Measuring harmonics would, at first, seem like a simple task for a tuned receiver such as the one in a Roos tester. Simply set the receiver to the harmonic frequency and measure. However, there are many pitfalls that can cause harmonic measurement to be a difficult task. This document describes some of the issues associated with harmonic measurement and provides solutions for those issues. Note that the equipment referenced are Roos testers, however the issues described here are applicable to any harmonic measuring equipment.
Spurious-Free Dynamic Range:
The biggest challenge is maximizing, or in some cases overcoming the spurious-free dynamic range of the tester.
Spurious-free dynamic range is not simply dynamic range. Dynamic range is the largest signal the tester can see minus the smallest signal it can see (limited by the noise floor). On a Roos tester, that is typically 140 dB, depending on frequency and setup.
Spurious-free dynamic range is the ability to measure a small signal in the presence of a large signal. If the large signal is too large, it will cause the tester's active components to generate harmonics, which can obscure the DUT's harmonics. On a Roos tester, the maximum spurious-free dynamic range is typically 60 dB, dependent on the signal's power and frequency, and the settings of the tester.
When measuring harmonics, if you have too much power into the measuring equipment, the equipment will generate its own harmonics. If you have you have too little power, you won't see the harmonic below the noise floor of the equipment. There is a 'sweet spot' where the tester's harmonic is just at the noise floor and you get the maximum spurious-free dynamic range.
On the Roos tester, with the proper tester setup, this occurs at about -10 dBm into the tester. With these conditions, and the proper tester setup, we can achieve approximately -60 dBc spurious-free dynamic range. However note that, at 60 dB, the tester is measuring at the noise floor, so the uncertainty will be large. For good uncertainty on measurements, the usable spurious-free dynamic range is typically 50-55 dBc.
Note that, on a spectrum analyzer, you can get a greater spurious-free dynamic range because the spectrum analyzer has resolution bandwidth filters that allow you to reduce the noise floor. The cost of this is, of course, speed. For low harmonics, it can take several seconds to get a measurement on the spectrum analyzer, which is why we don't normally recommend them for production measurements. Also, you need to be aware of cable loss with the spectrum analyzer. It's going to be twice as much at the harmonic frequency as at the fundamental. Depending on the amount of loss, this will make the spectrum analyzer's measurement appear better.
For best results on the tester, you need to adjust the settings, but a good starting point is:
receive path: Direct receive. (Although, surprisingly, the s-parameter path might give better results by lowering power into the testhead's LNA. You have to try both and see which is best.)
receive atten: 10 dB
IF Bandwidth: 7 kHz
Use the Receiver's 'Measure Voltage' button with a lot of averages (like 500) to lower the noise floor.
Overcoming Spurious-Free Dynamic Range
Sometimes the ability of the tester alone is not enough. Then we have to add a filter into the fixture, typically as shown:
The fundamental is measured with the thru path, and the harmonic is measured through the high pass filter. The filter lowers the fundamental into the tester so that the spurious-free dynamic range of the tester is no longer an issue.
A splitter is used in the front of the path because a switch would generate its own harmonics. With this configuration, the splitter doesn't generate harmonics. The switch won't generate harmonics in the filter path, because the fundamental is suppressed. If the switch makes a little harmonic when you're measuring the fundamental, it doesn't matter.
The fixture can be calibrated so that the test plan accurately measures the powers through the losses of the two paths.
The through path has 6-10 dB loss. If you can tolerate this loss, then this technique can be done with an electronic switch.
Occasionally, the splitter and electronic switch have too much path loss. In that case, they must be replaced with mechanical switches as shown:
Although low-loss, the mechanical switches have a limited number of cycles. A typical mechanical RF switch will have a life of 5 million cycles.
If you're measuring a broadband device (that can pass the harmonic), then you need to worry about the source harmonic. The Roos sources are quite good, with about 60 dBc typical harmonic. If your device would naturally filter that out, then the source harmonic is not an issue. Otherwise, you need a low pass filter on the device input to kill the source harmonic.
A lot of amps have a large harmonic internally, then use a filter to lower the harmonic at the output. We've found that these devices typically will leak harmonic power out the other pins of the part. For example, it's not unusual for a 1W amp to have +15 dBm of harmonic power floating around inside the part, just waiting to leak out anywhere it can. At Roos, we ALWAYS put RF filters, or big resistors, on all the DC and control pins to keep harmonic power off of those lines and out of the fixture.