Introduction
The Agilent/Keysight N9322C spectrum analyzer (SA) can be purchased with a built-in tracking generator option (option TG7), which means that it can generate signals as the analyzer sweeps measuring power across a frequency range. This means that you can provide stimulus to a circuit and view the output. It is great for checking out the frequency response of filters or amplifiers.
The SA also offers a reflection measurement option (option code RM7) which, together with the tracking generator, allows the possibility to find out how well signals can be transferred into a load. More on that further below.
This post examines frequency response measurements, how to build and test an amplifier or a filter, and discusses an amplifier test using the Texas Instruments CC11xL development kit Texas Instruments CC11xL development kit which allows for wireless connectivity at 868MHz.
Antenna reflection measurements are also discussed; these are useful to find out how much power is transferred from a transmitter to the antenna. This is of high relevance when good connectivity range is needed of course.
The topics covered in this post include:
Return Loss and VSWR (Voltage Standing Wave Ratio)
Spectrum Analyzer Calibration for Return Loss Measurements
Antenna Measurements (2.4-2.5GHz)
Building an Amplifier
I’d created a small circuit board for an amplifier, and it was time to select some parts and get soldering. The selected circuit was based around an MMIC amplifier. MMIC (monolithic microwave integrated circuits) are integrated circuits with RF features (such as transmission lines) built at the semiconductor level. They are the easiest way to obtain good performance with a flat frequency response across a very wide frequency range, and have 50 ohm impedance by default.
I used an Avago MGA-30889Avago MGA-30889 device; the circuit from the data sheet was used, although I didn’t use the exact part numbers recommended in the datasheet (and it had an effect – see later).
This is the circuit from the data sheet, it is a typical design for an MMIC amplifier:
It was not easy to solder, due to some 0402 sized components. Each part was soldered and inspected with a magnifying glass before moving on to the next one.
The gain was measured by connecting the tracking generator output to the amplifier input, and connecting the amplifier output to the spectrum analyzer input via a 15dB attenuator.
Since the MGA-30889 is supposed to provide approximately 15dB of gain, the 15dB attenuator would mean I should see around zero dB of gain. The SA will handle a higher input before damage occurs, but unless there is a good reason to remove the attenuator, it’s best to always be extra-safe and use it.
The screenshot below shows the output. The y-axis is set to 1dB per division. As can be seen, the gain was flat within +-0.5dB (indicated by the pink lines) for a frequency range of about 100MHz to beyond 2GHz. The SA can be programmed to alarm when limits are exceeded; this would be ideal for automated test scenarios.
The particular design that I chose from the datasheet should be capable of +-0.25dB of flatness from 200MHz to 2.6GHz, so my PCB layout and component selection (I used what I already had and did not purchase the exact parts specified in the datasheet) has had a negative effect, and the lesson is that a lot of care is needed to construct circuits to function in the multi-GHz range. I’d actually originally intended the board to be used for a 160MHz amplifier. I want the amplifier for 868MHz, so it exceeds the requirement, and it is good to know that the current board can be used to around 2GHz, and beyond this with some degradation.
As a final check, I connected the amplifier up to the Texas Instruments CC11xL development board’s CC115L transmitter, and set the transmit power to 5dBm. I was therefore expecting around 20dBm output. The screenshot below shows the original signal and the amplified signal. There is 20dB attenuation inserted, so the output signal was about 19dBm. A walkabout test also revealed the increased range.
The CC11xL range of devices from Texas Instruments (TI) can output at a decent power of 12dBm. With a suitable PCB or external antenna this allows them to comfortably cover an entire home and probably garden and quite a few homes away, depending on circumstances. With the amplifier attached, the range was noticeably extended; I could travel further from my home and still see data successfully received on the TI development kit LCD screen.
Although the amplifier was tested with the CC115L transmitter, in practice TI’s CC1190 range extenderCC1190 range extender integrated circuit would be more suitable for many applications since it contains the amplifiers and switching for transmit and receive applications using the CC110L transceiver.
Filter response measurement
Filters are used to remove unwanted signals, so that the frequency range of interest can be processed further. As an example, an amplifier would overload the input of a radio receiver without the ability to first filter off the unwanted parts of the spectrum.
There are lots of filter types and implementations. Designing filters can entail significant maths calculations, however a popular way is to either use tables of computed data to simplify the calculations, or to use software. A free trial version of ‘Elsie’ software is ideal to come up with a circuit for an inductor and capacitor (LC) filter, and is possibly the easiest software to use for this. It is possible to type in the requirements, and it will auto-generate a circuit along with the desired component values, and plot an expected response curve.
The tracking generator inside the SA is ideal for confirming filter behaviour too. The reason why you’d want to do this is that it can be extraordinarily hard to construct a filter and have a response as expected, without some adjustments.
The end result of using tables or software will be that the desired component values will be unobtainable values, so some experimenting is required to pick the closest values available (or use trimmer inductors or capacitors) and stray inductances and capacitances in the layout will have an impact too. For testing and aligning the filter, the tracking generator becomes highly important.
It was decided to test out the SA using an off-the-shelf filter initially. The selected one was a Coilcraft P7LP-157; it is a low-pass filter with cut-off (-3dB point) at 150MHz.
The filter has an extremely sharp cut-off, so it was important that the noise floor of the SA was sufficiently low. To achieve that, I heavily reduced the resolution bandwidth of the SA, at the expense of a slower display update. The result is shown in the screenshot below (frequency range shown is 100MHz to 1GHz, and the tracking generator output was set to -20dBm, with a display normalized to 0dB).
To summarise, free software and the tracking generator makes creating and testing custom filters very feasible. Probably it would be worth purchasing kits of inductors and capacitors if one is serious about designing filters. For inductors, the Coilcraft 0603HP range would be ideal for filters in the 100MHz to 1GHz range; kits C406A-2C406A-2 and C406B-2C406B-2 would be worth purchasing. For capacitors, Murata GQM series would be one option.
Return Loss and VSWR (Voltage Standing Wave Ratio)
Antenna return loss and VSWR measurement is a non-trivial task. I had some antennas that were designed for around the 868MHz frequency, and I was curious to perform these measurements for the antennas. In the past, I’ve not performed such a measurement (at least not with an SA).
Return loss is a measurement that reveals how close a load is to 50 ohms. The idea is to send a signal into a transmission line (e.g. a coax cable) to a device (which could be an antenna or a filter for example) and then measure the power that is reflected back. If the device is perfectly 50 ohms resistance, then no power is reflected back. If the original signal was (say) at a level of 0dBm, and the reflected signal was -50dBm, then the return loss is 0-(-50dBm), i.e. 50dB. Such a large value is desirable and it means that nearly all of the transmitted power into the coax was transferred to the remote device.
A return loss of 50dB is in fact an almost perfect match; in reality the return loss value could be (say) 10 or 15dB. Lower than that means that a significant percentage of the signal is being reflected back. Another representation is known as the voltage standing wave ratio (VSWR). A PDF document containing a table of conversions between return loss and VSWR is available (source: skyworksinc.com). The table is also helpful to show what this means for the amount of power transferred to the load, and the amount of power reflected, as a percentage. As an example, it can be seen that a return loss of 12.3dB means that 94.1% of the power in the transmitted signal can be transferred to the load, and 5.9% will be reflected back.
Note that it doesn’t mean that this much power was successfully radiated by the antenna. It just shows how effectively the power was transferred to the antenna.
Here is another way to see the effect of a mismatched load; the video below shows the voltage traveling across a transmission line (e.g. coax cable) where the coax cable is exactly the length of the wavelength of the signal. When the load is perfectly matched, there is no reflection, and no standing wave effect is created.
When the load is mismatched and is actually a short circuit, then the signal reaches the end of the transmission line and is reflected with a phase difference of 180 degrees. This results in a standing wave across the transmission line. Notice that the minimum amplitude is at the end of the transmission line.
When the load is replaced with an open circuit, the reflection has a phase of zero degrees. Again it results in a standing wave, but the maximum amplitude is at the end of the transmission line.
The demo above is from this website (needs Java).
Return loss measurements require a component such as a ‘coupler’ or ‘bridge’ device to allow the reflected signal to be measured. One popular device is a ‘directional coupler’ and the best way to examine its behaviour is to look at the graphic representation for it.
The coupler is designed to have low loss through its ‘main line’ which goes from the signal generator, through the coupler, to the antenna in the diagram above. Any reflected signal follows the same path from right to left in the diagram above, with a small percentage of it being passed up into the spectrum analyzer (in practice the signal source is set to a low level to prevent damage from reflected power).
With a system as shown in the diagram above, the it should be possible to obtain a line (frequency on the x-axis, return loss on the y-axis) on the spectrum analyzer display with (perhaps) an undefined return loss everywhere, except near the frequency that the antenna is designed for. Near this frequency, the return loss should be a high value, indicating that when used as specified, the antenna will not reflect back a significant amount of the signal.
The photo above shows a typical directional coupler. Also shown are some other parts that are often used with it. These are discussed next.
Note that with the RM7 option with the spectrum analyzer, there is no need for an external coupler type device. There is already a device built-in to the spectrum analyzer. This is really nice because it saves the need to purchase and maintain a high quality coupler and associated cables; it can be electronically enabled from the spectrum analyzer using the ‘Reflection Measurement’ option built-in. However, just as with an external coupler, the internal one needs calibration too.
Calibration
It is hard to get consistent, repeatable measurements without calibrating the system including the cables and connectors that will be used to perform return loss measurements. Furthermore conventional flexible coaxial cable is almost unusable. Instead, semi-rigid cable can be used because it will remain in a configuration long enough to perform the calibration and to then insert the device under test (the antenna in this case). Semi-rigid cable is (as the name suggests) more rigid than normal coax cable. It has a metal shield which can be bent by hand and remains there. The photo above shows RG-402/U semi-rigid cable soldered to an SMA connector.
The calibration is achieved by using some known quantities; typically an open circuit, a short circuit and a 50 ohm load.
The N9322C spectrum analyzer does come with a very nice calibration standard device (shown on the right side of the photo above; it is the T-shaped item) if purchased with the reflection measurement option. Sadly for work with smaller antennas, additional standards will be needed to suit the connectors in use. They can cost thousands of dollars, so instead a DIY open and short was used, along with a commercial 50 ohm SMA sized load. This is really just for the purposes of experimentation; the results will be inaccurate and for any real scenario the correct calibration standards must be purchased. Cal standards are not easy to make. The short and open circuit standards need extremely precise machining (and are kept as a pair) so that the reflection can be characterised properly. I used a junior hacksaw; not ideal.
To produce the DIY standard, an SMA connector was hack-sawed in half, as shown here, and one half was then discarded. The remainder had the center pin and outer shell flush.
Next, to produce the short circuit, a small disk was cut from a copper sheet or from a roll of copper tape (I used scissors but a paper hole punch could be used perhaps!). Then, a tiny amount of solder paste was added to the center pin and to the outer flush surface. With the copper disk in place, a soldering iron was applied on top until the solder paste melted and made the disk stick.
The actual calibration process is automated using the spectrum analyzer; the ‘Calibrate’ button is pressed, and the display prompts the user to apply the open, short or 50 ohm load to the connector on the spectrum analyzer.
In my case, since I wanted to calibrate an entire system of the spectrum analyzer plus the cable that would connect to the antenna, I had to apply the open, short and 50 ohm load to the end of a cable. I used a semi-rigid coax cable, with one end connected to the spectrum analyzer, and the other end because my reference point for the calibration. At that point, I connected the SMA-sized open and short connectors when prompted, and then the 50 ohm SMA termination when prompted during the calibration process.
Antenna Measurements
With the calibration complete, I removed the 50 ohm load, and replaced it with an antenna. Here are some antennas that I tried. These two are intended for 868MHz operation, and are manufactured by Anaren, model 66089-080666089-0806 and 66089-083066089-0830.
The photo below shows some more antennas that were tested, manufactured by Pulse Electronics and Antenova.
The screenshot below shows the return loss for one of the Anaren antennas (the longer one of the two Anaren antennas in the earlier photo). The x-axis is frequency (center is 868MHz, span is 500MHz) and the y-axis is return loss, with a high return loss (i.e. a good match) being a valley on the graph:
As can be seen from the text in the screenshot above, the valley occurred at 845MHz. This is slightly lower than the expected 868MHz. The test was repeated with the second Anaren antenna, and the result is shown below:
In this case, the valley was at 899.78MHz which is slightly higher than 868MHz. The return loss at 868MHz (center of the graph) was around 7dB for both of them, which according to the table corresponds to a VSWR of about 2.5:1. This would indicate that about 80% of the power is available for the antenna, and 20% is reflected back. Of course, there is no need to refer to a table, the SA can generate a view with VSWR as shown below; the center of the graph is at VSWR 2.5:1 as mentioned:
I then attempted the Pulse Electronics W5017 model. The results are shown in the screenshot below. As can be seen, the return loss was a much better value. The valley was at 897.6MHz, which is almost exactly the center frequency of the antenna (the W5017 is intended for 868-928MHz).
In terms of VSWR, the result looks like this:
The Anaren and the Pulse Electronics antennas exhibit similar return loss at 868MHz, although the Pulse Electronics one has far better return loss for its center band region (VSWR value is an excellent 1.3:1 at the center, as can be seen by the valley readout in the screenshot above). This corresponds to more than 98% of the signal being transmitted into the load, i.e. less than 2% being reflected back. But at 868MHz the Anaren and the Pulse Electronics antennas did not seem optimal, and would reflect about 20% of power at this frequency.
The question has to be asked, how accurate are these results? The answer is that it could only serve as a guideline. Accurate calibration standards would help, as would to mount the antenna on the board where it will be used, so that the antenna has the same ground plane as where it will be deployed. The semi-rigid coax leading into the SA would need ferrite toroids around it, so that the coax does not look like a ground plane extension that would not exist if the antenna was being used for real. A decent quantity of toroids is needed to attenuate the RF by a large amount; ferrite comes in a few different formulations, and type ‘61’ would probably be ideal, such as Fair-Rite model 2661540202Fair-Rite model 2661540202 since it is intended for attenuation up to 1GHz. If you’re testing 868MHz antennas then you’ll want to purchase a few dozen of these. There are some practical tips for measurements at this web page.
2.4-2.5GHz Antenna Measurements
After calibrating at 2.2-2.7GHz, I decided to test some Antenova antennas intended for 2.4-2.5GHz operation. I had three identical ones, so all three were tested. They are designed for 2.4-2.5GHz operation.
The screenshot below shows the return loss for the first antenna. The red vertical lines indicate 2.4 and 2.5GHz. It can be seen that the return loss is superb, although the valley of the curve is shifted left of the desired frequency band for the antenna. This is expected; there were no ferrite toroids on the semi-rigid coax.
This is what the VSWR looks like for the first antenna; it ranged from 1.36:1 to 1.6:1 for the 2.4 to 2.5GHz range, which is very good (98.3 to 94.7% power transferred):
The second antenna tested was just as good, and exhibited near-similar return loss/VSWR charts; the VSWR was 1.3 to 1.65 for the same frequency range. The Antenova data sheet has an expected return loss chart, so it was overlaid for a comparison with the results I had obtained for the two antennas that were tested:
This was great, it validated the test procedure, and ferrite toroids should improve the measurement further.
The third antenna tested had very different performance across the same range; it seems repeatable, and I can’t explain it; the return loss graph is shown below:
I believe these rigid antennas should exhibit more consistent behaviour, although a greater sample would help. The conclusion would seem to be that for testing out antennas it would be a good idea to obtain lots of samples.
Summary
This post grew a bit longer than expected, but this is because the tracking generator functionality is highly useful. It provides a lot of confidence with antenna and filter circuits. The spectrum analyzer can be used to support amplifier, filter and antenna design and/or test.
Quite a bit of practise and analysis is required to make use of the return loss capability for antenna measurement, and having good calibration standards and keeping ferrites and semi-rigid coax around both with connectors and bare for soldering onto the PCB containing the antenna will help.
Top Comments