In my role coordinating rush orders for RF and microwave test equipment, I've handled more than 200 urgent requests in eight years. The question I hear most often is: which Anritsu should we order? The two that come up most often are the Anritsu MG3694A and the Anritsu ML2438A. My first answer is always the same: it depends.
Not because I'm dodging the question. Because one of these instruments creates signals and the other measures power. If you don't know which one you need, you can end up with an expensive paperweight. And when your test window is 48 hours, the return process isn't your friend.
Here are the three scenarios I see most often, and the instrument I'd order for each.
The MG3694A is a microwave signal generator. In plain English, it creates RF energy at a controlled frequency and level. You connect it to a device to see how that device behaves. It's the tool you reach for when you're testing receiver sensitivity, checking a filter response, or characterizing an amplifier.
An example: an R&D lab I support was validating a 10 GHz transceiver. The receiver looked like it had a sensitivity problem. We checked the signal source first. The existing source had dirty harmonics. When we switched to the MG3694A, the same receiver passed. The problem was the test setup, not the device.
So if your goal is to put a known signal into something, the MG3694A is your side of the fork. Don't buy a power meter first, no matter how nice the spec sheet looks.
One caveat: the MG3694A tells you about the signal that goes into your device. It doesn't tell you what happens after the device. For that, you need a measuring instrument. If your test involves both, don't stop at the source.
The ML2438A is a power meter. It tells you how much RF power is actually present at a point in your system. This is usually where someone asks: can't I just use the best multimeter I can find?
Look, a multimeter is a fine tool. I keep a good one on my bench. But for RF at 2 GHz or 3.5 GHz or higher, a multimeter is not the right instrument. A power meter is designed to work with calibrated sensors that are characterized at those frequencies. The best multimeter in the world won't give you a trustworthy number if its detector and frequency response aren't rated for the signal you're measuring. I've learned that lesson the expensive way.
Last March, a field engineer was troubleshooting a remote antenna. He called and said the transmitter was dead. He had a multimeter and had measured DC at the power supply. I asked if he had a power meter. He didn't. We rushed him an ML2438A with a matching sensor. It turned out the transmitter was fine; the real problem was a damaged jumper cable with excessive insertion loss. He would have spent another day replacing a transmitter that wasn't broken.
If your question is how much power is actually here, choose the ML2438A. It answers the question a multimeter can't.
The ML2438A isn't a spectrum analyzer either. It won't show you harmonics or spurious emissions. It gives you one number: power. That's a feature when you need a confident, repeatable measurement for calibration or acceptance testing.
Sometimes the decision isn't source or meter. It's both.
Here's a recent example. A startup asked us for same-day delivery of two Anritsu instruments. They were testing a transparent smartphone prototype with a 7.1-inch display before a demo. The display was optically transparent, but it still affected RF radiation. They needed the MG3694A to generate a known signal for sensitivity testing, and the ML2438A to measure how much power actually left the antenna port.
I went back and forth with them on whether they could get away with one instrument. They couldn't. Oh, and I should add that they initially planned to borrow a generator from one lab and a power meter from another. That plan fell apart after we added up the shipping delays. In a transparent smartphone, the RF path is determined by materials that aren't in a standard phone. You need to both inject a signal and verify the output. We delivered both units in one box. They finished the test in one evening.
The surprise wasn't that transparent smartphones create new RF challenges. It was that the biggest delay had nothing to do with the hardware. It was the 20 minutes spent deciding what to order. Two instruments in one shipment beat two separate loans from two different labs every time.
If you're testing any device with non-standard materials, trace the RF path before you order. Ask where the signal comes from, where it enters the device, and where you expect it to leave. Write it down. That sketch will tell you which instruments you need. This is the checklist approach I use on every rush order, and it's saved us more rework than I can count.
Before you place an order, answer three questions.
One more pointer: if this is for a production line, redundancy matters. A single generator or meter can become a bottleneck during a rework. I've watched a failed power sensor stop a production line for a whole shift. Having a spare sensor or a second unit on hand is often cheaper than the downtime.
The MG3694A and the ML2438A are not competitors. They're two halves of an RF measurement loop. If you can only order one today, answer the create-or-measure question first. That answer will tell you which box to open.
And when you're in a hurry, remember this: 5 minutes of verification beats 5 days of correction. The instrument you choose is only as good as the question you ask before choosing it.