|Product Performed to Expectations:||10|
|Specifications were sufficient to design with:||10|
|Demo Software was of good quality:||10|
|Product was easy to use:||10|
|Support materials were available:||10|
|The price to performance ratio was good:||10|
|TotalScore:||60 / 60|
The R&S FPC1500 is a fairly unique instrument. Calling it just a spectrum analyzer is an understatement!
Just as oscilloscopes in recent times have managed to integrate a lot of functions such as logic decoding and arbitrary waveform generation, some spectrum analyzers too have integrated functions like modulation decoding and RF signal generation. The FPC takes quite a few more steps along the road though, as will be described further below. Together with a modern oscilloscope, the two devices would greatly enrich the desk of most engineers. What is interesting is that even if ‘radio’ is not in an engineer’s job-role, there are a surprising amount of tasks that the FPC1500 can accomplish. The FPC1500 is relevant for working with any modern circuitry including diverse topics like power supplies, wireless charging, electronic circuit tracing and troubleshooting, component characterization, wiring and network installations, and myriad other things you can think of – it’s all RF! It reminds me of a comedy series where the protagonist was always convinced everything was all Indian.
So, even if you’re not interested in traditional radio, I hope you read on, because it is surprising how much is relevant to other disciplines. Also, there is a video as an accompaniment to the review - I tried to keep it concise but there is a lot to go through:
I discovered that the FPC1500 offers a really nice combination of functionality that makes everyone RF-ready for all sorts of RF and non-RF related activities. The spectrum analyzer can be used to sniff out activity and problem in digital circuits. The in-built vector network analyzer allows all the components and component assemblies that a multi-meter or oscilloscope would be inappropriate for, to be measured; for instance a capacitor is not just a capacitor when it is inserted in a circuit such as an amplifier or even when used for supply decoupling. The FPC1500 will allow you to see what such components really look like. For diagnostics and troubleshooting, cable installers can benefit from being able to see the electrical properties of their cabling, and the FPC1500 makes that possible. All of these examples are activities that are not normally associated with RF.
Unlike many other test tools, the supplied software has to be good with a spectrum analyzer. I found that the FPC1500’s supplied InstrumentView software is not just great – it is truly fantastic. It was a pleasure to use, and the productivity gain was huge.
In the RF domain, one sometimes needs to work for hours or days on end with test equipment. The FPC1500 is totally silent and fanless – this is unheard of when compared to what was available a few years ago.
The FPC1500 has extremely impressive performance when compared to other low-cost spectrum analyzers. The FPC1500 can cost more, so it is not a fair comparison, but putting cost aside (there is always the possibility to negotiate lower costs when purchasing equipment) and focussing just on the technical capabilities, the FPC1500 offers excellent performance and very rich features that I would not have expected from such a test tool. It even feels awkward calling it a spectrum analyzer because it does a lot more. I will try to refer to it as the FPC1500 device : ) This is particularly because the FPC1500 contains functionality ordinarily seen in different test instruments – in particular the vector network analyzer (VNA) capability.
Everything about the FPC1500 isn’t perfect, but for such a recently released device I’m impressed that everything worked. There was the minor bug here and there, but nothing I could not work around. Also I think the features and performance mix is really good. Everyone wants more of course. The mix of design decisions and performance decisions that resulted in the FPC1500 seem sensible, and a lot can be done with this device – it is like a Swiss army knife. In this review which cannot possibly cover everything that the FPC1500 is capable of, I have tried to cover many of the broad interesting areas of functionality, and also different examples (RF and non-RF related) of how it can be used. It is not written in an overly technical manner since readers may have different backgrounds. It is written in sections so some can be skipped if you’re interested in just particular areas. This review begins with a look at the specification and then some traditional RF topics but then moves on to exploring other capabilities.
Probably like many engineers I build my first radio as a kid. Straight out of university I had a very traditional exposure to RF topics working with experts in designing state-of-the-art radio communications equipment used by military and police organizations. These high performance designs allow soldiers to communicate around the world even in the absence of mobile networks and satellites, operating off a few rechargeable batteries.
As expected, there were a fair amount of spectrum analyzers at our disposal, but just a few network analyzers because they were really exotic and pricey. We also had a few Rohde and Schwarz radio receivers in the lab and they were pretty much the only third party receivers that we used for verification. They were large 19-inch rack receivers and they are also likely used around the world by governments and for maritime use.
Although I had a traditional radio upbringing, the cool thing with radio frequency (RF) topics is that nearly everyone works with RF, even if they are not aware of it.
As an engineer even if you’re working on something that does not intentionally contain a radio system, RF impacts you. In digital electronics territory, anyone creating something to attach to (say) a Raspberry Pi will at some stage encounter the effects of high speed signals and what happens when these signals are not terminated correctly.
To take another example - any product that is sold needs to be compliant to electromagnetic compatibility (EMC) standards, so testing for the emissions from digital electronics and switching power supplies is something many engineers are aware of.
Even something as seemingly benign as installing a cable is not free from RF concerns. Cable faults can cause connectivity and data bit error problems in offices. I once spent days trying to figure out why a connection was flaky – it turned out that the length of cable was too long! This is very easy to occur when working in multi-story offices on a campus type of site – the underlying causes are RF related.
And as for intentional radio transmitters and receivers - I tried counting the amount of radio systems in the house (such as mobile phones, anything with Bluetooth, any computer with WiFi, home automation and so on, but setting aside the Raspberry Pi’s and other embedded boards because then the figure just gets crazy) and I had to stop at about 40. Each of those devices may have more than one radio, so a quantity of the order of 100-200 radios sub-systems inside a single home is not unrealistic.
Spectrum analyzers, like radio receivers, have to be able to pick up the tiniest of signals in the face of very large signals that may also be present. The ability to resolve a signal very close to another signal is important. Plus, spectrum analyzers internally generate their own signals, which could interfere. These are huge challenges, and was something very difficult to achieve without throwing a lot of hardware at the problem – the resulting cost explains why spectrum analysers were once rare in labs. Modern spectrum analyzers use digital techniques to integrate and reduce the hardware. Some of the spectrum analyser functions occur in the analog domain, and the remainder are achieved with digital electronics and software. Given Rohde and Schwarz’ legendary history, particularly in radio and spectral analysis, it is clear they can achieve superb results, but I was curious what level of performance was achievable in this low-cost general product. And by low-cost general product, I mean the general-purpose tool that one will use before busting out a high-end multi-$10k spectrum analyzer or other high-end test tool that ordinarily will not be on your bench unless you’re making use of it immensely. If the test bench has just the main useful tools, the FPC1500 would be a serious contender for desk space, for any engineer.
Although the FPC1500 suits RF and non-RF engineers, it is worth briefly addressing the concept that RF is black magic : ) This definitely was true up until relatively recently! For the reason that when test equipment costs tens of thousands of dollars, many engineers were just not exposed to working with such gear and it was a specialist domain. Even Lego is black magic if all you’ve seen is the end result and never touched it.
Another reason is that RF circuits had a lot of adjustments. A slight error could result in the difference between capturing a signal, or having a totally radio-deaf product. It was very frustrating without the right test tools. Here’s an example – this is a photo of a radio circuit from around the 1980’s and you can see all the expensive adjustable parts – no fewer than eight adjustments in a two inch square board.
Some of the types of adjustable components that were used, are no longer necessary in many modern designs. Today, it would be surprising to see even a single adjustable element inside a lot of products containing a radio sub-system because everything has been collapsed into microchips. Radio architectures have been developed that at one time were considered problematic or of low performance, but nowadays are easy to implement almost entirely in silicon, optimized for good performance, and therefore there is far less need for large coils or adjustments. This means that the black magic aspect has disappeared to the extent that barriers are reduced for engineers to successfully incorporate radio technology in their products. Anyone who can read a silicon chip datasheet can get some radio communication going. It might not be high quality (by high quality I mean the things that make a fantastic product, such as long range, low noise and so on) but it will function. Taking things to the next level requires test equipment.
Even when off-the-shelf modules are used, it is highly likely that far better performance can be achieved with custom modifications. Many companies spend money on consultants to perfect their antenna portion of the design for instance so these are skills worth learning.
One other point worth mentioning is that RF test equipment has often been really complicated to use to perform measurements. The procedures to make certain measurements are not easy, often the procedures require a lot of equipment, and slight errors or oversights in the procedures are easy to make but have big consequences. Modern test equipment helps because procedures can be automated, or the equipment can prompt the user to perform certain actions to complete the measurement. Possibly more than some other forms of test equipment, there is a lot of experience and knowledge wrapped up inside the software that runs in spectrum analyzers, and that is something of value and it makes the difference knowing that the people who designed the test equipment put their knowledge into the software to help you make better measurements.
When I first heard of the FPC1500, I must admit I was super-excited to learn that it had so many features! This isn’t like a bathroom buddy – instead this is a genuinely useful set of features that I knew I would love to make use of every day.
Most engineers have already experienced some spectral analysis capability, embedded within the oscilloscope’s computing capability (software or ASIC or FPGA) - all modern scopes have this mathematical capability for transforming the time domain view into a frequency domain view, also known as FFT capability. However, there is a huge difference between that and a spectrum analyser. The FFT capability doesn’t have much dynamic range. The oscilloscope is really a voltage sensing type of product, and has (usually) an 8-12bit analog-to-digital converter. The range that this can resolve is just in the tens of decibels (dB). For example, a 10-bit ADC can quantize a signal into a digital integer value as small as zero, up to a value as large as 1023 – a range of about 60dB. With RF signals, the amount of power for signals of interest spans a massive range – sometimes beyond 150dB! To do this using the same architecture would require a 25-bit high-speed oscilloscope, which doesn’t exist. The only oscilloscopes that I’m aware of that have a genuinely effective spectrum analyser capability in physical hardware are the Tektronix MDO 3000 and 4000 series but they are very unique and have a specific internal architecture for this.
The unit arrives with worldwide mains cables, a 2m length USB cable and paper copies of some brief documentation and licenses.
The unit looks smart, there is an integrated grip area to carry the unit, and it is fairly lightweight (3kg). One thing that sounds small, but is quite major, is that the FPC1500 has no fan, the unit is silent. This is a huge deal because it is not unusual for a spectrum analyser to be powered up for many hours at a stretch, and working next to equipment fans is really unpleasant. When tweaking RF circuits, it can frankly take hours and days. I have a very old HP-era spectrum analyser that sounds like a rocket powering up, and I find it really unusable. Even a four-year old basic spectrum analyser I use still has a fan – less noise than the HP-era product, but still unbearable after a couple of hours working a metre or two away from it. I had resigned myself to the fact that maybe a USB based spectrum analyser could be the only solution, but now the FPC1500 has shown that a dedicated standalone instrument can be silent – and in fact actually has no ventilation slots either. R&S were clearly very confident in the thermal design of their product.
I hope the drive toward silent test products continues. Right now it is very rare to want to set the keypad volume to 1% - that’s what I set the FPC1500 to, and the beep is still a bit loud : ) I may eventually set it to zero.
The display is large and a spectrum analyser can really benefit from this. The rubberised buttons are laid out conventionally, with little difference to other spectrum analysers. Peripherals such as a memory stick and a mouse can be attached using the front-panel USB sockets. The memory stick is a convenient way to transfer data. Firmware upgrades can be done using the memory stick too.
Additional connectivity includes an Ethernet and USB port on the rear, as well as 2.4GHz 802.11 (Wi-Fi).
The menu system is straightforward to use. All the usual information is well laid out on the display. Some things are positioned slightly differently to what I’m used to, but it takes just a few minutes to get familiar with it. The current settings appear above the main sweep display area, and the right side is dedicated to the soft-keys. The display is uncluttered compared to smaller display models which have always overlaid information onto the main sweep area display. Instead, almost every important setting is shown above the sweep area, leaving just the frequency settings displayed at the bottom where they should be (i.e. near the x-axis of the sweep area display chart).
The display does lose contrast when the instrument is viewed from above, but it is still usable.
As with most spectrum analysers, the top group of buttons select a function area, and then the soft keys can be used to narrow down which precise function is needed. The most likely function is already highlighted and ready to be used each time. So, if the ‘Freq’ button is pressed, then although the soft button selections allow for start and stop frequencies, the center frequency entry will already be selected by default so the user can immediately use the numeric keypad to start entering it if desired with no need to press any soft key. What was unique and that I really liked was that the units (kHz, MHz, etc) were real buttons next to the numeric keypad, instead of soft-buttons. I can quickly alter the top important settings in all the most important function areas (such as Frequency, Bandwidth, Span and Amplitude), without taking my eyes off the keypad area. No need to consult the soft-keys and the display while basically reconfiguring all the major settings!
The supplied software, R&S InstrumentView, has a lot of features. The software scales well, so it is possible to run other apps without needing to dedicate the entire screen to the one instrument. The software is extremely powerful and a delight to use. There is more on this throughout the review as the features are explored.
Different views can be brought up from the menu on the left. One basic mode is a remote view which functions adequately. Of course the view is dynamic. Incidentally the spectrum trace shown here was from a wireless Sennheiser microphone in a fairly quiet room. The two peaks either side of the center frequency are an out-of-band pilot tone (about 33kHz) as can be seen from the markers.
The FPC1500 is speedy. When configuring up the span and resolution bandwidth It will automatically select a sweep time as expected, which is usually very good, but it is possible to override it and do tricks like perform a 1GHz span spectrum sweep with 10kHz resolution bandwidth, and update the screen, in under 200 milliseconds flat. Clearly the underlying technology must be sampling a very wide bandwidth and then rapidly doing fourier transforms while the front end frequency-hops across the spectrum in chunks, but I’m still really surprised at the flexibility it has in sweep speed in what is supposed to be an entry-level spectrum analyser.
The difference is massive compared to previous generation basic spectrum analyzers. To take another example, and sticking with the automatic sweep time settings; a span of 20MHz and a resolution bandwidth of 100Hz results in a sweep time of 65 seconds. In contrast the six years old (but fairly decent) Rigol DSA815, will have a sweep time of around 25 minutes.
Taking a more contemporary Siglent SSA3032X, that performs a 3.2GHz sweep at 1MHz resolution bandwidth in just under a third of a second. The FPC1500 configured for a 3GHz sweep at the same resolution bandwidth will choose a sweep rate of 38 milliseconds. The sweep time is almost an order of magnitude faster. The software is doing lots of work however – although the sweep time is incredibly low, with this example the refresh rate on the display was about 2 updates per second. So, although the sweep was rapid, the digital signal processing is taking some time too - but this is still very quick. Different spans and resolution bandwidths produce different results of course – the FPC1500 automatically selects a speed for the internal hardware front end to sweep, and the correct sampling for the digital processing to work its magic, but the speed can often be manually adjusted and the FPC1500 will internally allocate the correct time between sweep and processing.
As one more example, if the FPC1500 is set to (say) 330MHz center frequency with a very narrow span of 8kHz and a resolution bandwidth of 10Hz, then the automatically chosen sweep speed is 819msec which is pretty good, but the Keysight Basic Spectrum Analyzer series N9322C will choose a sweep speed of 529msec. However, the N9322C cannot reduce its resolution bandwidth any further. The FPC1500 at 3Hz resolution bandwidth and all other settings the same will achieve a 2.2 second sweep speed by default, which is still very usable. At 1Hz however the sweep speed slows down a lot, to 24 seconds. Therefore it can be seen that the architecture is optimised slightly differently for the FPC1500 compared to other spectrum analyzers because the FPC1500 has more granular resolution bandwidth. Sometimes the FPC1500 will win, sometimes the N9322C will win. If the span is increased from 8kHz to 1MHz, at 10Hz resolution bandwidth the N9322C will take 40.7seconds for its sweep. The FPC1500 will take just 32.5 seconds, a good 20% improvement.
In summary, the FPC1500 is no slouch with its sweep times although it will depend on the precise frequency span and resolution bandwidth settings, and if the auto-selected sweep time is too slow, often it can be speeded up a lot through manual selection.
Incidentally, the user interface is always quite responsive regardless of the sweep time or the level of digital signal processing. I did not experience any lock-ups or lag.
The noise floor, or Displayed Average Noise Level, is a measure of the smallest signals that can be detected by the Spectrum Analyzer. It is determined by the quality of the circuitry inside the Spectrum Analyzer. If the circuitry is too noisy, then the DANL may be too high to identify very low level signals. Such signals could be amplified, but then there is a risk that a nearby higher-amplitude signal that will also be amplified, could exceed the upper threshold of the spectrum analyser (in practice there are several ‘upper thresholds’ and the relevant one depends on precisely what is being measured, and that will be explored in the next section concerning dynamic range. In this section we are concerned about the lowest theoretical power levels that can be measured with the FPC1500).
It can get tricky fully understanding the capability of the spectrum analyser because the DANL can vary depending on the frequency range of interest. It would be nice to have a graph with frequency on the x-axis, and the corresponding DANL on the vertical axis. I decided to plot it out:
The blue bands indicate the region that the spectrum analyser supports – with and without a preamplifier. The typical level of DANL is indicated by the lower edge of the bands, and the specified limit is indicated by the upper edge of the bands.
The diagram below shows how the DANL compares with other vendor spectrum analyser specifications. The key to viewing this diagram is to pick a color, and then look for the two horizontal band paths for that color. They will represent the DANL values with and without the preamplifier enabled. Long story short, the FPC1500 has the lowest DANL versus the competition, across nearly the entire spectrum that it supports. It is exceptionally low noise. (By the way – regarding the spectrum analysis performance, even though I’ll refer to the FPC1500 everywhere in the text, the information applies to the FPC1000 too. I’ll just refer to the FPC1500 in the text for convenience).
Across nearly the entire spectrum, the FPC1500 has a better (i.e. lower) DANL. Only below 10MHz do a couple of others have an advantage. From 10MHz upwards, the FPC1500 has better DANL than the nearest-priced competition. The Keysight and Rigol models perform pretty well in this comparison too but technically are still no match for the FPC1500 above 10MHz. Both of the Siglent models struggle on DANL at the useful 2GHz+ area of the spectrum. Another key point is that although the comparison is normalized at 1Hz, the FPC1500 actually does have a 1Hz resolution bandwidth and the values in the diagram are actually attainable for real measurements. Some other spectrum analysers may have 1Hz resolution bandwidth but are compromised with reduced DANL for that bandwidth setting.
Also, just for interest, the pink line is for a very old but mid-range spectrum analyzer from the early 1990’s, the HP 8561E. These are listed on ebay occasionally, at crazy prices. It goes to show that even some modern cost-effective spectrum analysers can perform better than a mid-range spectrum analyser from a few decades ago.
However, DANL alone is just one aspect (but an important one).
Another area where the FPC1500 excels is in measurement display range; the span between the maximum allowed input level, and the noise floor, with all other parameters fixed. Note that this isn’t the same as dynamic range, because of distortion that can occur toward the upper limit of the measurement display range (the difference between the measurement display range, and the dynamic range, can be quite large, so it cannot be just assumed that dropping the input level a few tens of dB below the maximum allowable input will prevent distortion; anyway, more on that further below).
The FPC1500 has a clear 10-15dB of extra display range compared to other low cost spectrum analyzers. The diagram below shows in light blue that this range extends very low (as shown in the earlier DANL discussion, down to -150dBm without the preamplifier, across the majority of its spectrum range) and extends up to +30dBm whereas many other spectrum analyzers (overlaid in colors in the diagram) only reach +20dBm. In other words, across the majority of the spectrum, the FPC1500 has a very impressive 180dB of measurement display range.
The diagram belows shows this clearer, for an important part of the spectrum that will cover sub-1GHz frequencies (used for home automation, smart meters, and low-power wide-area networks LPWAN), and 802.11 wireless and Bluetooth. The FPC1500 has a significant 15dB or higher difference.
Having this display range is great, but for some measurements the distortion from the spectrum analyser needs to be considered.
This is important, because when you’re observing signals with a large amplitude, you are still often highly interested in the low-level content around that large signal. Without adequate dynamic range, one is forced to adopt workarounds such as notch filters to temporarily remove the large amplitude but keep the skirt. That is actually a very effective way of doing things, but such filters are expensive and difficult to make.
Across a chunk of the display range the response from the analyser will be linear, but at high power levels some distortion can occur. In the frequency domain, this looks like additional spikes on the spectrum display that shouldn’t be there. If there are spikes in the spectrum very close to the signal(s) of interest, then they mess with your measurements. One easy way to create such an unwanted situation is to connect the spectrum analyser to two high input signals that are very close in frequency to each other. Some spikes will appear on the display that are related to the input frequencies, but are not harmonics (it is known as intermodulation distortion). The solution is to reduce the input signals, or to increase the attenuation setting on the spectrum analyser, until such spikes disappear into the noise floor.
There are different types of distortion, and a couple of types are explored here to see how well the FPC1500 performs. The one that exhibits itself as spikes close to the signals as discussed above can be represented by a cubed portion of a non-linear function – the parameter that defines it is called the third order intercept (TOI). The parameter can be used to determine what is the dynamic range of allowable input such that those unwanted spikes do not appear on the spectrum analyser display.
The chart further below titled ‘Dynamic Range Calculator’ was drawn based on -150dBm DANL typical value and 7dBm TOI from the FPC1500 datasheet, so that it can be referred to whenever one wishes to apply signals and eliminate such distortion from the spectrum analyser display. This is highly important, because you want to be sure that any distortion effects you’re seeing on the spectrum analyser display are those that are present on the input signal (which could be the output from an RF amplifier for example), and not internally generated by the spectrum analyser. Then the engineer can work to eliminate that distortion from the equipment under test (e.g. the RF amplifier). Such distortion is very undesirable because it is hard to filter out after it has been generated, because it is within the bandwidth of the wanted signal; it isn’t a far-away harmonic. Here is an example showing why third order intermods are so bad; this spectrum shows an amplitude modulation (AM) transmission centered at close to 330MHz (330.000174MHz), consisting of two tones at 1200Hz and 1600Hz. It could be an audio transmission, or modem signals. If you look to the right of the centre carrier, the large twin peaks are the two tones. The separation is just 400Hz since 1600-1200=400. Now if you look on both sides of the twin peaks, there are a couple more peaks of lower power, marked 3 and 4. These are just 400Hz away from the 1200Hz and 1600Hz signals respectively. These would be very difficult to filter out of a received signal since they are within the bandwidth of the usual audio channels (e.g. 2.5kHz bandwidth etc). Although these intermods are visible on the spectrum analyser display, the question is, are they really there in the radio signal (in which case the engineer has a problem to solve!), or are they being caused by the spectrum analyser dynamic range – this is the reason it is important to be aware of what dynamic range is available. (Incidentally, this screenshot shows intermods coming from a transmitter set to a high output, causing the transmitter output stage to generate the intermods – so these intermods are really there in the connected signal, they are not internally generated by the spectrum analyzer).
The chart below has a set of parallel blue lines that indicate the noise floor (DANL) of the spectrum analyser, for different resolution bandwidth settings and different attenuation settings. The dashed blue line represents the cubed portion of the distortion function.
To use the chart, select one of the blue lines that reflects the resolution bandwidth that the spectrum analyser is configured to. At the point where it intercepts with the dashed blue line, you can read off the available dynamic range on the right side axis markings, and you can read off the ‘Power at (first) Mixer’ value at the bottom. The spectrum analyser internally has an adjustable attenuator followed by circuitry known as the first mixer. The power at mixer value is the maximum value that the mixer should see to avoid the intermodulation spikes from appearing on the spectrum analyser display. Usually this is just the input signal level minus the spectrum analyser attenuation level setting.
As an example, at a resolution bandwidth of 100Hz, the maximum power at the mixer should be -38dBm. If the input attenuator is set to (say) 10dB, then the connected signal to the spectrum analyser should not exceed -28dBm, to keep the power at the mixer at -38dBm. The dynamic range possible for such a scenario is 91dB (read off from the right of the chart), i.e. from -28dBm input to -119dBm.
The table here can also be used:
The pink and green horizontal lines in the diagram above represent another type of distortion, that has an effect when trying to measure small signals that are in the ‘skirt’ of larger signals. The spectrum analyser internally contains an oscillator (actually a frequency synthesized oscillator), and any jitter from that oscillator results in something known as phase noise, often described as a skirt around signals in the frequency domain. If a small signal is inside the skirt, then it won’t be visible on the display. The measure of the purity of the oscillator is a phase noise value that is indicated at an offset from the large frequency. The actual shape of the spectrum analyzer’s internal oscillator phase noise depends on implementation; it is based on a crystal oscillator and a phase locked loop (PLL) system. By careful design, the Rohde & Schwarz engineers will aim to minimize the noise level within the PLL bandwidth, but outside the bandwidth the VCO noise will dominate and so a specific bandwidth is chosen to get the best out of the synthesizer design. In the case of the FPC1500, it has a good quality VCO because the FPC1500 datasheet lists a typical value of -103dBc/Hz at an offset of 100kHz (and specified at less than -98dBc/Hz). This is excellent to see – it is far better than other low-cost spectrum analyzers at this very usable offset, especially when the resolution bandwidth can be set down to 1Hz.
That’s at 100kHz offset. I was curious about the performance at even at lower offsets, and I was pleasantly surprised. To see how, the pink and green dashed lines in the diagram above must be examined.
The pink lines are to be used for 100kHz offset, and the green line is for 30kHz offset. They can be used to see what size a small signal at such an offset would have to be, in order to become visible above the skirt of the larger signal. They describe the minimum power level relative to the larger signal, at an offset of 100kHz. As an example, with 100Hz resolution bandwidth, the smaller signal must be at least -83dBc for it to be visible. So, for that -28dBm example earlier, then at (or within) 100kHz offset, no additional signal will be visible unless it has an amplitude of -111dBm or higher. If such signals are of interest, then this has further reduced the dynamic range by another 8dB (from the lower limit of -119dBm, to -111dBm).
At an offset extremely close to the oscillator frequency, the phase noise of the internal frequency reference will dominate, and the specification states a value of -92dBc/Hz at 30kHz offset. It is also possible to feed in an external 10MHz reference. The devil is in the detail, and although -92dBc/Hz may sound poor, it must be interpreted in conjunction with the resolution bandwidth. Fortunately, the FPC1500 can go down to 1Hz resolution bandwidth, and at that setting the green dashed line in the diagram above can be used. This means that at 30kHz offset, the minimum signal power level must be greater than -92dBc. This is really good for a basic spectrum analyzer.
At 10kHz offset we can expect the phase noise value to be higher. It isn’t specified, however by applying a pure signal we can peek at the phase noise of the internal oscillator. For the input signal I use an Analog Devices AD9954, which should easily reveal the shape of the spectrum analyzer phase noise. I applied a signal at an arbitrary 47MHz at about -18dBm level, and set the sweep to the slowest 1Hz setting, with a span of 100kHz. As expected, we can see the shape of the PLL loop filter inside the FPC1500. It reveals that for phase noise measurements at around 10kHz offset (at the marker M1 position in the screenshot), there is about just under 10dBm increase compared to the 30kHz offset specification (see the second marker M2), which is not bad. Incidentally the AD9954 is a direct digital synthesis (DDS) device, so has very low phase noise, however you can see that the peak was not displayed here because I used a ‘sample detector’ mode, which is ideal for seeing the shape of the phase noise, but isn’t good for seeing the input signal with was the 47MHz signal at -18dBm.
With other low cost spectrum analyzers, they may specify a slightly lower phase noise at these very close offsets, but unless they have a 1Hz resolution bandwidth, the minimum power level required at that offset for a signal to be visible is actually higher than that needed for the FPC1500. The FPC1500 really does have excellent performance in its price range.
Although this section of the review examined the FPC1500’s own internal oscillator phase noise, it is also possible to use the FPC1500 to measure the phase noise from external signals. It makes use of a three excellent features of the FPC1500 that will be discussed further below – see the Measuring Phase Noise of a Signal section if you want to skip to it now.
I love automated measurements in spectrum analyzers because they reduce errors massively. It is easy to make a mistake with manual measurements with any spectrum analyzer unless you use it all the time, so I do rely a lot on the automation. Also, lots of time is saved. To take the example of third order intercept measurements, an earlier screenshot showed a radio transmitter generating intermods from its output stage, and markers were manually placed on the two tones, and the third order distortion. The FPC1500 can handle it automatically. I applied the same transmitter signal, and then centered the display so that the four peaks were visible, and then selected the Third Order Distortion measurement mode on the FPC1500. It automatically put markers on the four peaks and revealed the TOI value. Looking at the markers, it can be determined that the intermodulation products are around -24dBc, i.e. 24dB lower than the wanted signal. Ordinarily one would have to start doing some arithmetic to determine the TOI value, which is fine for a one-off measurement, but really awkward when tweaking a circuit and wanting to rapidly know the new result in real-time. The FPC1500 of course provides a dynamic TOI value changing in real-time, not a static measurement. In summary the TOI feature is executed nicely.
A spectrum analyzer can be used to measure power specifically in a radio channel. It can then be compared with the output from a power meter, to know precisely how much power is being wasted. To perform this task, the spectrum analyzer integrates the area under the sweep trace, across a certain bandwidth that the user can define. For simple shapes it is easy, for example if the spectrum analyzer displays an output of -40dBm across a 100kHz channel with a resolution bandwidth of 3kHz, then the channel power is
However an automated measurement will do this far more accurately, and will automatically select the correct resolution bandwidth and even the measurement bandwidth if it is to a standard (or you can enter custom manual values too of course).
The screenshot here shows an example of a mobile phone network signal simulation (it was generated with a LimeSDR software defined radio shown in the photo above. It connects to a PC and allows any signal or modulation to be created on-demand). This is a 3G network simulation of a Wideband CDMA (Code Division Multiple Access) channel which can carry a hundred phone calls.
The FPC1500 makes channel power measurements surprisingly easy. Once it is showing the correct portion of the spectrum, select the RMS detector so that the reported levels are suitable for a channel power measurement, and then in the Measure menu hit the Channel Power button. A softkey marked ‘Standard’ reveals itself, and you can use this to choose the particular signal type and it automatically picked the correct channel bandwidth of 3.84MHz and 30kHz resolution bandwidth for me, and instantly displayed the channel power. It couldn’t be easier. No longer does one need to approximate or remember bandwidth settings. 2G, 3G and 4G are covered, as is narrowband IoT. You can create your own settings too and save them into the instrument or onto a USB stick.
Incidentally, one incredible feature of the R&S InstrumentView software is that regardless of what you’re doing with the spectrum analyzer, you can save all the detail into a single file, for review at any time. This is very different to saving a PNG or JPEG screenshot - here is an example:
The screenshot above shows a single dataset file that can be recalled at any time. I saved two sweep traces in it. On the right side, in columns, are all the detail relating to the two sweeps. Furthermore, the dataset view can be manipulated at any time – markers can be added or modified, as well as horizontal limit lines, even new sweeps can be overlaid the following day! and re-saved. This dataset feature is fantastic, because one can share the file between engineers, and as long as the InstrumentView software is installed, it can be viewed and printed and modified. It is like self-documentation for spectrum output, one could print it out and put it into a notebook too. It is a stunning feature.
A very useful feature is the ability to automatically measure harmonic distortion. Such distortion is created by every oscillator or frequency synthesizer, and the extent depends on how well it is designed. It is something that is almost impossible to identify from an oscilloscope display. Here is an example:
It looks like a perfect sine wave. And if you zoomed in, you probably couldn’t find any other sinewaves at multiples of this main (i.e. fundamental) frequency. However, the same signal viewed by a spectrum analyser shows a different story. Now you can see that this signal has some content at two and three times the fundamental frequency (and four and five times and so on although that is not shown here). The FPC1500 can quickly count up the contribution from each harmonic (you can choose how many harmonics) and work out the harmonic distortion as a percentage or more usefully, just to see what the highest harmonic output was relative to the carrier (in this case, it shows it as -45.62dBc or 1/30000th of the power of the fundamental, which is quite excellent but something that would have been impossible to know for sure from the oscilloscope view). For this test, the signal was generated by an Analog Devices AD9851 direct digital synthesis (DDS) chip followed by a multi-element antialiasing filter.
This automated measurement feature works adequately, but I had to manually set the resolution bandwidth to something where the harmonics were visible; it didn’t automatically set it for me. This is minor, but I hope it gets added as an enhancement to a later release. As another enhancement the FPC1500 could advise when it thinks the harmonics are buried in the noise.
Amplitude modulation is not as popular as it once was, but nevertheless the capability exists in the FPC1500 to ensure AM transmitters are precisely adjusted. To try this feature out, an AM circuit was used, built around a Freescale/Motorola MC1496 integrated circuit. It is a kind of universal analog chip for radio projects that can be assembled into circuits to perform different tasks. It was fed with a 7.4MHz carrier signal, and a 1kHz input tone.
An AM signal is not the easiest thing in the world to capture on an oscilloscope:
By observing the calculated modulation depth on the spectrum analyzer, it is possible to tweak the carrier and input tone levels, and the carrier insertion adjustment potentiometer, to get as close to 100% modulation depth as possible. Old school circuits like this had a lot of adjustments! The result was a very high quality AM output.
The screenshot below shows the FPC1500 view showing the center carrier frequency, and the two sidebands carry the 1kHz tone.
You can deep-dive a lot further into the modulation detail with the FPC1500, because it has additional applications installed for this task. It is explored further below.
The Modulation Analysis license provides both analog and digital modulation analysis apps for the FPC1500. In the analog modulation mode, it can reveal the demodulated trace as expected with some information displayed up top:
More interestingly a summary screen provides a lot of detail about the modulation. It is a fantastic level of detail. I really liked that you can load up your own limits and immediately see a pass/fail status on the screen too.
The limit files are created on the PC, using the R&S InstrumentView software. It is a smooth process, very intuitive.
The created limit file is transferred into the FPC1500 using the PC too. It is very slick – no need to mess about with USB memory sticks for this.
In summary I thought the AM modulation features were excellent on the FPC1500.
Frequency modulation has a complicated looking spectrum. It varies a lot depending on the ratio between the amount of frequency deviation, and the modulating frequency. Here is an example from a wireless Citronic beltpack UHF transmitter with a 1kHz input signal at 10mVp-p into its microphone input. An automated Occupied Bandwidth feature was used to automatically draw vertical blue bands around the portion of the signal with more than 99% of the energy.
You can see the peaks are all spaced 1kHz apart and there are lots of them. By changing the aforementioned ratio it is possible to make each sideband spike increase or decrease, and at certain ratios even the carrier can disappear altogether. That’s something ordinarily difficult to visualize, but fortunately with a Spectrogram view it all becomes clear. Here is a view showing the frequency deviation kept constant, but the modulating signal (i.e. the input audio signal) sweeping from about 5kHz down to 100Hz. The interesting pattern is caused when the ratio is just right to cause some of the sideband content or the carrier to reduce or disappear.
Going into the FM analysis application, a summary screen appears, much like with AM analysis:
From here the frequency deviation and other measurements can be read off. A demodulated trace can also be viewed if desired, like the AM analysis feature shown earlier.
Frequency Shift Keying (FSK) frequency modulation is highly popular for creating the air interface in many systems. I was curious to see if sub-1GHz devices could be analysed with the FPC1500. With an antenna attached, there was lots of activity to each side of the European 868MHz ISM band. The left side shows a couple of 4G/LTE channels, each 10MHz wide. The right side reveals several 3G WCDMA channels. Right in the center was the 868MHz signal I was interested in (from a Texas Instruments CC110L device in a test mode).
It is inadvisable to run such a test for long with an external antenna, because the band is only licensed for short bursts. Zoomed in, you can see that my transmission was very bad and in the spectrogram, you can also see a well-behaved device that only transmits intermittently a few hundred kilohertz higher.
Zoomed in further, the transmission from the CC110L can be seen to occupy around 10kHz of bandwidth.
The FSK analysis application can be enabled and it will display a demodulated trace. By default it can look a mess because parameters need to be defined. To set the parameters, I used trial-and-error. It isn’t too difficult, because I know that the CC110L is used for sending data at a few kbit/sec in a typical configuration. As a first step, I enabled ‘Burst Processing’ in the FSK analysis application because the CC110L sends bursts (packets) of data, not a continuous stream, so I wanted the FPC1500 to ignore the periods when there is no FSK being transmitted. It cleaned up the output slightly as a start.
Next, I varied the Frequency Deviation setting until the displayed output was of the correct amplitude and filled the screen height (up to the +1 and -1 values on the y-axis):
Finally I tweaked the symbol rate until the eye diagram appeared out of the mess:
The eye diagram shows a good received demodulated signal, and an engineer would observe this while tweaking their design for the best clean output.
A popular use-case with spectrum analyzers is to be able to probe signals on a circuit board without overly loading the signals. This is useful to see what activity is going on, and also to see what is being radiated. Two ways to see the emissions are to use magnetic field probes, and electric field probes. Two features in a spectrum analyzer that could help are (a) high speed to be able to quickly examine a large portion of the spectrum before narrowing down into a particular area, and (b) a low noise floor to see the weak signals. The FPC1500 in theory should work well for this, so I wanted to try it out.
It is possible to either purchase or use home-made probes. Good probes will have a defined response, however home-made probes are sufficient for at least seeing activity. I used home-made probes that are a work-in-progress.
Armed with the magnetic field probe I set the FPC1500 to its default configuration (by pressing the Preset button) and then the FPC1500 performs sweeps of the entire 3GHz spectrum at a 3MHz resolution bandwidth every 540 msec (I counted how many updates visibly occurred in a minute and then divided). This is more than fast enough to move around on a circuit board and see signals come and go. For the board under test, I used a Raspberry Pi 3B. I could clearly see activity around the CPU of course, and the USB/Ethernet chip, and around the power supply section, in different portions of the spectrum for each of these sub-systems. The power supply section had activity of interest in the lower portion of the spectrum, and executing a ‘stress’ command on the Pi caused it to visibly rise. This immediately shows the effect of processing activity on the power supply components. I reduced the span of interest to 50 kHz to 300 MHz, and the resolution bandwidth down to 100kHz. The update rate was very usable again – this time faster, at around 300 msec per update (it was hard to count at this speed so it is approximate), and again this is ideal for probing around when you’re unsure of the particular area of the spectrum that you will be interested in. The noise floor is down to -90dBm at this point with no averaging.
The screenshot shows the probe activity when it is not near the Pi in yellow, and the activity picked up by the probe in green when placed on top of the Pi’s inductor near the power supply section. The brown trace shows the effect of enabling the stress command.
Incidentally the InstrumentView software was really handy for this use-case, because I didn’t need to fiddle around with placing markers while probing the board. I could do that afterwards with the captured data, because the software captures the detail including all settings, and not just a screenshot. I could then explore further to see the spacing between the humps on the screen (about 8MHz in this case but it varied) and the power levels.
Once you have a broad idea of the area where the signals of interest are present, then there could be a desire to more accurately narrow things down in space to a single trace on a PCB. An electric field probe comes in handy because it has a single area of sensitivity ideally, instead of a loop. A coax cable with a slightly exposed (but insulated!) end away from the coax braid could help. Often the signal is weak so some amplification helps. Again, I used a DIY work-in-progress probe with the FPC1500.
Here is an example signal, probing at the traces running from a 100MHz oscillator on a circuit board (an AD9954 frequency synthesizer board).
The great thing about using a spectrum analyzer is that it doesn’t load the circuit much at all with an E-field probe (because it doesn’t touch it – so it looks like a very small capacitance) compared to using an oscilloscope and x10 or x100 probe. Intrusively measuring with such oscilloscope probes can sometimes be very difficult with high frequency and high impedance signals.
I was able to trace over tracks on the PCB and see that this 100MHz oscillation was combining with the intended signal, so basically the 100MHz clock was leaking where it shouldn’t. In the screenshot below, the second hump is the correct intended signal; the first hump is the leaking clock.
This was great insight, and it really helps to know that the FPC1500 is speedy and handy to have around to trace signals over a PCB.
So far all the features that require external stimulus only have been explored. But the FPC1500 also supports signal generation capabilities and it opens up a range of possibilities for engineers.
A typical use-case is to set the signal output into a ‘tracking generator’ mode so that the output frequency follows or tracks the trace sweep on the spectrum analyzer input. It simplifies the task of measuring the frequency response of filters and amplifiers.
In normal use a cable would be used to initially join the input and output to normalize the system, and then the cable would be disconnected and the circuit under test (for example a filter or amplifier) would be inserted. An interesting capability is that the FPC1500 can offset the tracked frequency by a fixed amount, which is very unusual.
I initially used the tracking generator to test an amplifier circuit based around the Avago/Broadcom MGA-30889.
First the cables were attached without the amplifier. I used two attenuators, and a back-to-back adapter to connect the two cables together. As expected there was some unevenness:
Hitting the Normalize button sorted that out:
Next the back-to-back adaptor was removed from the cables, and the amplifier circuit was inserted instead and powered up.
I adjusted the scale, and it can be seen that the output is quite flat with approximately 15dB gain to beyond 2GHz.
This doesn’t quite meet the stated performance in the Avago/Broadcom datasheet, but then my circuit was constructed on a cheap PCB and I used slightly different components compared to the datasheet circuit.
Earlier in this review we examined the phase noise of the oscillator within the spectrum analyzer. The next question is, how can we measure the phase noise of an input signal?
In nearly all cases where you might want to measure it, the input signal phase noise will be lower than the height of the phase noise skirt in any basic spectrum analyzer. The typical scenario when that would occur is when you’re trying to measure the phase noise of an external oscillator, and it normally requires a high-end spectrum analyzer. However one method available with a low-cost spectrum analyzer is to adjust the test such that the external signal is at a convenient frequency where you have a very sharp external filter available. You can use the filter to only pass one side of the external oscillator’s own phase noise skirt, allowing you to see the detail! This sort of test entails having a very sharp filter however (i.e. a crystal filter) – but these can be purchased, especially at typical frequencies used inside radios, such as 10.7MHz. It is a convenient way of performing what is a more sophisticated measurement, with a low-cost spectrum analyzer.
You can even make your own crystal filter to perform the tests at a more suitable frequency. It requires precisely characterising the crystals because each one can be different. The FPC1500 has vector network analyzer (VNA) capability which helps with that! The VNA functionality is discussed further below.
This low-cost phase noise measurement procedure requires the FPC1500 to have the tracking generator capability, and also ideally the preamplifier license.
For a test, I used a 10.7MHz filter with an advertised 7.5kHz bandwidth. I was able to find the frequency response of the filter by using the FPC1500’s tracking generator capability. The response is shown in the screenshot below; the difference between the two markers is 10kHz and almost 85dB. This is very useful, because it will be possible to place the input signal (for which we want to measure the phase noise of) in the stop band area on the left, and see the noise in the passband of the filter, without the measurement being affected by the phase noise of the spectrum analyzer’s internal oscillator.
Next, I switched off the tracking generator function and attached the filter instead to the signal for which I wanted to measure the phase noise (it was the AD9954 board used earlier too).
The test signal was set to 10.69MHz, so that it would sit in the stopband of the filter. Without the filter being present the level would have shot up very high, and caused the phase noise skirt of the spectrum analyzer’s internal oscillator being visible. But because the test signal was in the stopband of the filter, it just revealed itself as a low power level (-110dBm) as visible in the screenshot below. Then, looking 10kHz to the right, I could see the phase noise of the test signal peeking out above the noise floor of the spectrum analyzer (the preamplifier is useful for this). From here it is surprisingly easy to work out the phase noise.
So, by using the tracking generator and the preamplifier, we are very close to the final measurement. The penultimate task is to set the marker mode to Noise – and it will magically work out the phase noise based on the position of the first marker and the second marker. It is the penultimate task because we need to make a manual correction to account for the filter being present. This is easy; subtract the 85dBm that we measured earlier being the difference between the stopband and the passband. So, the phase noise of the AD9954 circuit at 10kHz offset is about -130dBc/Hz. This I believe sounds correct. The theoretical typical value from calculations based on the AD9954 datasheet would suggest -137dBc/Hz is achievable at 9.5MHz (I tested at 10.69MHz). More work would have to be done to investigate if (say) the power supply for the crystal oscillator on the AD9954 board needs improving for example. Without the FPC1500, I would have no idea if the performance of the AD9954 board was up to scratch or not.
In summary it was great that the FPC1500 worked really well for phase noise measurements, even for relatively high-performance signal inputs such as the AD9954. It cannot replace a high-end spectrum analyzer, but to work on a budget, the preamplifier and tracking generator can help a lot, and the noise marker saves a lot of headache.
In recent times some low-cost spectrum analyzers have come equipped with tracking generator capabilities, usually as an option. It is a convenient feature, but anyone who has signal generation capability usually can get away without requiring it. In brief a tracking generator is a signal generator that can be set to sweep frequencies as the spectrum analyser sweeps its input response. The tracking generator output can be plugged into the equipment or device under test to provide stimulus, and the spectrum analyser input captures the response of the DUT. If the DUT is a filter, then the tracking generator plus the spectrum analyser can be used to provide a frequency response curve. The tracking generator plus spectrum analyser functions as a kind of scalar network analyser – the network being the device under test. The actual signal from the tracking generator often isn’t of as high a quality as from a dedicated standalone signal generator.
There is another very attractive use-case for a tracking generator or signal generator, when it is combined with a directional coupler. It results in a system that can perform a 1-port network analysis. In other words, the device under test has its response measured at the same connection that the stimulus is sent to, instead of passing the stimulus through the DUT and measuring the output from another port on the DUT. The way the 1-port analysis works is that a signal is sent into the DUT, and if the DUT looks like a purely resistive 50 ohm load, then no energy is reflected back. If some energy is reflected back then it is displayed on the screen of the 1-port analyser. To achieve this capability, the directional coupler serves to direct reflected energy into the input on the spectrum analyser.
Nearly all spectrum analyzers require the coupler to be an external separately attached item, so it is an additional cost. The FPC1500 is amongst the few exceptions because it has it built-in (some Keysight basic spectrum analyzers have it built-in too).
Where the FPC1500 is extremely unique (the upcoming Siglent SVA1000X also has a similar capability, but can only operate to 1.5GHz) is that not only can it perform the scalar network analyser capability, but it can also perform a vector network analysis in that 1-port mode of operation! The vector capability is significant. With just scalar capability one can see if the match is good or not, but there is little insight into how any mismatch can be resolved. The VNA function provides deep visibility into the attached device/circuit under test. It helps design better circuits. A VNA has always been an exotic instrument, but it shouldn’t be. They used to be even more expensive than spectrum analyzers, but costs have come down. Nevertheless, when I decided to build a home-made VNA a few years ago based on a design I found online, it ended up costing around $500 (not including time!), and it only supports up to 60MHz capability. The VNA option on the FPC1500 costs less than $1000, and extends to the spectrum analyser frequency capability. It would have been the icing on the cake to have full 2-port VNA capability too, but as it is the FPC1500 supports 1-port VNA capability, and 2-port scalar network analyser capability.
This section (and the next section) will go into some detail to describe Smith charts because they are highly useful, and the FPC1500 can directly output such charts. If you're not interested, or are already used to them, you can skip the two sections without losing context provided you're basically aware that these diagrams allow you to see what components to use in series and parallel to make an impedance match). Smith charts look like unusual astrological drawings often with strange loopy shapes and back in the days engineers would pore over printed versions of them on drafting film or tracing sheets. The charts provide for a graphical way of solving a number of problems when designing RF circuits. One typical use-case is when an antenna needs to be matched to a radio circuit. If a radio transmitter is connected directly to an antenna without any regard for matching, then some of the transmit energy will be reflected back into the transmitter circuitry, and it could damage the transmitter. A scalar analysis using a directional coupler will indicate if the antenna is a good 50 ohm match or not. But it won’t tell how to correct for it – i.e. what circuit to design and what component values to use, to match the transmitter output impedance, to the antenna impedance? Vector analysis capability, and specifically a representation on a Smith chart, will provide easy insight to proceed with designing and testing a matching circuit.
About the best reference to understanding the Smith chart can be found in RF Circuit Design by Chris Bowick.
The diagram here shows a typical Smith chart. The following things can be observed:
The red horizontal line through the center is the line of pure resistance. The resistance is zero ohms on the left, and reaches infinity on the right side. The markings along that line are the resistance in ohms, or can be scaled up or down – often things are normalized to a scaling of 50:1. The example point marked A represents a resistance of 50 ohms if that position is normalized to the multiplier 50, or anything else if suitably scaled.
Anything above the horizon has a positive reactance (inductance). Anything below has a negative reactance (capacitance). The solid circles are lines of constant resistance, and the dashed lines indicate reactance values. So, the point marked B has a reactance of +j0.5 ohms (+j25 if the chart is normalized), as well as a real component of 0.2 ohms (10 ohms if normalized). Point B represents an inductance with a bit of resistance. Point C is below the horizon so it represents a capacitance, in this case the reactance is between -j1 and -j2 ohms (because point C is in-between two dashed lines) so approximately -j1.3 ohms perhaps (-j65 ohms if normalized), and the resistance is 2 ohms (100 ohms if normalized).
The actual inductance and capacitance values for points B and C depend on the frequency at which the reactance values were determined. So, if B and C were measured at 3MHz, then point B, which represents a reactance of j0.5*50 = j25 ohms has an inductance of 25/(2*pi*3MHz) which is about 1.3 microhenries. The point C reactance is -j1.3*50 = -j65 ohms and the capacitance would be 1/65*2*pi*3MHz which is about 0.8 nanofarad.
The workflow with the FPC1500 for this task is pretty straightforward. The goal could be to get better communications range and performance, by matching up the antenna more accurately with the transmitter. The first step is to understand what impedance the transmitter has; usually this is 50 ohms. Nearly all RF products use 50 ohms impedance to connect up circuit building blocks internally, and it can be verified in the product datasheet and by connecting to a spectrum analyser (through an attenuator, or through a coupler connected to the transmitter and a 50 ohm load) and confirming the expected power is being transferred. If the impedance is different, this is ok; it can still be matched. Next, the impedance of the antenna is measured. This is achieved using the 1-port VNA capability on the FPC1500, and the result can be observed on the Smith chart view.
As an example, if a 433MHz antenna is measured and it has an impedance of 12 – j60 ohms, and it has to be matched to (say) a 50 - j5 ohm source, then the antenna impedance is plotted on the Smith chart, and so is the complex conjugate of the source impedance (i.e. change the sign of the imaginary part, so that the complex conjugate is 50 + j5 ohms).
It would look like this on a normalized Smith chart (S* is the complex conjugate of the source impedance, and L is the load i.e. antenna impedance):
The task is to get from the load value to the source, and it can always be done with a combination of series and parallel (shunt) components. If the source is also generating harmonics, then the desire could be to make the matching network also act as a low pass filter. A series inductor and a parallel capacitor would achieve that. In the example here, inductive elements are needed to get from the load to the source location. Some elements could be in series, and some in parallel. A method is needed for seeing the reciprocal, i.e. admittances, so that the Smith chart can be used for parallel components too. The trick is to take the Smith chart, rotate it by 180 degrees and overlay it. The blue curved lines emanating from the point on the left side are lines of susceptance values. Any path along a blue circle therefore represents a parallel (shunt) inductance or capacitance. So, one example path from the load to the source points could be a series inductance (indicated by the red arrow path in the Smith chart below) up to the intermediate point marked ‘I’, followed by a shunt inductance (blue arrow).
The lengths of the red (series) and blue (shunt) paths are j0.8 and j1.9 respectively. The series reactance is j0.8*50 = j40 ohm (i.e. the normalisation is removed) and the inductance is therefore 40/2*pi*433MHz = 0.015 microhenries or 15 nanohenries.
For the shunt inductance, to convert to an impedance, the reciprocal of the shunt path length is 1/j1.9 = j0.53 which after the normalisation is removed is j26 ohms. This reactance at 433MHz is 26/2*pi*433MHz = 0.009 microhenries or 9 nanohenries.
The matching circuit is solved!
To those new to the VNA world, the calibration procedure initially sounds bizarre. It requires precision machined connectors that either seem to do nothing, or are a short circuit, and cost a thousand dollars. The secret to understanding the procedure is to temporarily picture the situation from the time domain instead of the frequency domain: when trying to determine the precise behaviour of the attached device under test, the VNA generates a signal that sweeps across a frequency range. You can imagine that the signal will travel along the coax cable, and then it will be either absorbed fully by the load (if the load is a perfect 50 ohm impedance) or a part of it will get reflected back if there is a mismatch, possibly out of phase, if the load has some capacitance or inductance. By examining the returning signal and its phase, it is possible to gain insight about what the load looks like. But to do that, the the location of the load matters too; if it is connected slightly further or nearer (for example by using slightly longer or shorter cables) then this impacts the delay for the reflected signal to return, and it is visible to a VNA. Armed with this knowledge it is clear that we need some way to calibrate the VNA and any connected cables, such that the VNA displays an open circuit result on the Smith chart when no DUT is attached but the cables are present. The procedure to do that is to use a calibration kit that consists of known impedances with the same connector as the device under test, and calibrate the VNA with the known impedances. One typical way is to use an open circuit, a short circuit and a 50 ohm load, i.e. impedances that are far apart on the Smith chart. But the open circuit has some capacitance at the end, and the short circuit has some inductance across the short, and the 50 ohm load may not be perfect either. The calibration impedances will ship with coefficients or a dataset that will describe them in detail. Thus, once the VNA has been loaded with the coefficients and a calibration has been performed, the VNA should display an open circuit until the load is attached.
The slightest change in capacitance or inductance can have a significant effect in the measurement and therefore a lot of care needs to be taken during calibration and during any measurements, to not disturb the environment or the connections. If a connection loosens slightly, or the room temperature changes, it will impact the results. It is good to keep an eye on the connections and not move the cables too much. It is also important to allow equipment to warm up, so that it is in equilibrium, and to recalibrate if the environment or the cabling gets moved.
The calibration kit is expensive because it is machined to accurate dimensions, and it comes with measurements and coefficients that are unique to that kit. My calibration kit unfortunately does not have coefficients, so any measurements will have error. The R&S InstrumentView software has a Calibration Kit configuration page where the coefficients are entered and a file is uploaded into the FPC1500. I manually had to provide guesses for the coefficients, and I experimented a bit until I got some vaguely sensible results. For any real use, a proper calibration kit with supplied data is essential. To save immense headache, it would be highly recommended to purchase the electronic calibration (Ecal) kit for the FPC1500. Without it, one needs to resort to manually attaching calibration standards as mentioned, and the scope for error is very high without significant experience, since the conections have to be unscrewed and screwed to swap out each known impedance. Incidentally the calibration method has different names; often it is referred to as OSL (open-short-load) or OSM (open-source-match) or SOLT (short-open-load-through) where the latter has a through-connector for calibrating a network analyzer with two ports. The electronic version internally contains electronic switching that is commanded by the USB connection to the FPC1500. The coefficients are accessed automatically through the USB connection too, so it is a 1-touch button operation to calibrate it all.
The FPC1500 supports the ZN-Z103 electronic calibration device which appears to be good value. It would be a fairly essential purchase to have a reliable and fast calibration procedure.
Once I had my cable attached to the VNA port on the FPC1500. I invoked the calibration procedure and followed the on-screen instructions that prompted me to attach the mechanical calibration components, i.e. the open, short and load impedances to the cable. After this, the calibration device was removed and the Smith chart showed the impedance to be close to the right side (infinite resistance). It was not perfect, and I will have to eventually obtain a calibration kit or ecal device.
Incidentally, high quality cables and connectors are needed for any VNA use. The slightest fault is visible. The cable in the photo above is hand-made, from semi-rigid coax with male SMA connectors on each end and ferrite beads on the outside – that is useful for some test-cases, to act as a high impedance for signals on the outside shield. The more ferrites the better, for good attenuation at a range of frequencies. The semi-rigid coax helps to keep things stable, because if the cable flexes during calibration or test, then an error could be introduced. Having said that, I also tried a short length of RG316DS cable, and results were good with that too. I used a female SMA to female type N adapter, and the end of that became the calibration plane (technically the calibration plane is about 9mm deeper, inside that female N connector). It means that once the calibration procedure has been established, any device under test must attach to the same calibration plane location.
Although this will not be anywhere near as accurate as the methods mentioned above, for basic testing, a home-made Open/Short/Load can be used. One method is to obtain connectors, and trim off the end and then attach either a short (e.g. copper sheet) or a load (e.g. a couple of 100 ohm resistors connected electrically in parallel).
Alternatively, at one point Texas Instruments offered an Antenna Design Kit with code CC-Antenna-DK however it is no longer available for sale. The kit included a printed circuit board version of an open/short/load with SMA connectors. Since the Gerber files are still available, it could be possible to get the PCB manufactured and build that up.
I edited the Gerber file, to remove the antennas, and it is attached below, ready for sending off to a PCB manufacturer.
The FPC1500 shines for antenna measurements and adjustments. After VNA calibration I attached a dual-band Wi-Fi (2.4GHz and 5GHz) antenna to the calibration plane position and observed the 2.4GHz match on the VNA return loss display. This instantly tells you if the antenna is tuned to the correct frequency, by looking for the dip. Where the dip occurs, it means that there is little reflection, and so the antenna is ‘doing something’ with that power – hopefully radiating most of it.
The FPC1500 can directly display return loss or VSWR. Return loss can be converted via a table or graph into a voltage standing wave ratio (VSWR) value if desired.
With the antenna hooked up, the return loss chart showed that either my measurement, or the antenna, was off very slightly, because the specification for the antenna was 2:1 VSWR between 2.400-2.483 GHz. It is highly likely to be calibration error since I do not have accurate coefficients:
Unlike a scalar network analyzer or spectrum analyzer with just a tracking generator, the VNA provides much more useful information. Not only does it reveal if the antenna is tuned, but it will also indicate the phase information, allowing the engineer to create a matching circuit. Without that, one would be blind about how to modify the circuit to achieve a match. In the Smith chart mode, all is revealed:
The area of interest is between the M2 and M3 markers, and the complex impedances are also indicated on the left side of the chart. Near the center frequency for the antenna, at marker M1, there is almost a perfect match at 47 ohms and very little reactance. If we wanted to move the M2 position closer to achieving the specification, then the Antenna Matching Procedure in the earlier section would be used, and again the return loss view could be used to see the effect across the band.
In summary I was very happy with the VNA view, and the result seemed very stable even after an hour had elapsed! It does require decent quality cables of course, and care not to move or bend cables too much between calibration and test. There are some techniques to ensure an accurate measurement; for example the antenna should be positioned away from the user, in the same orientation as it will be used, and with a ground plane if that will be used in the deployment too. Also, as mentioned in the VNA calibration section earlier, to reduce RF over the shield of the cable, ferrites can be stringed over the coax.
For another look at antenna measurement, this time with a very wideband antenna, see Antenna Measurement with the R&S FPC1500 Vector Network Analyzer
There are myriad use-cases for a VNA. One very useful one is to example the Q factor for inductors. Ordinarily inductors do not come with a N or SMA connector, so some adaptation or test fixture is needed to work with the VNA. Since my calibration kit has a type N connector, I used a procedure where I would swap out the SMA-to-N adaptor from the end of the cable with a SMA to wire end coax, cut to be as close to the same electrical length as the adapter that is removed. The final result was good, but it took three attempts much like Goldilocks; the first was too short by a few millimetres, the second was too long and the third was pretty good. It is worth buying several SMA connectors for this exercise, because unfortunately some will be wasted unless you can clean them out. Once you have made up the ideal cable/test fixture, keep it for future use too because it can be reused provided the calibration always uses the adapter that was used previously and then removed.
I don’t have a calculated procedure to do this, I just manually tried different lengths until the impedance looked entirely real (i.e. with no imaginary component) on the Smith chart, and then applied a short across the cable and observed the short to be in the correct position on the Smith chart too.
As an experiment I wished to examine the Q for a Wurth wireless charging coil which is used for mobile phone charging. It was part of Wurth’s Medium Power Wireless Charging Demo Kit. According to the data sheet it has a typical Q of 30 at 125kHz. The VNA operates down to 2MHz so I was curious what the Q could be at a few MHz because this was not in the Wurth datasheet since wireless power charging does not operate at these frequencies. Knowing Wurth, I figured they were conservative with their Q value. But I didn’t know if the Q would be higher at such frequencies, or lower.
The strategy for this scenario was to solder the inductor to the cable and place a silver mica capacitor in parallel. They are ideal because they have extremely high Q, to minimise the effect on the resistance value which we wish to read off in the next step.
With the parallel inductor and capacitor, the Smith chart can be examined and at the point where the impedance is purely real (i.e. anywhere on the center horizontal line) the resistance can be read off. This is the dynamic resistance of the parallel LC tuned circuit. From there it is quick to calculate the Q, being
If you’re interested in the Q at a particular frequency the capacitor value can be changed by adding more or less capacitors in parallel. It is worth keeping various value silver mica capacitors in a box, for such use. They are not cheap, but they can be desoldered, saved and reused. The Cornell Dublier CD15 series is suitable – for about $40 a good selection can be purchased.
The chart shows that at 2.888MHz the impedance was almost entirely real. I would have needed a more granular sweep near there in order to get the value even closer to a pure real value.
Plugging in the values, this means that the Q is about 35 at 2.888MHz. The value can jump about significantly, and this is because the VNA is designed to operate in the ballpark of 50 ohms. It isn’t a purpose-built Q meter or vector impedance meter, however it is good to see that it can be used as such, provided you can arrange your test to fall into an area where the resistance is less than around 5-10k. You can still do a lot with this, and the technique shown here could be improved too for more accuracy. Nevertheless it showed that very useful component parameters can be determined with the FPC1500 - it was impressive that it was quite easy to measure inductor Q with little effort, and the trial-and-error time taken to build the short length coax test fixture will come in handy for other uses with the VNA too.
The VNA capability has a cable Distance to Fault (DTF) mode where it can take the frequency domain information and convert to a time-domain view but with cable distance on the x axis. In theory it should provide extremely deep insight into the cable behaviour, so I was curious to try it out.
The feature is really neat. It can provide detail about the match to 50 ohms all the way along the length of the cable. You can also spot any connectors along the way, and any terminations too, because they will never be a perfect match to the cable, so the transitions will be apparent.
To use the feature, a calibration procedure is first executed as normal. Next I attached a very high quality cable from Times Microwave – a 3 metre length of LMR400. It was possible to follow the return loss, or the standing wave ratio (SWR) all the way along the x-axis up until the 3 metre position, and there the termination or lack of termination was very clear. The screenshot shows at the 3 metre point in the faint trace that the SWR was off the scale when the cable was not terminated. With the correct 50 ohm termination (the bright yellow trace), there was a small peak at 3 metres to a SWR value of 1.1. This shows that not only was the cable performing extremely well across the whole 3 metre distance, but that the termination was quite good too.
This type of test will show up the slightest imperfections in the cable or the termination. I dread to think how bad home television antenna cable installations are in comparison, given the junk cables sold by DIY stores. For a professional installer, it could be worth building up a set of adapters to interface with such cables – it is hard for a customer to object to swapping out a cable when it is plainly visible on a graph how poor the existing one is performing! I think the most interesting cable installation I’ve seen was where the antenna installer did not cut the cable to the correct length, but instead stapled a very long length of home antenna cable to the outside of the home in a several-feet high meandering square-wave like pattern. I would have loved to see what the FPC1500 made of that : )
The FPC1500 also has a mode where it can measure cable loss too. Again it uses the VNA capability for this, so it can perform the measurement even when the cable is installed and there is only access to one end of the cable for connecting to the FPC1500.
Although the cable DTF feature described above is ordinarily used with coax cables, I was keen to see if it could also work with network cables such as office local area network (LAN) wiring using Ethernet. If it worked, this would be extremely helpful for working at nearly every site because LAN cabling is used nearly everywhere. Compared with coaxial cable, the technology is similar but slightly different. Office LAN cabling such as CAT5 cabling relies on unshielded twisted pairs with a 100 ohm characteristic impedance, whereas coax cables are very often 50 or 75 ohms. So, some method is needed to transform the 100 ohms to look like 50 ohms to the FPC1500. I used what I had, which was a 0.2-400MHz capable transformer model PWB-2-BLB with 1:2 impedance ratio. Unfortunately this is not ideal, because the FPC1500 will want to use higher transmit frequencies if possible, to better characterise the cable. 400MHz is extremely low for this task. However, I tried it anyway. I knew it was unlikely I would get good results for the reason mentioned, and also because my manual calibration is only adequate – a decent calibration kit or the ZN-Z103 automatic electronic calibration device will provide far better results.
I soldered the transformer to a RJ45 socket for easy connection to a LAN patch cable for testing. For now, I just chose one of the four twisted pairs to test. For the far end of the cable, I took another RJ45 socket and soldered a 100 ohm resistor across two pins.
For the cable model, as a quick test I entered a velocity factor of 0.7, and did not make any other tweaks. I used an 8m length of LAN patch cable for the test.
The faint trace hump at the marker M1 shows the unterminated result, and the bright trace shows the terminated result. I still need to interpret how the test works using the transformer for impedance transformation, and I will refer to R&S to find out if they have suggestions for improving this test. One method could be to just ignore the impedance transformation, and directly connect the network cable and accept the mismatch, but I will ask to find out if they have other ideas too. This is an unusual test, since the FPC1500 is not intended for network cable testing.
Still, I was very happy to see that in principle the procedure could work! I have been in the situation in some businesses where, configuring end user equipment, days were wasted troubleshooting 100m-long connections, and while there are tools to support such tests, all too often the procedure after cabling is to wait until end user equipment is attached (which could be days or weeks later) and then see if it functions adequately. Using the FPC1500 in theory should provide awesome immediate insight. It is hard to hide a fault from a VNA capable device – even a slightly crushed or overly bent cable should be identifiable, since if the wire twists are disturbed, it will impair the characteristic impedance of the cable. Bit error tests with other equipment would be blind to this unless the impairment was more extreme. Either an eye diagram view, or a VNA is required for such detailed tests.
In order to get stuff done as described in this review, in terms of costs, it is hard to say which licenses should be purchased because it depends on the requirements and tasks of interest. For non-RF engineers and for minimizing budget, the base FPC1500 plus the VNA license would be really welcome for any engineer, for the huge mix of capabilities that it provides.
A near-essential extra would be the ZN-Z103 electronic calibrator for the VNA function to be useful, reduce the risk of error and save a lot of time during calibration (calibration is a necessary task when using any VNA).
The preamplifier is a nice-to-have but not essential initially.
For RF engineers, then of course the frequency upgrade licenses would be particularly useful, and all the various options related to radio tasks could be of interest.
You'll also need good quality (i.e. a lot better than usual when using the VNA) assembled cables, connectors and attenuators. Despite having a tracking generator, an external signal generator is extremely handy too.
Here are the order codes for the FPC1500 (click on them for the latest pricing). The top level order codes get you the physical hardware, and a frequency upgrade if desired. Otherwise the frequency upgrade licenses can be purchased later.
The feature licenses are in rows marked 6-11 below, and I've tried to place them in an order that I think would be useful to most engineers, i.e. uppermost rows being of most interest (subjective of course). RF engineers will know what they want anyway.
|Top Level Order Codes|
|1||FPC1500 Base Unit (1 GHz Frequency Range)|
|2||FPC1500 Base Unit with 1-2 GHz Frequency Upgrade|
|3||FPC1500 Base Unit with 1-2 GHz and 2-3 GHz Frequency Upgrade|
|Optional Frequency Upgrade Licenses (can be purchased afterwards)|
|4||1-2 GHz Frequency Upgrade|
|5||2-3 GHz Frequency Upgrade|
|Optional Feature Licenses (can be purchased afterwards)|
VNA (Vector Network Analyzer)
Don't forget to obtain a calibration tool such as the
(Spectrogram, 3rd Order Intermods, Harmonics, Channel Power,
TDMA Power, Occupied BW, AM Modulation Depth)
|9||AM, FM, ASK, FSK Modulation Analysis|
|10||Receiver Mode (Radio channel views)|
|11||WiFi Connection support (2.4 GHz)|
All sorts of cables, adapters and attenuators could be required. For VNA use, they need to be of a good quality, and by that they ought to come with a datasheet. Generally ebay ones will not be appropriate, and could even damage your good adapters, so don't mix them up if you already have some cheaper parts.
These are just some various low-mid price suggestions that I've used or could be suitable alternatives. Not all are needed, these are just some ideas.
Cylindrical type. Suitable for 10-1000MHz. 6.35mm inner diameter, 9.9mm outer diameter, 19.5mm length
Ideal for stringing on coax before the connectors are soldered.
|2||Split, snap-on type. Suitable for 300-2500MHz. 8mm inner diameter. Ideal for coax with RF connectors already soldered.|
|Coax and Assembled Cables|
|3||Semi-rigid, 3.58mm outer diameter. Bare, so you may wish to apply a sleeve to it.|
|4||250mm length RG316-DS, pre-assembled with SMA connectors|
|5||500mm length RG316-DS, pre-assembed with SMA connectors|
|6||SMA connector for RG402 coax|
|7||53S132-K00L5||N male to SMA female (ideal for leaving attached to the FPC1500) - Rosenberger|
|8||N male to SMA female (ideal for leaving attached to the FPC1500) - Radiall|
|9||53K132-K00L5||N female to SMA female (ideal if you have an OSL calibrator with N male connectors)|
|10||32S103-S00L5||SMA male to SMA male - Rosenberger|
|11||SMA male to SMA male - Radiall|
|12||32K101-K00L5||SMA female to SMA female - Rosenberger|
|13||SMA female to SMA female - AMP/TE Connectivity|
|50 Ohm Termination|
|15||49.9 ohm 0603 sized resistor (useful when trying to calibrate on a PCB)|
|16||49.9 ohm 0402 sized resistor|
|17||K1-VAT+||Set of five 1W max: 3, 6, 10, 20, 30dB - Mini-Circuits|
|18||K2-VAT+||Set of ten 1W max: 1,2,3,4,5,6,7,8,9,10dB - Mini-Circuits|
|19||10dB 1W max - Radiall|
The review started off by exploring who would need such an instrument, and found that in theory the feature set would benefit many engineers and not just in RF engineering disciplines.
The usability of the device is very good, not least because this instrument is dead silent, but also because everything worked, and the user interface was responsive. I found no bug I couldn't work around, which really impressed me in such a new product.
The InstrumentView software is so good, it is hard to ignore that it too forms a big part of the usability of the FPC1500. Results can be documented extremely well using the detailed trace capabilities that capture all settings and allow the user to add markers and thresholds after the data has been captured to the PC. This is very different from a screenshot. All of the configuration files were easy to create and transfer with the software. InstrumentView was reliable and easy-to-use.
The FPC1500 has a number of modes and features, and in its spectrum analysis mode it performed extremely well. It was speedy and users are likely to notice a difference from earlier generation spectrum analyzers.
It has really impressive range and noise floor performance even without the preamplifier. The internal oscillator is designed to keep a clean noise floor beyond a few tens of kHz, and close-in phase noise was not bad either. The very low noise floor and the preamplifier will really help with measuring phase noise of external signals using techniques such as the one described in the review, and the FPC1500 offers assistance with phase noise marker capability which will automatically compensate for the sweep settings.
For traditional RF use the FPC1500 provides helpful features to confirm modulation levels, channel power, harmonic distortion, and so on. To achieve some of the functionality the FPC1500 will demodulate popular analog and digital methods like AM, FM, ASK and FSK.
The very good sensitivity due to the low noise floor, and the configurable sweep speeds allowed for very rapid examination around circuit boards using H-field and E-field probes. In this scenario users may not know what frequencies to expect activity at, so the speed really helps otherwise the user would have to move the probe at an awfully slow speed. Also, I really liked that I did not need to turn on the preamplifier - the sensitivity was awesome.
The signal generator operates in the usual tracking generator mode, but R&S have opened up the possibility to use it for other tasks too. It comes into its own when the FPC1500 operates in the vector network analysis (VNA) mode of course. Even without the VNA capability, the tracking generator is useful for quickly seeing the frequency response of amplifiers for instance. I liked that I could use it for testing a filter, and then using the filter to perform a phase noise measurement. But the VNA capability really opens up the a lot of use-cases.
With the VNA capability it is possible to use the FPC1500 to not only examine if an antenna is matched, but also to precisely see how to modify the circuit to achieve a better match. This is because the one-port VNA capability provides the very useful S11 measurement which shows the phase and the magnitude of reflected power at different frequencies. With a bit of simply maths you can do really interesting things like seeing how components or networks of components behave at different frequencies, measure the Q of an inductor and so on. The VNA feature was rock-solid; I could leave the FPC1500 running for extended lengths of time with little observable change in measurements. This is good to see because it means that internally the FPC1500 has well-designed thermal capabilities which do not cause much drift that would require recalibration frequently.
There are also built-in applications inside the FPC1500 to use the VNA capability to perform cable measurements even if the far end of the cable cannot be connected to a test instrument. It performed really well with coaxial cable, and the possibility exists to use it with network cable installations too.
Throughout all these tasks the FPC1500 performed very well. If I had to suggest anything that I would have wished was different, it would not be a technical capability, it would be the LCD screen. While it is usable, in a typical use-case the FPC1500 will be sitting at desk level with the screen perpendicular to the desk and not tilted, for perhaps the most stable way of connecting cables to it. Unless the desk really is large and the user is further away from the FPC1500, the screen angle of view does cause a drop in screen contrast that is noticeable. Other than that, I would struggle to find any significant flaws.
My concluding thought is that the performance and feature mix is excellent because it provides all engineers with a large suite of capabilities. This is an instrument that will be switched on a lot more than traditional spectrum analyzers, and consequently will result in better engineering. I loved that it was silent, compact and nothing like the boat-anchors from the past - it will be switched on and see use regularly by any engineer, almost as much as an oscilloscope. For that alone, this surely represents good value.
The wealth of tasks that can be performed with the FPC1500 makes it like a Swiss army knife. It’s a multi-function instrument, not a spectrum analyzer.
Note: I hope there are no errors in this review, but if there are, please let me know. It is very easy to make a mistake, especially when covering diverse features in a multi-function instrument. Thanks for reading this far : )