Build Your own 2.0-2.7GHz Scalar Network Analyser

Created 8 December 2003

Here's another one of those projects I've been thinking about a lot but doing very little with. Until now. But, bear in mind that this is a work in progress.

Like I said, it's a 'work in progress'!

For a while now I've been dabbling on eBay for that elusive Network Analyser that'll cover the 2.4GHz band. I had to think to myself - do I really want to spend several thousand dollars on a piece of kit that'll be stuck in the cupboard for 99.9% of its life?

So I thought I'd try buildinh my own. Firstly, let's consider the basic components of a conventional Network Analyser with a T/R test set...

Basics of a conventional network analyser with T/R test set

The key components here are a frequency agile sweep oscillator, some directional couplers and some detectors. 'DUT' is the device under test. The three detectors provide the amplitude of the power applied, the amount of power returned and the amount of power passed through the device respectively.

Sometimes, in the case of an antenna or dummy load, the DUT will be a single ended device. In this case, only the forward and reflected powers of the first two detectors are used.

The sweep oscillator scans a range of frequencies, and the amplitudes of the three detectors taken. In its most basic form, the sweep oscillator frequency is shown on the display's x-axis and the amplitudes are combined to show return loss, power passed through and/or VSWR on the y-axis.

The detectors are low capacitance diodes specifically designed for this job. They operate in the square-law region, well below the forward knee voltage. In this region, allegedly the voltage across the diode will be in proportion to the power applied.

On some network analysers, the directional couplers are placed into a unit called a 'test set'. A T/R test set allows measurements to be made as above. An S parameter test set is more comprehensive and includes many more combinations of options suitable for testing a device both ways around without removing it from the test set and turning it around.

Practically, for the average home constructor, how do you approach building a network analyser? Practically, you really need to know what you want to achieve. For example, all I wanted was to be able to show was at around 2.4GHz for tuning purposes, and the functions I wanted to test were (a) return loss from an antenna and (b) filter passband tuning.

Cheap sweep

To provide the sweep oscillator, we need to be able to scan a range of frequencies, and initially I looked at using a VHF DDS synthesiser and multiplying it up. But, my RF design is lacking, so I looked at using the 13cm TV transmitters created by Giles Read G1MFG ( These provide about 30mW (15dBm) with a coverage between 2.0 and 2.7GHz. With no video applied, they generate a carrier.

Measuring the return loss of a one-port device

Next problem - how do you program these 13cm transmitters? Well it turns out they're based on the SP5055 PLL chip. This chip is programmed from the I2C bus, so a bit of PIC programming was called for. You program the chip with a fifteen bit PLL divider value of 16,000, and out pops 2.0GHz. Program it with 20,000, and you have 2.5GHz. The conversion factor is Freq = 0.125 * n where n is the value set in the PLL. It's also possible to read whether the PLL is locked or not. The smallest PLL step is 125kHz, fine for our application.

Measuring the response of a two-port device


The detector

I managed to pick up an 18GHz SMA (m) to SMA (f) detector at a hamfest for about $25. Because they're highly non-linear, and because they are likely to present a large mismatch, it's wise to place an attenuator (10dB+) before them and the port you're testing. That way, they'll still appear to present a match fairly close to 50 ohms. Take the output from the detector and pop it into a high gain op-amp. This then feeds the ADC input of the same PIC device programming the scan frequencies. I used a simple positive  LM358 with a variable gain set by a multi-turn pot. Gain was of the order of 60dB.

The directional coupler

Again, at a hamfest, I picked up a Narda 1.7-4.2GHz 20dB directional coupler for about $30 or so. These devices are quite neat. They sample a proportion of the power going through them into a third port which is what you measure. Depending upon which way around it is, it will measure either the incident or reflected power. The Narda unit I used samples the signal 20dB down from the power passing through. 20dB down is about the value you want in a directional coupler for this purpose.

In retrospect, a nice dual directional coupler would have been nice. That way you can simultaneously compare incident and reflected power.

The PIC Controller

The controller maintains three functions: (a) set the sweep frequency, (b) read the detector value and (c) take commands and report back to the PC. If the PC Board looks like an LVB Tracker board, then that's because it is! The PC communications is a standard 19,200bps RS-232 link.

Displaying the results

I hacked together some VB code to show what we can do with this...

Like I said, this is work in progress. When I have good enough results for you to use, I'll release the information. The biggest problem right now is calibration and result accuracy. The good news is that results are repeatable. The bad news is trying to figure out what they mean!

Mail Howard, G6LVB