If you are about to buy a signal generator, do not begin with a model number. Begin with the signal that must arrive at the device under test (DUT): its carrier frequency, level, modulation, occupied bandwidth, timing, phase relationship to other channels, and required spectral purity. Only then should you translate the application into generator specifications and options.
This is the practical difference between a generic signal generator buying guide and an engineering selection process. A generator that reaches the right maximum frequency can still be unusable because it lacks modulation bandwidth, output power at the top of the band, low enough phase noise, pulse performance, coherent channels, GNSS software, fading, or the required reference and trigger interfaces.
Short answer: choose the minimum complete configuration that can reproduce the waveform your DUT must see under the worst test condition — not the instrument with the largest headline specifications.
Start with the signal, not the catalogue
Before comparing instruments, write the stimulus in one sentence. Examples:
- “A clean CW carrier at 18 GHz, +20 dBm at the DUT, with very low close-in phase noise.”
- “A Wi-Fi 7 vector signal at 6 GHz with 320 MHz channel bandwidth and enough EVM margin for receiver testing.”
- “Four GNSS constellations, at least 12 satellites, with multipath and fading.”
- “Two phase-coherent RF channels for beamforming verification.”
- “Pulsed X-band stimulus for radar receiver sensitivity and blocking tests.”
This sentence usually determines whether you need an RF/microwave signal generator, a vector signal generator, a GNSS simulator, a multi-channel coherent source, or a more specialized platform.
For the deeper technical background on microwave versus vector generators, clock sources, phase noise and modulation architecture, read our complete signal-generator technical guide.
Which type of signal generator do I need?
Pure CW, swept RF, LO stimulus or very low phase noise
For component characterization, mixer LO drive, local-oscillator substitution, receiver blocking and many radar tasks, an analog RF or microwave signal generator is often the most economical and technically appropriate choice. Priorities are normally frequency range, spectral purity, phase noise, output level, switching speed and pulse performance.
Digitally modulated RF, QAM, OFDM, Wi-Fi or 5G
If the DUT must see a defined I/Q waveform, choose a vector signal generator (VSG). The decisive parameters become modulation bandwidth, waveform memory, baseband/sample architecture, EVM, supported standards, fading and channel count. A wide RF frequency range alone is not enough.
GNSS receiver testing
A normal VSG can generate prerecorded or custom GNSS-like waveforms, but full receiver validation may require a real GNSS simulator. Check constellations, satellites/channels, ephemeris control, trajectory, atmospheric models, multipath, fading, interference, synchronization and scenario automation.
Phase-coherent or MIMO testing
For phased arrays, MIMO, direction-finding and multi-channel receivers, verify that channels are not merely simultaneous but phase coherent with deterministic relative phase. Clock and LO sharing, phase repeatability, trigger alignment and calibration matter.
Requirement worksheet: what to know before you compare products
| Requirement | Question to answer | Datasheet term to compare |
|---|---|---|
| Carrier | Lowest and highest RF frequency? | Frequency range |
| Level | Minimum and maximum power at the DUT? | Output power range, level accuracy |
| Purity | Weak signal near a strong carrier? | Phase noise, harmonics, spurious |
| Modulation | CW, AM/FM/PM, pulse, I/Q? | Modulation modes |
| Bandwidth | Widest instantaneous waveform? | RF modulation / I/Q bandwidth |
| Timing | Pulse width, rise/fall, trigger? | Pulse modulator, timing resolution |
| Channels | One, two, four, coherent? | Channel count, phase coherence |
| Standards | Wi-Fi, 5G NR, LTE, GNSS, etc.? | Software/application options |
| Synchronization | Common 10 MHz, trigger, clock? | Reference I/O, trigger I/O |
| Automation | Bench or automated test? | SCPI, LAN, APIs, list/sequence mode |
Frequency range: buy what the measurement really needs
The highest required carrier frequency is only the first boundary. Add margin for harmonics, blockers, image-frequency tests, frequency conversion and future variants only when those are realistic parts of the test plan.
A common mistake is buying a 40 or 44 GHz generator for an application that will never leave 6 or 8 GHz. Frequency range is one of the largest cost drivers in RF instrumentation. The opposite mistake is forgetting that an 18 GHz DUT may need blockers or LO frequencies above 18 GHz.
Always verify whether quoted frequency coverage is native or depends on an external multiplier, frequency extender or different output connector.
Output power: check it at the frequency that matters
“Maximum output power” is not a single universal number. Available level often changes with frequency and with options. If your requirement is +23 dBm at 20 GHz, compare the guaranteed or specified level near 20 GHz — not a best-case value at 1 GHz.
Work backwards from the DUT:
Generator level = required DUT level + cable loss + attenuator loss + fixture loss.
If the DUT needs -110 dBm, also check low-level accuracy, attenuator behavior and leakage. If it needs high power, consider an external amplifier and whether the generator can still provide the required spectral purity.
Phase noise and spurious: the hidden selection criteria
For receiver sensitivity, reciprocal-mixing tests, radar, frequency conversion and high-order modulation, phase noise may be more important than maximum frequency. Search results and vendor buying guides repeatedly surface phase noise, spurious performance and signal purity because these specifications determine whether the source itself becomes the measurement limitation.
Compare phase-noise values at the offsets relevant to your test. A single number at 1 MHz offset is not sufficient if your blocker is 10 kHz away from the wanted channel. Also check non-harmonic spurs and harmonics under the actual output-level conditions.
For metrology background, see the NIST Phase Noise Metrology Group, the PTB work on highly stable microwave signals, the NPL good-practice guide to phase-noise measurement, and the National Institute of Metrology of China microwave and millimetre-wave laboratory.
Modulation bandwidth: do not confuse channel bandwidth with generator margin
If you need Wi-Fi 7 at 320 MHz, buying exactly 320 MHz of modulation bandwidth gives little engineering margin. The required generator bandwidth depends on the waveform, filters, oversampling, adjacent-channel content and the measurement you want to make at the DUT.
For 5G NR, Wi-Fi, satellite and proprietary OFDM, verify:
- usable RF modulation bandwidth;
- sample rate and waveform length;
- EVM at the required carrier frequency and bandwidth;
- crest-factor handling;
- digital standards and waveform software;
- fading and MIMO features if required.
This is why “which vector signal generator do I need?” cannot be answered from frequency range alone.
EVM: the source must be better than the DUT requirement
For high-order QAM, the generator contributes to the total EVM budget. If the DUT must be verified at a stringent EVM limit, the generator needs enough margin that its own imperfections do not dominate the result.
The practical procurement question is not “does it support 4096-QAM?” but “what EVM does the configured generator achieve at my carrier frequency, bandwidth, level and selected options?”
Pulse and radar requirements
For pulsed RF, check more than the presence of a “pulse modulation” checkbox. Important parameters include minimum pulse width, rise/fall time, on/off ratio, pulse delay, repetition rate, pulse train/list capability and trigger latency. Radar and EW work may also require frequency hopping, fast switching, chirps or phase-coherent pulse sequences.
GNSS: when a vector generator is not enough
For GNSS receiver tests, list the scenario rather than the waveform format:
- GPS, Galileo, GLONASS, BeiDou;
- number of simultaneously simulated satellites;
- static or moving receiver;
- multipath and fading;
- atmospheric effects;
- interference and jamming;
- multiple RF outputs or antenna elements;
- scenario duration and automation.
If those requirements are central, compare purpose-built GNSS simulation capability rather than assuming any VSG with GNSS software is equivalent.
Hardware, software and accessories are part of the instrument
The base generator is rarely the complete purchase. A technically correct comparison must include options that enable the required bandwidth, modulation, low phase noise, pulse function, standards, GNSS, fading, MIMO, memory, coherent channels, high output power and remote-control features.
Also include external attenuators, amplifiers, combiners, phase-stable cables, adapters and calibration accessories. Two instruments with the same front-panel model name may be radically different once their installed options are considered.
Four worked selection examples
Example 1 — Receiver sensitivity at 6 GHz
Requirement: weak CW or modulated stimulus around 6 GHz down to approximately -120 dBm. Prioritize low-level accuracy, leakage, phase noise, spectral purity and a controlled attenuation path. Wide modulation bandwidth is unnecessary if the receiver test is CW.
Example 2 — Wi-Fi 7 receiver development
Requirement: 6 GHz carrier, 320 MHz channel, high-order QAM. Prioritize vector modulation, bandwidth above the occupied channel, low EVM, waveform software, enough memory and possibly fading/MIMO. Buying an ultra-low-phase-noise microwave generator without I/Q capability would not solve the problem.
Example 3 — X-band radar receiver
Requirement: 8–12 GHz, pulse modulation, low phase noise, precise timing, blocking/interference scenarios. Prioritize pulse performance, phase noise at relevant offsets, fast switching and synchronization. Digital communication standards may be irrelevant.
Example 4 — GNSS navigation receiver
Requirement: GPS + Galileo + GLONASS + BeiDou, 12+ satellites, fading and multipath. The critical decision is simulator capability and scenario realism, not merely RF frequency coverage.
Common overspecification mistakes
- Buying the highest frequency range “for future proofing” without a credible future use.
- Paying for multi-gigahertz modulation bandwidth for narrowband receiver work.
- Buying multiple channels when only sequential tests are required.
- Selecting every wireless-standard license instead of the standards actually used.
- Paying for ultra-low phase noise where the DUT and measurement do not benefit.
Common underspecification mistakes
- Checking only maximum frequency and ignoring output power at that frequency.
- Forgetting cable/fixture loss when defining required output level.
- Buying a VSG with insufficient usable modulation bandwidth.
- Ignoring source EVM for high-order modulation.
- Assuming GNSS software automatically means full multi-constellation simulation.
- Forgetting coherent channels, reference I/O or trigger requirements.
Build a procurement requirement before asking for a quote
Use this checklist:
- Carrier frequency range.
- Required level at the DUT and expected path loss.
- CW / analog / pulse / vector modulation.
- Maximum occupied and required generation bandwidth.
- Phase-noise offsets that matter.
- Allowed spurious/harmonic level.
- EVM target if digitally modulated.
- Pulse timing requirements.
- Number of channels and coherence.
- Required standards, GNSS, fading or MIMO options.
- External-reference and trigger interfaces.
- Automation interface and sequence requirements.
Once those are known, comparing instruments becomes much more deterministic.
Related measurement instruments
A generator is only one side of many test benches. For transmitter and modulation work, see how to choose a spectrum analyzer. For component S-parameters and impedance, see how to choose a VNA. For traceable absolute RF power, see how to choose an RF power meter and sensor.
Questions people ask before buying a signal generator
Which signal generator do I need for receiver testing?
For sensitivity and blocker tests, prioritize low-level accuracy, phase noise, spectral purity and the required modulation. For modern wireless receivers, also verify modulation bandwidth and source EVM.
Can an arbitrary waveform generator replace a vector signal generator?
Sometimes at baseband or low IF. At RF, a VSG integrates upconversion, calibrated output level, RF filtering, frequency synthesis and usually better synchronization and standards support.
How much frequency headroom should I buy?
Enough for realistic harmonics, blockers, conversion products and planned product generations. Headroom without a test requirement is expensive insurance.
What is more important: output power or phase noise?
It depends on the application. High-power component stimulation may be power-limited; weak-signal receiver and close-in blocker tests are often phase-noise-limited.
Independent technical references
- Germany: Physikalisch-Technische Bundesanstalt (PTB) — RF and microwave measurement publications
- United Kingdom: National Physical Laboratory (NPL) — phase-noise measurement guide
- United States: NIST — Phase Noise Metrology Group
- China: NIM — Microwave and Millimeter Wave Metrology
From requirement to independent comparison
Once you can describe the test in engineering terms, the next step is not to browse catalogues one by one. WaveMatch is designed to compare validated technical data and complete configurations across vendors from the requirement you provide.
Example request: “I need a vector signal generator for Wi-Fi 7 at 6 GHz, 320 MHz bandwidth, very low EVM, and two phase-coherent channels.”
Request a WaveMatch demo and enter the requirement in natural language.