Picking the Right Signal Generator Isn't a One-Size-Fits-All Thing
If you've been searching for a Keysight RF signal generator—say, models in the PSG, MXG, or EXG family—you've probably noticed the specs are different, and so are the price tags. There's no single 'best' generator; the right choice depends on what you're actually testing.
Let's break this down by the three most common scenarios I see in my work (and in our own lab at a mid-sized RF design firm). We'll cover when to lean on an old workhorse, when you need a feature-packed modern box, and when a basic generator actually makes more sense than a high-end one.
Scenario A: The 'It Just Needs to Work' Workhorse
Who this is for: Production test, RF troubleshooting, or replacing an aging unit where the core specs (like phase noise at 1 GHz) are 'good enough' and you need something reliable, serviced, and available.
Think of a used Keysight 8648C (note: these are older, but still common) or a newer EXG X-Series (like the N5172B). The core job here is straightforward: generate a clean CW signal or basic AM/FM, and do it day in, day out.
What you gain: Low upfront cost, proven reliability, and a deep pool of technical knowledge online (for the older models). For a company running a production line that doesn't need to change frequency or modulation every 10 seconds, this is the right call.
The hidden catch (note to self: check calibration cycles): Old generators drift. A 10-year-old 8648C might be off by 25-50 ppm in frequency accuracy unless it's been recently calibrated. That's fine for a simple power test, but terrible for any kind of filter alignment. In our lab, we bought a used 8648C for $1,200 (in 2022) and spent another $400 getting it calibrated. It now works flawlessly for our 1-2 GHz spurious checks.
Scenario B: The 'I Need Every Feature' Lab Standard
Who this is for: R&D, design verification, or anyone testing modern wireless standards (5G NR, Wi-Fi 6/7, etc.). This is where the MXG X-Series (N5182B) or the PSG (E8267D) lives.
This is a different beast entirely. These generators aren't just about generating a signal; they're about generating a precise, repeatable, modulated signal that mimics what the device will see in the real world. The PSG, for example, can handle complex I/Q modulation up into the mmWave bands.
What you gain: Superb phase noise (think -140 dBc/Hz at 1 GHz offset for the MXG), fast frequency switching (under 100 µs), and the ability to generate custom waveforms. You can literally load a 5G NR test model from PathWave and have it come out of the RF port.
The assumption most people make (and it's wrong) is that the more expensive generator automatically gives better measurements. Actually, it gives different measurements. In Q3 2024, I compared a PSG vs an EXG on a simple harmonic distortion test at 6 GHz. Both gave the same result within 0.1 dB. The real difference is with complex modulation or very tight receiver testing.
Causation reversal alert: People think expensive generators deliver better results because they're expensive. Actually, they cost more because they solve a specific problem (accurate modulated signals, low close-in phase noise for EVM testing) that R&D teams are willing to pay for. If you don't have that problem, you're paying for a solution you don't need.
Scenario C: The 'I'm on a Budget (Or Need a Backup)' Smart Choice
Who this is for: A startup, a teaching lab, or a team that needs a second generator for basic sweeps but can't justify $10k+ for the latest model. This is where a used Keysight E4438C ESG (a classic) or a less-common model like the N9310A makes sense.
This is the 'value' play, but it requires some due diligence. The E4438C, for example, is a workhorse from the early 2000s, but it can still generate vector modulation (like QPSK, 16QAM) if you have the option boards. They pop up on the used market for $3k-5k.
The legacy myth here: The old thinking was 'digital generators from 10 years ago can't handle modern signals.' This was true when the software and digital backend were limited. Today, with software like PathWave (or even old versions of Signal Studio), you can program an old ESG to output a perfectly decent LTE waveform (as long as the options are installed). The hardware is usually fine; the limiting factor is whether the I/Q bandwidth is wide enough. In our case, we kept an old E4438C for basic EVM checks at 20 MHz bandwidth, and it works fine for 90% of our legacy products.
A real example from our lab (in 2023): We needed a signal generator for a one-week project on satellite payload testing. Buying a new N5172B was not in the budget. We rented an N5182B MXG for 4 days for $900 from a local rental house. It was the exact right call—we got the specs we needed for the critical measurement, and didn't add to the capital expense.
How to Decide Which Scenario You're In
Here's a quick checklist I use with our engineers when they ask for a new generator:
- Are you testing a receiver's EVM below -35 dB? → Go to Scenario B (MXG or PSG). You need low phase noise and high I/Q bandwidth.
- Do you just need a stable CW source for a filter or amplifier test? → Scenario A (EXG or older workhorse).
- Is your budget under $5k? → Scenario C (used ESG or rental).
- Do you need to generate a signal to test a 5G mmWave device at 28 GHz? → Scenario B (PSG with a upconverter). Don't even think about other options.
Pricing is always a moving target. For reference, a new Keysight N5182B MXG RF Vector Signal Generator (100 kHz to 6 GHz) lists around $22,000 (as of early 2025; verify with your Keysight rep). A used E4438C in good condition can be found for $3,000-$5,000 on sites like eBay or TestWorld. Calibration status often matters more than the age of the unit.
The Bottom Line
Don't buy the most expensive generator. Don't buy the cheapest one. Buy the one that matches your workflow and test requirements. If you're doing R&D on next-gen comms, the MXG is worth every penny. If you're running a production line on a mature product, an older or cheaper model will serve you just fine. The most expensive mistake is buying a $20k generator for a job a $3k one can do.
