT15 or similar Schottky in SOT-323), with construction in ground-plane PCB style rather than perfboard.

The probe is most useful for: trimming wire antennas in the field (measuring the relative voltage at the high-Z end as you cut), checking PA stages on the bench, finding hot spots on parasitic-radiation problems (probe along a coax shield to find common-mode current), and as a sanity-check for any high-impedance RF voltage measurement where you don’t want to disturb a 50 Ω instrument.

13. Commercial buys

Sorted by price tier (USD, mid-2026):

Budget (~$30-$150). Diamond SX-200 cross-needle meter ($150) — the workhorse SWR/power meter for an HF station. Mini-Circuits TERM-50W-N ($30) for SWR-reference termination. MFJ-260C 1.5 kW oil-filled dummy load ($130) — the budget high-power dummy load (you supply your own mineral oil). Diamond SX-400 cross-needle for VHF/UHF ($170). MFJ-849 broad-coverage cross-needle ($130). ADF4351 AliExpress eval board with SPI ($15). F4HWN ADF4351 standalone signal generator with OLED ($45). Tayloe probe DIY kit (Far Circuits, $25 PCB-only).

Mid (~$200-$500). Bird 43 body used + 5 slugs ($400-$500 complete set) — the most flexible HF-through-UHF wattmeter. Daiwa CN-801HP ($260) for VHF/UHF cross-needle high-power. Diamond SX-1100 wideband ($310). HP 8657A signal generator used ($400-$600). Bird Termaline 8201 50 W dummy load used ($200-$300). BladeRF 2.0 micro xA4 ($530) — entry-level SDR sig-gen.

Premium ($500-$5,000). Bird 4391A digital wattmeter ($700-$1,500 used). HP 8648C/D wider-frequency sig-gen used ($1,000-$2,500). Bird Termaline 8401 500 W oil-filled dummy load used ($400-$600). Bird Termaline 8930 1 kW air-cooled ($1,000-$1,500 used). USRP B210 ($2,000) — laboratory-grade SDR sig-gen. Keysight FieldFox handheld combo (sig-gen + spectrum analyzer + VNA in one box, ~$15,000-$25,000 new; $4,000-$8,000 used for older models).

Reference / instrument-grade (~$10,000+). Agilent E4438C vector sig-gen used ($3,000-$5,000) for digital-modulation testing. R&S SMA100B premium reference sig-gen ($30,000+ new). Bird 8890 high-power air-cooled dummy load ($1,500 new). Narda SRM-3006 selective field-strength meter ($15,000-$30,000) for EMC compliance work.

Avoid. Cheap Chinese “QRP wattmeters” without published SWR data — these typically have undocumented insertion loss above 30 MHz and may not be 50 Ω at all (some are 75 Ω or have hidden mismatch). Generic AliExpress “1 kW dummy loads” with unspecified resistor architecture — these often use cheap wirewound resistors that work at DC but are dangerously reactive at RF, and many “1 kW” units cannot dissipate more than 100 W continuous without melting. Used HP 8657A/B units sold with NiCd battery still installed — assume the battery has leaked and the PCB is damaged; ask the seller for photos of the cleaned PCB before buying.

14. Common gotchas and myths

“The Bird 43 reads accurately to 1 W on a 250 W slug.” False — the 5% of full-scale accuracy spec means a 250 W slug reads ±12.5 W everywhere on its scale, so a 1 W indicated reading is somewhere between 0 W and 13.5 W. Use the right slug for the power level; the practical rule is full-scale within 4× of the measurement.

“I’ll measure SWR through the tuner.” This is one of the most common confused-measurement mistakes. The SWR meter must be between the transmitter and the tuner input — measuring there shows you what the transmitter sees, which is what matters for the transmitter’s protection circuits and tube/transistor health. An SWR meter on the output of a properly-tuned tuner reads 1:1 regardless of what the antenna actually presents, because the tuner has transformed the antenna’s impedance to 50 Ω at its input and to whatever-the-antenna-needs at its output. The “real” SWR on the coax between the tuner and antenna is what it is; the meter at the rig side of the tuner doesn’t see it because the tuner is hiding it.

“Sig gen + oscilloscope = good RF measurement above 100 MHz.” Only with active FET probes. A passive 10× probe has 10-15 pF of input capacitance, which forms a 10-Ω-or-less reactance at 1 GHz — the probe shorts out the signal. Active probes (LeCroy ZS1500, Tektronix P7600) cost $2,000-$10,000 and are required for any meaningful >100 MHz scope measurement. For RF amplitude work above 100 MHz, use a spectrum analyzer (see Vol 27 (Spectrum analyzers)) or a calibrated power meter, not a scope.

“Used HP 8657 is good as new.” Partial truth — the chip-level circuitry on an 8657 is long-lived and stable. The wear-out items are: the reed-relay step attenuator (100,000-cycle life; if the previous owner used it as a workhorse for ten years it’s at end-of-life and starts skipping steps or sticking); the GPIB interface card (often fails on units that have sat unused for years); the NiCd battery for last-state-memory (leaks alkaline electrolyte that eats PCB traces). Always pull the NiCd before powering a used HP instrument. Always test the attenuator across its full range before paying for the unit.

“A dummy load is purely resistive.” True at DC. At RF, the resistor leads have inductance and the package + heatsink geometry has capacitance. A nominally 50 Ω dummy load is 50 Ω at one specific frequency and somewhat different at others. The cheap MFJ-260C cantenna is 50 Ω at 5 MHz, drifts to 45-65 Ω across 1-30 MHz, and is well past 1.5:1 SWR by 100 MHz. The Mini-Circuits TERM-50W-N maintains 1.2:1 SWR to 8 GHz because it’s a SMD design with sub-millimeter lead lengths. Always sweep your dummy load with a NanoVNA at the frequencies you intend to use it; “50 Ω dummy load” is not a permanent property.

“A field-strength meter measures EIRP directly.” No, it measures field strength at a single point in space. EIRP is computed from field strength using the Friis equation, which requires knowing the distance, the receive-antenna gain, and the propagation environment (free space or two-ray-ground). Wrong assumptions on any of these — wrong distance estimate, wrong gain assumed for the reference antenna, polarization mismatch, ground-reflection interference — give EIRP estimates that can be off by 10 dB or more. EIRP measurement is a discipline; treat it as a multi-distance, multi-azimuth average rather than a single-point reading.

“My SDR’s TX output is calibrated.” No SDR’s TX output is calibrated to bench-instrument standards. The HackRF, BladeRF, and USRP all have output levels that vary ±3-5 dB with frequency across their range, vary with software-controllable gain settings (which have non-monotonic actual-gain-vs-setting curves), and vary unit-to-unit. For amateur antenna work this rarely matters because relative measurements dominate. For absolute-power measurement, calibrate the SDR against a power meter at the frequency of interest, or use a bench sig-gen.

“I can measure my transmitter’s harmonic content with a wattmeter.” A through-line wattmeter integrates total RF power in its band, regardless of harmonic content — a transmitter putting out 100 W at the fundamental and 10 W at the second harmonic reads as 110 W on a wattmeter whose slug covers both frequencies, or 100 W on a slug that only sees the fundamental. To measure harmonics, you need a spectrum analyzer (Vol 27) or a tunable-bandpass filter + power meter combination.

15. Resources

Authoritative texts and application notes:

  • Bird Electronic Corporation, Model 43 Operation Manual — the original. PDF available from https://www.birdrf.com
  • Bird Electronic Corporation, Slug Selection Guide — full catalog of slug models with frequency/power matrix
  • ARRL Antenna Book (25th+ edition), Chapter 27 “Antenna and Transmission-Line Measurements”
  • ARRL Handbook (current edition), chapter on test equipment
  • Hewlett-Packard Application Note 64-1, “Fundamentals of Microwave Power Measurement”
  • Hewlett-Packard Application Note 1303, “Spectrum Analysis Basics” (cross-reference for Vol 27)

Datasheets and chip-level references:

Tayloe probe original source:

  • Dan Tayloe N7VE, “An Improved RF Probe,” QST, December 1995 — the original description of the twin-Schottky topology
  • W7ZOI’s notes on RF probes, EMRFD (Experimental Methods in RF Design) by Hayward, Campbell, and Larkin

ADF4351 firmware and software:

EIRP and field-strength measurement references:

  • FCC Office of Engineering and Technology Bulletin 65 (OET-65), “Evaluating Compliance with FCC Guidelines for Human Exposure to Radiofrequency Electromagnetic Fields” — cross-reference for Vol 31 (Regulatory & RF safety)
  • ARRL RF Exposure and You — practical worksheet-driven calculations
  • IEEE Std C95.1, “Safety Levels with Respect to Human Exposure to Radio Frequency Electromagnetic Fields”

Vendors:

Sibling volumes for cross-reference: