Noise power ratio measurement

Intermodulation distortion (IMD)

One issue with multicarrier signals is that these carriers can mix with each other, creating intermodulation distortion (IMD). This occurs primarily in amplifiers, but it can happen in any component in the signal chain.

There are two problems that can be caused by IMD:

• If the products fall outside of the carrier channel, they can create interference and lead to a decline in signal-to-noise power ratio (SNR) in adjacent channels.
• IMD can fall within the carrier channel itself and degrade its own SNR. This is a common concern in satellite systems, since these links do not normally have direct neighbors that could be affected by IMD falling outside the channel.

The classic method for measuring IMD is a third-order intercept (TOI) measurement. This methodology uses two separated input tones and is the standard way of quantifying device linearity. However, traditional TOI measurements are not applicable to multicarrier systems with many contiguous carriers. Instead, IMD in multicarrier systems is typically quantified with the noise power ratio (NPR).

Noise power ratio (NPR)

NPR is tested by generating a wideband signal that contains an unused channel - a deep “notch.” The signal width is usually the same as the nominal channel width, and a deep notch is created somewhere within the channel. This notched signal is then input to the device-under-test (DUT), e.g., an amplifier. IMD will cause the notch to fill with noise; this is sometimes referred to as spectral regrowth. You then measure this new notch depth using a spectrum analyzer.

NPR is the ratio, in dB, of the carrier power to the notch power. A higher NPR or “less noise in the notch” means lower IMD. Note that the size, location and number of notches are chosen based on the given DUT or application, and NPR tests are often repeated with different notch sizes, locations and power levels.

Measuring the depth of the notch is relatively easy, so the real challenge in NPR testing is the creation of the test signal. There are two main ways of generating a test signal when measuring NPR:

• With an analog signal generator or noise source combined with a notch filter.
• With a vector signal generator.

You want the test signal to be as close as possible to the “real” signal that passes through the DUT. For example, you want the same (or similar) modulation and peak-to-average ratio, since these will affect the level of intermodulation generated by the device. The notch should also be a deep as possible, especially if the IMD level is low. Depending on the test requirements, you may need more than one notch or notches with different widths. Finally, you may need to change the position of the notch to measure the IMD at different points.

Generating test signals with an analog signal generator

Let’s start with looking at how NPR test signals are created with an analog signal or noise generator. There are largely two steps:

• Generate band-limited noise, ideally with the same width as the channel you are emulating.
• Use a notch filter to create the notch.

The use of traditional, analog notch filters is problematic because deep, narrow notch filters are hard to make and relatively expensive. These filters are also usually not tunable, i.e., they have fixed notch stop band, you need a separate notch filter for each notch frequency.

One way to get around these limitations is to create the test signal at one frequency before converting it to the test frequency. However, this approach has its own problems: the frequency conversion process creates intermodulation distortion and degrade the notch depth.

Analog NPR test signals also suffer from a serious drawback: they are not very realistic. A broadband noise signal generally will not create the same levels of intermodulation distortion as the actual modulated signal, especially when the modulated signal has a high crest factor or peak-to-average ratio.

Generating test signals with a vector signal generator

Using a vector signal generator is a more flexible and accurate way to create NPR test signals. With a vector signal generator, you can use a realistic test signal, i.e., a signal that is much closer to the ones normally used with the DUT. It’s also easier to create multiple notches, and you can typically get better notch quality, i.e., steep, deep and narrow.

There are several ways to create notches in the vector signal domain:

• Offline processing: The notch is statically added to an arbitrary waveform file using a tool like MATLAB.
• Multi-carrier continuous wave (CW): Large numbers of unmodulated carriers are used to “fill up” the test signal bandwidth, minus the notches.
• Digital notch filter: Digital notch filters are used to create notches in any type of signal.

Using digital notch filters is the preferred method for creating NPR test signals. They offer the highest level of realism because they can be used with the actual modulated signal types carried by the system. They are also completely flexible, so you can create multiple notches of arbitrary width and position. And finally, they can be defined, updated and enabled or disabled in real-time, with no need for offline pre-calculation of waveforms.