This blog post is part of a collection, click here for links to the entire set, and a review of the products.

Introduction

The Agilent N9322C Spectrum Analyzer (SA) has some built-in capabilities to examine modulated signals. The post here examines Frequency Modulation (FM) analysis capabilities. The capabilities are useful for developing, testing and aligning FM systems.

 

The photo here shows a tone being transmitted using frequency modulation.

452hz.jpg

 

To see some more photos of the SA itself, click here.

 

What is Frequency Modulation (FM)?

In a nutshell, it’s the process of taking a signal of one frequency (known as the carrier frequency) and adjusting its frequency over time, in proportion to the amplitude of another signal known as the baseband signal (the baseband signal would typically be speech or music for example). The resultant signal is frequency modulated.

 

So, for example, if you had a voltage controlled oscillator and you drove the voltage control input with (say) a microphone such that the frequency would deviate up or down by an amount, then the output would be an FM signal (this assumes the VCO output frequency varies linearly with the voltage input).

 

The diagram here shows a sine wave baseband signal, and the resultant FM signal in blue (source: Wikipedia, Berserkerus)

Amfm3-en-de.gif

 

Spectrum Analyzer FM Analysis

The screenshot below shows the spectrum of an FM signal as captured by the SA. This happens to be a 1kHz tone, FM modulated on a carrier frequency of 145.5MHz as an example.

mod-analysis-fm-sa-spectrogram2.jpg

 

The SA has the capability to demodulate this input to recover the original tone. There is a built-in speaker on the SA and it is possible to listen to it (above the noise of the fan on the SA! – it’s not particularly quiet).

 

More importantly from my perspective, what is really great is that the SA can show a view like this:

mod-a.png

 

Here it is possible to see the trace of the demodulated 1kHz tone (in the stats shown circled, the PC client software appears to round down the value to 999.99Hz, whereas the spectrum analyzer on-board display rounds up to 1.00kHz). The SINAD value here is of extreme interest to engineers, because it will show you if the communication is likely to be intelligible to the user. Here the signal is very strong and therefore the SINAD value is high. In a more normal test scenario, one would seek 10 or 12dB SINAD by dropping the input signal to a receiver progressively. Once this threshold is reached, the input level would indicate the sensitivity of the receiver device.

Normally the value is directly measured from the audio output of a receiver, but this requires specialist equipment. The fact that the SA provides a computed SINAD value from its internal demodulation is fantastic and will be extremely useful.

 

The THD figure isn’t right; it may be missing the negative sign. I’ve asked Agilent.

 

The other very useful figure that is calculated is the frequency deviation. This is useful to know if a transmitter is adjusted correctly and is not trampling across adjacent channels (the FM spectrum spreads to infinity, but most of the power is concentrated in a certain bandwidth as will be seen later). In this example, it can be seen that the signal had +-2.5kHz frequency deviation.

 

The information in the screenshot also shows that the carrier frequency was not exactly 145.5MHz, but had a small offset of 26Hz shifted left. This is because the signal source does not have an oven or any temperature compensation (I could have corrected for this by programming an offset, but I didn’t in this case [see below for a digression]).

Another very nice feature is the ability to compute the occupied bandwidth of the modulated signal. When the option is selected, the SA will automatically integrate and draw a block around the portion of the input signal that occupied 99% of power.

occupied-bw.png

 

As a quick check, the value above is what is approximately expected, since the input signal was a 1kHz tone with +-2.5kHz frequency deviation. There is a rule (Carson’s rule) which provides an estimate that the occupied bandwidth will be (1kHz+2.5kHz)*2 = 7kHz. At this level of frequency deviation, it would be possible to space channels 12.5kHz apart, and have signals below about 3kHz, enough for speech as an example.

 

It’s always interesting to see signals in the frequency domain. Here is my attempt to get the carrier to virtually disappear, i.e. all energy is now in the sidebands (achieved by fine-tweaking the tone frequency). It is a characteristic of FM signals that causes this. Depending on the ratio between the frequency deviation and the tone frequency, each sideband power will increase or decrease, and at certain ratios the carrier can disappear. There is a table of sidebands vs. different ratios here.

1040hz-make-carrier-disappear.png

 

Summary

The N9322C goes beyond traditional spectrum analyzer products in offering modulation analysis capabilities. The execution here is good enough to perform some useful tests on FM transmitters and receivers to help align and troubleshoot them. As a next logical step, it would be nice to confirm the sensitivity of a real receiver based on SINAD, and confirm that it functions to its specification; this will be revisited in a later blog post.

 

Digression

Back to the temperature effects on the oscillator; touching a finger on the oscillator causes temperature dependency to be exhibited as shown in the photo below (no modulation, FM mode disabled for this screenshot for clarity):

finger-on-oscillator.jpg

 

Here is a short video showing what happens as I touch and let go of the crystal oscillator. It was very responsive!