3GHz Spectrum Analyzer - R&S® FPC1000 - Review

Table of contents

RoadTest: 3GHz Spectrum Analyzer - R&S® FPC1000

Author: Instructorman

Creation date:

Evaluation Type: Test Equipment

Did you receive all parts the manufacturer stated would be included in the package?: True

What other parts do you consider comparable to this product?: Although not a directly comparable instrument, the MDO4104-3 mixed domain oscilloscope contains a 3 GHz spectrum analyzer, so I used it as a comparison product for this review.

What were the biggest problems encountered?: A few user interface issues that are detailed in the full review. Nothing serious.

Detailed Review:

July 2, 2018:  New content is available at the bottom the the review.  The long awaited addendum with James Hoffman testing h-field probes is now posted

 

Introduction

My experience with electronic test equipment reaches back to 1974 informally and formally, to my post secondary years learning electronics starting in 1978.  In school, and in the lines of work I pursued after graduating, Spectrum Analyzers (SAs) were always rather rare and exotic instruments.  Not many work places had them in their labs, and they were a shared resource at the polytechnic institute I attended.  If you work with Radio Frequency (RF) electronics on a regular basis as a scientist, engineer, technologist, or technician, you probably work with spectrum analyzers all the time.  For those not steeped in the arcane and mysterious RF world, the time domain based oscilloscope is go to signal analysis instrument.  That said, it is clear that modern electronics technology is very dependent on an assortment of sophisticated wireless data communication technologies and that dependence isn't likely to go away.  Spectrum Analyzers provide frequency domain views of signals.  That means the horizontal axis on a spectrum analyzer displays a span of user selected frequencies, lowest on the left, highest on the right. Oscilloscopes provide time domain views of signals.  That means the horizontal axis on an oscilloscope displays a span of user selected time, earlier on the left, later on the right..  An SA will show you the energy content at each frequency within a range of frequencies, an oscilloscope will show you the way a signal voltage varies over a span of time. Time domain information about a signal can be manipulated mathematically via the Fast Fourier Transform (FFT) method to produce information about the signal in the frequency domain.

 

You may not need a dedicated SA on your bench because many contemporary oscilloscopes provide an FFT option that turns an oscilloscope into a basic spectrum analyzer on those occasions when a frequency domain view is needed. Of course, a dedicated SA instrument will provide a richer set of analysis features and will typically have greater flexibility than a built in FFT feature on an oscilloscope.

 

The Rohde & Schwarz FPC1000 examined in this Road Test is a dedicated entry level SA instrument. It's whole purpose is to allow exploration and measurement of signals in the frequency domain ranging from 5 kHz up to an (option enabled) 3 GHz.  Although SAs have been rare and exotic instruments in my 44 years of experience, I have always found them to be fascinating and useful tools.  The perspective they provide into the world of signals is quite enlightening and separate from the perspective provided by oscilloscopes, which, in full defense of oscilloscopes, is also rather enlightening.  Together these two instruments provide a robust picture of complex signal behavior.

 

A new review feature

This review introduces stylistic departure from my previous reviews.  For the first time I will be working with another element14 member to provide a dual perspective review.  The first two parts of this review are presented from my perspective.  The third part relies heavily on the perspective of a relatively new element14 member, James Hoffman, an RF and mm-wave engineer working out of California.  We are collaborating remotely with the goal to provide a richer and more interesting review.  Let us know via comments if this format works to achieve that goal.

 

My parts provide a "first date" impression of the user experience followed by a deeper conversation comparing features of the FPC 1000 with comparable features available in the 3 GHz spectrum analyzer embedded within the Tektronix MDO4104-3 Mixed Domain Oscilloscope.  The third section is written as an application case study using the FPC1000 to simulate troubleshooting on a Watkins-Johnson WJ-8611 UHF receiver.

 

Part 1:  A "first date" with the FPC1000

 

More buttons and a new screen to feed my gizmo addiction

The FPC1000 is a handsome instrument that would look impressive on most any bench.  It has a sleek modern and professional front panel (what we used to call the escutcheon in years gone by) complemented with thoughtfully engineered features including a large glare free screen and a clean button layout with easy to read legends.  Though the screen is sized generously, it is not touch sensitive and that is a little disappointing on a new instrument like this.  Many new instruments are coming out with touch sensitive screens.  The ability to touch a waveform and manipulate markers or instrument settings is very appealing because it allows for intuitive and distraction free interaction. With a touchscreen you don't have to pull your gaze away from the details of the waveform to locate and then fiddle with knobs and buttons.  Nevertheless, this instrument does not have a touchscreen, so I will delve into the behavior of the screen and the traditional button and knob interfaces.

 

The handsome and professional looking Rohde & Schwarz FPC1000 Spectrum Analyzer

 

Although the screen is glare free and large, it has less than ideal off axis viewing.  Contrast falls away rapidly as viewing angles increase away from the perpendicular.  The video below provides a fairly reasonable reproduction of what my eyes see as the viewing angle is changed to a right, then left, then above, then below aspect.  The screen is still view-able under all of these aspects, but the contrast is reduced, and this becomes a problem as we will see later on.

 

Effect of viewing angle on screen contrast

 

 

While on the topic of the screen, I took some measurements to see how the viewable spectrum area on the FPC1000 stacked up against the viewable spectrum area on the Tektronix MDO4104-3.  The two instruments are shown side-by-side below.

 

Rohde & Schwarz FPC1000 and Tektronix MDO4000 side by side

 

Initially, the FPC1000 screen looks quite generous, but not all of the screen real estate is devoted to displaying the spectrum.  From the total screen area subtract the menu text area on the right side and the parametric information area across the top. To get a better sense of the differences in screen area available to display spectrum traces, look at the diagram below, which is drawn to scale.

 

The R&S screen has 45 square inches of usable image space (9" x 5"). However, only 32 square inches are set aside for the spectrum traces.  Of the 53 square inches available on the MDO4000 screen, fully 43 square inches can be used to display spectrum traces.  The MDO4000 spectrum trace area is about 34% larger than the spectrum trace area on the FPC1000.  Tektronix and Rohde & Schwarz made graphical user interface design decisions during product development. Tektronix decided to develop a graphical interface with removable menu overlays that temporarily occupy a portion of the spectrum trace area, but can be removed by pressing a front panel "Menu Off" button.  R&S decided to permanently assign screen real estate to display menu and parametric information.  As far as I can determine, these areas are static and cannot be removed.

 

On a clear day you can see your menu options from here

One final observation about the graphical user interface on the FPC1000 before moving on to the button layout.  Some of the on screen selection choices use a very low contrast highlighting scheme that can leave a user (like me) stuck in a state of indecision not knowing which choice is actually highlighted.  See the video illustration below that shows the final step in setting up a WiFi connection.  Now, I will say that if you look carefully, the highlighted selection can be discerned, but my point is that the user should not have to look carefully.  Highlighted menu selections should be bold, unmistakable, and obvious.  The camera actually improves the contrast relative to what my eye sees in a well illuminated room. Can you immediately, and easily, tell which choice is highlighted?

 

Contrast issues on menu selection during WiFi set up

 

Here's hoping your WiFi password is 123

 

The FPC1000 does not seem to store WiFi passwords in non-volatile memory.  When power is cycled off and on, the previous WiFi connection details are lost and the WiFi password must be reentered to establish a wireless connection again.  Passwords are entered by the repeated press method using the front panel alphanumeric keypad.  I attached a USB keyboard to both of the front panel USB ports and neither port recognized the keyboard as a device for entering alphanumeric values for the password.  My particular 12 character WiFi password requires a series of 36 key presses, at best.  This assumes I don't mess up and press too many times on each key to reach the character I'm looking for.  Though tedious, I can see the security advantage this approach offers.  In a shared user environment, holding on to WiFi passwords in non-volatile memory can be a security risk.  It is also possible that the instrument will not be shared.  In cases where use is restricted to one user, it would be nice to have an option to store WiFi details in the instrument to avoid the hassle of entering them every time the WiFi feature is going to be used.

 

IMPORTANT UPDATE: April 14, 2018

 

With the release of firmware V1.30 R&S has eliminated this problem.  I loaded V1.30 firmware on the FPC1000 and checked to see if my WiFi password was retained through a power down - power up cycle.  The password was retained.   Thank you to R&S for responding to this issue in a timely fashion!

 

The alphanumeric keypad used to enter WiFi passwords

 

 

What is up with the cognitively dissonant cursor keypad?

In large part the button layout on the FPC1000 is sensible and user friendly.  You press, for example, "Freq" to get access to Start, Stop, and Center frequency controls, as well as Center frequency Step Size and Frequency Mode (which allows the user to access a variety of tables containing preset parameters for CDMA, GSM, LTE, TV, etc., up and down-links).  There is, however, one cluster of front panel buttons that immediately, though temporarily, led me astray.  The four button cursor navigation cluster, shown below, can, in my opinion, be a bit confusing.

 

The FPC1000 cursor controls

 

The button shapes kind of look like stylized arrows, all pointing toward the center of the cluster, so one might be led to believe the arrow shapes correspond to cursor navigation directions.  This belief, if taken, would be found to be bereft of truth.  It is, of course, the smaller white icons on each button that indicate the cursor navigation direction.  The cognitive dissonance of this cluster of buttons violates the Gestalt Symmetry and Order design principle. The shape of the buttons draws your attention toward the center of the group, suggesting that when you push a button you are bringing the cursor from where it is toward the center.  Instead, the white overlay is the feature you should be concentrating on.  Just for comparison, look at the cursor keys on the Keysight 34470A 7-1/2 digit multimeter.

 

The Keysight 34470A cursor controls

 

The button shapes are angled in a way that draws attention to the central "Select" button, but the buttons themselves are not reminiscent of arrows.  The cluster of buttons gives a sense of an exploding oval, suggesting that when you push a button you are sending the cusor away from where it is in the direction of the back arrow overlay icons.  This whole discussion about cursor button shapes may seem overly nit picky, but I think button layout is an important design consideration that supports, or detracts from, the user friendliness of an instrument.

 

The FPC1000 is a lightweight: No, literally it doesn't weigh much

Another notable characteristic of the FPC1000 that can probably be attributed to its modern design is that it is remarkably light.  It only weights 3 kg (6.61lb).  You know how your brain estimates the force necessary to lift an object based on accumulated past experiences?  Usually, this estimating process works well and you end up lifting objects smoothly.  Sometimes your brain underestimates the lifting force required and the object doesn't move, or, in the case of the FPC1000, the lifting force estimate is too high and you end up lifting it with a jerky awkwardness, like I did the first time I picked it up.  So, from one perspective a super light spectrum analyzer is evidence of modern, efficient design that eliminates unnecessary bulk.  On the other hand, from a practical user level perspective, the instrument is so light that it slides back with nearly every push of a button.  See the video illustration below.

 

FPC1000 is very light and may slide on your bench

 

Watch the far edge of the instrument slide backwards as a coax cable is connected and various buttons are pressed (mostly in a panic to turn off the preamplifier and crank up the input attenuation because the IF Overload alarm was going bonkers while this video was being shot.  Not to worry, the input signal was only at -3 dBm, but the PREVIOUS set up was for signals coming in from an antenna.  (Oops).  The right side of the instrument moves back noticeably as buttons are pressed in this short video.  Imagine what would happen over several hours of use.  I have resorted to pressing my fingers down on top of the instrument and activating buttons with my thumb.  This technique reduces the amount of sliding.

 

A silent piece of test equipment: Nice!

From yet another perspective, the efficient design of the FPC1000 seems to have allowed the instrument to be fabricated sans cooling fan, or ventilation slots.  The FPC1000 makes no discernible noise whatsoever. Nearly every other instrument on my bench has a variable speed cooling fan that hums along incessantly when the instrument is powered.  The following video clip provides a rough (i.e., not calibrated) demonstration of the accumulating fan drone produced by a modern electronics test bench as instruments are powered up in sequence, starting with the outlier FPC1000.

 

Starting with all instruments off in a quiet room, the dB meter shows about 30 dB (A weighted).  The FPC1000 is turned on first, and the noise level does not change, because, the FPC1000 makes no noise.  Then the following instruments are turned on, in order:

 

  1. Tektronix MDO4000 series oscilloscope
  2. Keysight Waveform Generator (33622A)
  3. Keysight Counter (53230A)
  4. Keysight DMM (34470A)
  5. Keysight Source Measure Unit (B2902A)
  6. B&K DC power supply (9130)  NB: Now replaced with a Keysight E36313A.  YES!
  7. Keysight oscilloscope (DSOX1102G)

 

By the time the last instrument is up and running, the background noise level has increased to about 58 dB.  An increase of about 28 dB, just from fans.  I have gotten used to the drone of the fans.  It is actually mildly comforting.

 

Fan noise from a modern test bench

 

A minor quibble

Question for Rohde & Schwarz: What is the point of having a dedicated front panel button that brings up a menu with only one option, and that option doesn't actually do anything?  Just asking.  This is the situation I encountered when my curiosity caused me to press the very intriguing "Mode" button on the front panel.  Pressing the "Mode" button brings up a menu with one item: "Spectrum".  Pressing the "Spectrum" button appears to do only one thing.  It makes the FPC1000 beep.  See video below.  My assumption is that more options appear once additional features are enabled through installation of licensed upgrades.  If this is the case, perhaps the disabled options could appear as dimmed out options in the list, rather than as nothing at all, because not having any other options listed leads one to ask the question I started this section off with.

 

The menu button that doesn't do anything

 

It sure is great to have iOS and Android apps for remote control of test equipment...I guess

I have a Samsung Galaxy smartphone and an iPad mini (see previous admission of gizmo addiction), so I naturally downloaded the free Rohde & Schwarz apps for each platform to try them out.  Both apps provide remote control over the FPC1000 and live updates of the screen.  Both apps sort of feel like pointless additions to the Internet of Things scrapheap.  For decades I have managed to be productive at my bench, to design, prototype, troubleshoot, and tinker, without ever feeling the need to control my instruments remotely. I have difficulty conjuring a scenario where I would have a serious need to remotely monitor and control the instruments on my bench.  My life has never been busy enough that I would feel compelled to adjust an instrument setting remotely because I just couldn't manage to be at the bench to make the adjustment in person.  Let's explore another scenario.  Suppose I need to monitor the status of a hazardous experiment (rapid lithium battery charging perhaps) to make sure everything is okay in the lab.  Wait, why on earth would I leave a hazardous experiment running unattended?

 

So, I  really don't see the need for remote instrument control.  On the other hand I do see the benefit in having programmable instruments and the ability to generate automated test scripts.  Once upon a time I was a production technologist responsible for validating the performance of hundreds of complex PCBs coming off a production line.  The suite of tests that had to be performed to verify compliance to specifications took about 15 minutes per board and involved many settings changes on test equipment.  I spent many late nights in my youth earning overtime while grinding though repetitive test procedures.  I would have loved to have programmable test equipment back then.  I have also done large volume environmental stress testing of production units and prototypes.  Automated test scripts would have been a blessing.  But remote control?  Still can't see it as  a must have.

 

The video below illustrates the Android and iOS remote control apps in action.  They both work as advertised, and if you have a need to control your spectrum analyzer from another time zone, well, there is an app for that!

 

R&S Mobile View App on Android and iOS devices

 

A well thought out feature. . . that mysteriously only works in some cases

Because I have been frustrated in the past by clutter on instrument screens, especially screens that tend to have information rich features like annotated markers, I was delighted to see how R&S stepped up their game by making frequency markers hop over each other if they get too close on the screen (see video).  I immediately gave silent praise to the team that delivered on that innovation.  However it was not more than a day later that my praise turned to bafflement when I was working with the spectrogram feature and ended up using horizontal based time slice markers.  The nice user friendly marker hopping feature was not carried though to the time slice markers (see video).  Why not?  Different design team?  Just forgot to follow through?  Not sure, but I was disappointed nonetheless.

 

Cool and not so cool marker behavior

 

                                             

Time Marker behavior improved in Firmware Release V1.30

 

IMPORTANT UPDATE: April 14, 2018

With the release of firmware V1.30 R&S has addressed the overlapping Time Marker issue.  The Time Markers now behave just like the frequency markers.  Starting with V1.30 a time marker being moved by the user now  jumps politely over a stationary marker to avoid obstructing the users view.  See the improvement in the video clip above.  R&S were probably scheduling a new firmware release for the FPC1000 series spectrum analyzers to coincide with release of the FPC1500 model, so the reviews posted on element14 helped identify important beta testing bugs that came in just in time to get fixed and included in the V1.30  release.  Still, I am impressed that R&S responded so quickly to firmware issues identified by users.  Not all vendors are so responsive.  Thank you R&S!

 

 

So, after a first date with the FPC1000 (with the MDO4000 tagging along), I'm generally impressed.  There are always quirks in user interfaces and honestly, what strikes me as bizarre, or awkward, or just a wee bit odd, may not bother other users.  At any rate, I wanted to learn more about the detailed features of this spectrum analyzer, so I carried out a series of experiments to satiate my curiosity.  These experiments are detailed in Part 2 of this Road Test, which follows immediately:

 

 

Part 2: Comparing the R&S FPC1000 with the Tektronix MDO4000

1.  A dead simple demonstration of using a spectrum analyzer:  Looking at a 1.000 MHz sine wave.

Here I step through the process of measuring a 1 MHz sine wave on the Rohde & Schwarz FPC1000 and on the spectrum analyzer in the Tektronix MDO4000 mixed domain oscilloscope.  Although both instruments behave as spectrum analyzers, they have different internal architectures that affect the steps required to get best results.  The FPC1000 is a traditional swept spectrum analyzer with digital user interface and analysis features.  The Tektronix MDO4000 uses a wideband digitizer and a digital "spectrum engine" (explained in this video) to achieve true spectrum analyzer performance (well beyond the basic FFT feature available in many oscilloscopes).

 

Test equipment was set up to allow precise generation and measurement of a 1 MHz sine wave on either the Tektronix MDO4000 or on the R&S FPC1000.  The configuration is diagrammed below:

Equipment set up for measuring 1 MHz sine waves

 

 

The Agilent 33622A is a precision waveform generator.  It was set to produce a -10 dBm sine wave at a frequency of 1.000000000 MHz into a 50 ohm load  The instrument screen under these conditions is shown below.

Agilent 33622A generating a 1 MHz sinewave

 

To verify the frequency output of the 33622A, I connected a 53230A counter with an oven controlled oscillator.  The counter input was set to 1 M ohm input impedance to reduce loading on the waveform generator output.  The images below show the counter measurement to 7 digits on the left, and on the right, to 15 digits (yes, it can do that).

 

                 

 

So the generator is producing a sine wave accurate to with in about 130 mHz of 1 Mhz.  Not bad.

 

  1. Steps to measure a 1 MHz sine wave on the MDO4000

            Since this will be a strictly frequency domain measurement, we begin by shutting off all the analog inputs and activating the spectrum analyzer by pressing the RF button.

Because we know the signal we are measuring is a pure sine wave, we expect all the spectral energy to be found at the frequency of the sine wave (1 MHz in this example).  The table below summarizes the parametric settings that will be entered into both spectrum analyzers.

 

Center Frequency

1.00 MHz

Span

20 kHz

Start Frequency

990 kHz

Stop Frequency

1.01 MHz

Resolved Bandwidth (RBW)

100 Hz

Reference Level

0 dBm

 

The MDO4000 performs mathematical analysis on the digitized RF waveform to produce a spectrum trace.  The MDO4000 provides six “window functions” that are used to optimize the digitized signal for spectral analysis.  The available window functions are: Kaiser, Rectangular, Hamming, Hanning, Blackman-Harris, and Flat-Top.  See this blog post to get some background on window functions.  Caution: the topic of window functions can get complicated in a heartbeat.

 

The window function makes a significant difference in the resulting spectrum trace, so it is important to select an appropriate window function. For example, the screen captures below allow you to compare the result of two different windows used on the exact same 1 MHz -10 dBm sine wave.  A Flat-Top window was selected on the left and a Hamming window on the right.

         

An important take away from this comparison is to understand how instrument settings influence the way information is presented to you.  The Flat-Top window method results in the best amplitude accuracy, so I used it to complete the remainder of this comparison between the MDO4000 and the FPC1000.  The FPC1000 is a traditional swept spectrum analyzer, so there is no need to select a window method.

 

As an aside, it took about 30 minutes to warm up all the instruments to steady state conditions before taking the screen shots show here. I gauged steady state to be the point in time where the frequency counter readout showed no further drift (see an example trend chart of oscillator warm up behavior in my review of the Keysight DSOX1102G) and the MDO4000 showed a stable amplitude in agreement with the waveform generator setting. In 30 minutes of warm up the measured amplitude drifted down from -9.93 dBm to -10.0 dBm.

 

While watching the spectrum trace as the instruments warmed up I noticed an anomalous signal at about 990 kHz.  This signal shifted up and down in frequency with an overall drift to lower frequency over time.  I turned on the max hold spectrum trace to capture the behavior of this signal over time. You can see the record of this unknown signal in the screen capture below.  It appears as a rectangular block on the white max hold trace toward the left side of the image.

The markers indicate the unknown signal drifted from about 991.1 kHz to 990.6 kHz in a span of about 10 minutes.   To see if this signal was coming from the waveform generator, I disconnected the generator from the spectrum analyzer.  As you can see below, the anomalous signal remained (but it has drifted lower in frequency), so it does not seem to be coming from the generator. That means it might be coming from something inside the MDO4000.

Before moving on to the FPC1000, here are three things to remember about the MDO4000 spectrum trace of the 1 MHz sine wave:

 

  • The MDO4000 measures the signal at -10.0 dBm, in very good agreement with the -10.00 dBm setting on the Keysight 33622A.
  • The MDO4000 measures the signal frequency as 1.00000 Mhz, also in very good agreement with the 1.000000000 MHz setting on the 33622A.
  • The spectrum trace updates every 5 seconds on the MDO4000 under the conditions described in the table above.

 

B. Steps to measure a 1 MHz sine wave on the FPC1000

To obtain the screen images in this next section I used Rohde & Schwarz’s Instrument View application, which is available as a free download from the R&S website.  Additionally, I made a WiFi connection to the FPC1000 and had the Instrument View application use the WiFi connection to obtain the screen captures.  The same basic set up parameters entered into the MDO4000 were entered into the FPC1000 (table of parameters reproduced below for easy reference).

 

Center Frequency

1.00 MHz

Span

20 kHz

Start Frequency

990 kHz

Stop Frequency

1.01 MHz

Resolved Bandwidth (RBW)

100 Hz

Reference Level

0 dBm

 

The FPC1000 has a nice feature available through the Setup button that captures all pertinent details of the instrument setup in a single screen.  The setup for the tests that follow is shown in the screen capture below:

 

 

This is a great feature that helps with documenting instrument configuration for reports in professional settings and in educational settings.  I draw your attention to the sweep time as it will be relevant in the following discussion.  It is set to 116 ms.  This is the fastest sweep time I could obtain within the constraints set by the other parameters.  Why not go for the fastest possible sweep speed? Well, turns out there are good reasons to slow things down.  The screen capture below show the spectrum trace that results from a 116 ms sweep.

 

 

There are a few things wrong with this measurement.

 

      • The peak according to the M1 marker appears at 1.001117 MHz, or at 1.0011147 MHz according to the M1 frequency count (I haven’t figured out why these numbers differ). That is over 1 kHz higher than the 1.000000000 MHz set point on the Agilent 33622A waveform generator.  This offset (about 1117 Hz) is roughly 10,000 times greater than the 130 mHz offset measured on the oven controlled 53230A counter. The R&S FPC1000 is a high-quality test instrument, so my first instinct is to question the measurement set up rather than question the instrument accuracy.
      • The amplitude of the peak is showing as -16.6 dBm.  That is 6.6 dBm lower than the -10.00 dBm setting on the waveform generator that has been confirmed by the MDO4000.  Again, I question the instrument set up rather than the instrument accuracy.
      • The spectral detail is quite low considering the resolution bandwidth (RBW) is set to 100 Hz.

 

Let’s try slowing the sweep to 5 s.  That will match the update rate on the MDO4000.  The screen capture below shows the result of a 5 s sweep.

 

 

 

The slower sweep speed provides much better results. The filters have more time to react and produce a more accurate output.  Things to note in this measurement:

 

      • The frequency peak measured at M1 is 1.0000003 MHz.  That is much closer to the set point frequency on the generator and is in much better agreement with the value measured by the high precision Agilent 53230A counter.  The difference between the two instrument readings is now about 170 mHz.  Again, I do not know why the M1 marker frequency (1.000203 MHz) is so far off the expected value.
      • Measured amplitude is now -10.1 dBm.  This is 0.1 dBm off the set point value in the waveform generator and close enough to be considered reliable.
      • There is much more detail in the spectrum trace on either side of the sine wave peak. This is in keeping with what one would expect from a 100 Hz RBW.
      • There is no anomalous signal lurking around 990 kHz as noticed on the MDO4000.

 

Just for fun, I set up another test to average 10 sweeps with each sweep taking 60 seconds.  The idea was to average out most of the noise in the trace.  The result is shown below.

 

 

Sure enough, averaging the sweeps has averaged out a lot of the noise.  The peak hasn’t moved much in terms of frequency or amplitude, so no improvement there.  But, the M1 marker frequency is a bit worse at 999.983 kHz.  Not sure how to fix that.  I assume I have mis-configured some setting, but have not yet determined which one.

 

In summary

A very simple measurement of a pure sine wave at a fixed frequency and amplitude was illustrated on two spectrum analyzers.  Both instruments achieved very good results on this simple test.  It is, however, up to the instrument user to properly configure the measurement parameters to allow the instruments to deliver their maximum performance potential.  For this test, I prefer the swept spectrum result provided by the FPC1000 because it is free of artifacts caused by windowing a digitized signal.

 

2. Modulation analysis of an RF decorative lighting remote control.

 

On the right is a photograph of a simple RF remote control used to control strings of decorative LED lights in residential applications.  The remote control uses RF (hence the antenna) to control strings of LED lights.  The purpose of the On/Off button is self-evident.  The Functions button allows the user to cycle through eight lighting effects, and the Sync button forces several series connects light strings to align on the same lighting effect with unified timing.  In this section of the road test I will use the R&S FPC1000 to investigate the characteristics of the RF signal transmitted by this hand held remote and compare what the FPC1000 can do against the capabilities in the Tektronix MDO4000.  The characteristics I want to discover about the transmitted RF signal are:

 

  • Transmitted carrier frequency, or bandwidth of transmitted frequencies.
  • Modulation scheme used to encode commands (On/Off, Functions, and Sync)

 

The FPC1000 has a couple of very useful features that allow investigation of the modulation characteristics of simple RF remote control devices.  The first is the ability to turn the FPC1000 into a frequency selective oscilloscope by activating the Zero Span feature.  The second is the Marker Demodulation feature that provides the ability to demodulate the signal at the M1 marker frequency using either an AM or FM demodulation processes and have the resulting base band audio signal drive an internal speaker. 

 

The Zero Span option, as implied in the name, reduces the spectral span to 0 Hz. The important thing about this feature is that span determines the range of frequencies that are swept and displayed on the screen around a Center Frequency.  So, Zero Span means the display will show the amplitude behavior of the selected Center Frequency, and only the Center Frequency.  Essentially, Zero Span turns the FPC1000 Spectrum Analyzer into a single frequency oscilloscope.  In the next few paragraphs I'll show how the FPC1000 can be used to provide insight into the two transmitter characteristics I mentioned above. 

 

First, to determine the RF carrier frequency I made an educated guess about the approximate range of frequencies that I expected the RF remote to use.  Generally, low cost consumer grade RF remotes operate in the Industrial, Scientific, and Medical (ISM) portion of the radio frequency spectrum.  The length of the pull out antenna suggested the transmitted frequency might be around 430 MHz, which is in one of the ISM bands.  My frequency estimate is derived from two assumptions and a quick calculation.  First, it looks to me that the total length of the antenna (external plus internal parts) is about 7" (17.5 cm) and the antenna is a 1/4 wavelength monopole.   From these assumptions the rough operating frequency of the antenna can be obtained by multiplying the length of the antenna by 4 (to get a full wavelength), then divide the result into the speed of light (the approximate propagation speed of RF signals).

 

Using the assumption of a roughly 430 MHz transmitted carrier frequency, but knowing nothing about the modulation scheme, I attached a basic antenna to the FPC1000 and configured the following sweep parameters:

 

  • Range 200 MHz to 600 MHz
  • Max hold on (to capture and display any activity within the range)
  • 20 ms sweep time.  Nice and fast to make sure any short bursts are captured.
  • RBW and VBW set to 3 MHz, mostly to allow the short sweep times.

 

Amplitude accuracy is not a concern in this investigation.  I only need to determine the transmit frequency and hopefully learn something about the modulation methods used.  After pressing the ON/OFF button on the remote, the following spectrum was captured.

 

Carrier detected between 200 and 600 MHz

 

There is a single, strong response somewhere just above 400 MHz.  To find out the exact frequency of the carrier I turned on marker 1 (M1) and had the FPC1000 establish the center frequency at M1.  The result is shown below.

 

Carrier frequency measured and centered

 

The carrier appears to be operating at 433.8 MHz, which is in the ISM band, as assumed.  Now that the carrier frequency is known and has been centered, we can switch to Zero Span and see if anything can be learned about the modulation technique used to encode commands.  At this point, my educated guess is that the remote is using either On-Off-Keying (OOK) modulation, or perhaps Frequency-Shift-Keying (FSK).  Considering the remote is a low cost consumer product, and that the spectral peak looks pretty narrow and clean, I'm leaning toward OOK modulation because that is the simplest to implement at low cost.  If FSK was the modulation method, I'd expect to see a slightly broader spectral response as the carrier would be shifting between two frequencies.  However, the RBW is huge (3 MHz) and the 20 ms sweep time probably wouldn't capture both frequencies.  I also expect the command code to burst out repeatedly as long as the button is pressed.  What I don't know is the burst repetition rate or the duration of the command code frame.

 

To accommodate a range of probable command code frame duration and repetition rate, let's slow the sweep down to 100 ms.  From experience, I expect that will allow us to see at least one complete code frame, and probably more.  Remember that setting a Zero Span turns the FPC1000 into a frequency selective oscilloscope.  In this case, the FPC1000 will show us what happens at exactly 433.8 MHz over a span of 100 ms.  Now, about triggering.  We don't want to run continuous sweeps.  We are only interested in capturing what happens when a button is pressed on the remote.  For this reason I selected Max Hold on Trace 1 (to retain whatever happens on the screen) and set the sweep to trigger on Video 50%.  This means the FPC1000 will sweep when the Video filter output is at 50% of its range.

 

At this point I added a 10 dB attenuator to the antenna (see photo) because I am holding the remote fairly close to the antenna and there is a possibility of overloading the IF section of the spectrum analyzer.

 

10 dB in-line attenuator added to antenna

 

I am delighted to report that the set up worked perfectly and the FPC1000 was able to essentially decode the modulation of the 433.8 MHz signal.  See the screen capture below.

 

On-Off code frames from RF remote

 

This is nice.  I wasn't sure the FPC1000 (without the modulation analysis option) could deliver this level of demodulation detail.   A couple of transmitter characteristics can be derived from the image above.  First, it appears the remote does indeed use OOK modulation.  This is revealed by the pulse nature of the trace which shows that the remote transmits fixed amplitude 433.8 MHz bursts (On key), spaced by zero amplitude intervals (Off key).  Second, the code frames are about 25 ms long and they do indeed repeat as long as a button is held down.  Now we can zoom in on a single code frame and, by capturing the code frame for each button, determine the code pattern for each function.  Two minor irritations to report at this point.  First, the M1 marker is located at the far left edge of the screen capture marking the start of the trace.  The trigger point itself is marked by a barely visible green icon to the left of the -60.0 dB legend.  I can't seem to move the trigger point to the right to make it more visible.  Perhaps the trigger can be moved to the right, but I just haven't figured out how to do so.  If it can't be moved, that would be annoying.  Below are the code frames transmitted by each of the buttons on the remote.  Second, while working on this investigation, the WiFi connection to the FPC1000 was lost more than once.  Not sure if the issue is with the FPC1000, or my router, but everything else that is WiFi enabled doesn't lose connection, so the FPC1000 is currently suspect.

 

A single On-Off code frame

 

A single Functions code frame

 

A single Sync code frame

 

A couple more observations about the user interface came up during this investigation.  First, there are 6 markers available on the FPC1000.  The first four are accessed on one menu screen (see screen capture of the On-Off code frame), but the other two markers (5 and 6) are accessed on a separate menu list (see screen capture of Functions and Sync code frames).  Having to switch between the two menu lists is a minor inconvenience.  Second, each marker can be set to display an absolute value (time in this case, but normally frequency), or each marker can be set to display the difference between itself and marker 1.  This is great because sometimes absolute values are useful and other times delta values are useful.  However, I would like to be able to specify the two markers that are used for each delta.  For example, in the screen captures above, I used all the makers with the hope of being able to measure the duration of various pulses within the code frames.  To measure the duration of the narrow pulses following what I think might be an automatic gain control (AGC) pulse at the start of each frame I put a marker at the start and end of the first narrow pulse.  If I select delta, or difference mode for these markers, I get the delta, or difference of these markers relative to the position of marker 1.  I would like to get the delta between markers 2 and 3 instead.  Not a big deal.  Just a little arithmetic is needed, but I'd like the instrument to do the arithmetic for me.

 

From an educational perspective, I can see how the FPC1000 could be used in a lab setting to help students explore OOK modulation, just as we have done here.  I am impressed with the capabilities of the FPC1000 in this regard.

 

There is one more cool feature, mentioned earlier, that I would like to illustrate.  That feature is Marker Demodulation.  The base version of the FPC1000 (no K7 Modulation Analysis option installed) allows AM or FM demodulation of marker 1.  In the case being described here, AM demodulation of marker 1 (located at 433.8 MHz) allows you to hear the code frames as they are transmitted by the RF remote.   I found this to be an interesting new perspective on data communication and a potentially useful teaching tool.  Check out the short video below which lets you hear the On-off, Functions, and Sync commands as they are transmitted by the RF remote.

 

Marker demodulation of an OOK signal

 

 

For comparison purposes I have put together a video that runs through the same modulation analysis on the Tektronix MDO4000 series mixed domain oscilloscope.  If you wish to take 5 or so minutes to see the details, they are in the video.  For those that wish to move on, I'll summarize by saying that the MDO4000 has a few more capabilities in this department compared to the FPC1000.  The Tektronix MDO4000 can show the spectrum trace and RF versus time traces simultaneously.  The RF versus time traces include Amplitude (effectively reproduces what the FPC1000 can do), plus Frequency (shows detail about what the carrier is doing when keyed on), and Phase (shows the relative phase of the carrier when it is keyed on).  The base FPC1000 doesn't have the ability to show frequency or phase correlation to the spectrum trace. The MDO4000 screen capture below gives you an idea of the extra depth of signal insight provided by the MDO4000 in this particular application. In the photo you can see the spectrum trace on the bottom half and two RF versus time traces on the top half.  The pulse signal is the amplitude trace which shows the decoded OOK packet.  Above the decoded packet is the Frequency trace which shows a slight carrier drift of about 4 kHz (embedded in a bunch of noise) during the first carrier burst on the left.

 

The MDO4000 has a nice set of features that give extra signal insight

 

Here is the video run through of the MDO4000 performing the same OOK demodulation shown above on the FPC1000 (minus the cool audio decode feature).

 

MDO4000 series OOK demodulation Part 1 - Amplitude v Time analysis

 


MDO4000 series OOK demodulation Part 2 - Phase  v Time analysis

 

3. Modulation analysis of a 1.92 GHz DECT 6.0 Cordless Telephone.

In this section I use the FPC1000 and the MDO4000 series spectrum analyzers to explore the signals that are transmitted by a common DECT 6.0 household wireless telephone.

The phone is illustrated in the photograph on the right.  I did some preliminary web research on DECT 6.0 technology to get an idea of what I should be looking for prior to using the spectrum analyzers.  Getting some background on DECT 6.0 guided my use of the spectrum analyzers and saved some time that would have been spent going down blind alleys.  RF Wireless World has a brief but information rich tutorial on DECT 6.0.  You can find it here.

 

I will provide a written and illustrated walk through of how the two spectrum analyzers were used to explore DECT 6.0 signals as well as a video clip that allows for a real time experience of the signals and operation of the instruments.

 

From background study I determined that the Panasonic DECT 6.0 phone I was exploring operated at about 1.9GHz.  I set up the FPC1000 to sweep from 1.5 GHz to 2.5 GHz With an RBW and VBW of 3 MHz and continuous 20 ms sweeps.  The Max Hold trace was activated to capture whatever RF activity occurred within the established span.  Detector mode is set to RMS. The objective in this first step is to verify the center frequency of the DECT 6.0 spectrum.  The outcome on the FPC1000 is shown below.

 

 

 

 

The FPC1000 locates the DECT 6.0 signal

 

The FPC1000 captured the DECT 6.0 signal emanating from the phone and the marker automatically found the peak at 1.92688 GHz.  The signals on the right edge of the span are from WiFi activity generated by my laptop.  The net step is to have the FPC1000 move the center of the span to the marker position.  This is a single button operation.  The spectrum must be recaptured under the new center frequency setting.  The result is shown below.

 

 

The FPC1000 centers the span on the marker

 

Next, the span is reduced to get a closer look at detail in the DECT 6.0 signal.  When this is done, the frequency hopping nature of the DECT 6.0 signal is revealed (this is much more evident in the video clip).

 

DECT 6.0 Frequency Hopping Envelope

 

The FPC1000 can be used to squeeze a little more nuance out of the DECT 6.0 signal by activating the spectrogram feature.  This captures individual sweeps and presents them below the spectrum display as color coded slices.  In the image below, the FPC1000 is in in continuous free run mode which causes it to sweep as fast and as often as it can.  The spectrogram builds up a history of the sweeps with the most recent sweep at the top of the spectrogram.  The color palette can be selected from a variety of choices and the range of amplitudes displayed can be adjusted as well.  In the image below the darker red dashes indicate lower amplitude signals and the brighter pink/purple dashes are higher amplitude signals.

 

The FPC1000 reveals frequency hopping nature of DECT 6.0 in spectrogram mode

 

The vertical frequency markers can be used to get an idea of the bandwidth used in a single time slice.  M1 and D2 above show the bandwidth of the slice is about 507 kHz.  The sharp vertical edges on the slice are a little suspicious to me.  Not sure if we are seeing an artifact of the measurement setup here.  The 507 kHz bandwidth suggests there is something going on inside each time slice.  There is more than one frequency present.  The T1 and T2 horizontal markers can be used to measure absolute event times within the spectrogram or relative time differences between sweeps.

 

It appears from the image above that the DECT 6.0 signal is hopping from frequency to frequency within the band at random.  To investigate this observation further, I used an external pulse generator to feed the rear panel BNC input and configured the FPC1000 to use the rear BNC signal as a trigger to initiate a sweep.  Because the fastest available sweep time is 20 ms there is no point in attempting to trigger sweeps faster than every 20 ms (corresponds to 50 Hz).  I experimented with various triggers occurring at rates below 50 Hz and discovered patterns emerged at several key frequencies including 18 Hz (55.5 ms), 12 Hz (83.3 ms), 8 Hz (125 ms), 6 Hz (166 ms), 4 Hz (250 ms), and 2 Hz (500 ms).  There are obviously several multiples in that list, so some of the patterns are related to each other.  The pattern that emerged at 12 Hz is shown below.

 

DECT 6.0 spectrogram triggered at 12 Hz rate (20 ms sweep)

 

Notice that this particular frequency hop appears to occur at at fairly regular interval of about 250 ms (as revealed by the T1 and T2 markers in relative mode).  This is the limit of detail that I could extract using the FPC1000 (without the K7 Modulation Analysis option).  A lot of signal detail has been extracted using the FPC1000, but to dive deeper into the individual time division/frequency division bursts I had to move to the spectrum analyzer in the MDO4000.

 

For those that wish to view a video demonstration of the FPC1000 being used to investigate DECT 6.0 signals, the clip below runs through the same steps described in the text and screen captures above.

 

FPC1000 DECT 6.0 signal investigation

 

The video clip of the MDO4000 DECT 6.0 analysis (inserted below) gets into more detail, but the screen capture below shows the result of the deeper signal analysis provided in the MDO4000 spectrum analyzer.  In the image you can see the actual data structure within one up-slot of a code frame in the top portion of the trace.  This string of bits is generated by FSK modulation that has been revealed by the MDO4000's Frequency vs Time trace feature.  The various bit fields can be correlated back to the framing diagram found in the tutorial mentioned above.  It appears the MDO4000 has captured a single 416.7 microsecond slot within a 10 ms frame, beginning with 16 preamble bits.

MDO4000 decodes bits within a slot of DECT 6.0 code frame

 

MDO4000 series DECT 6.0 signal investigation

 

I would like to emphasize that both spectrum analyzers were able to provide meaningful insight in to the nature of the RF signals transmitted by the DECT 6.0 cordless phone.  These are insights that would be very difficult, or impossible, to extract using a time domain oscilloscope.  It is also important to note that if you need to examine the detailed structure of modern digital data communications signals you will need a spectrum analyzer with modulation analysis capabilities that go well beyond those found in entry level analyzers.

 

4. Third Order Intercept (TOI) measurements

I have to say, working on this section of the review reminded me why I am drawn to digital microcontroller design where most everything can be described using ones and zeroes and processes are defined by the well behaved rules of Boolean logic.  Nonlinear analog circuits and RF signals are fascinating and oh so very important in modern electronics, but working with them can lead you into rabbit holes where mysterious and magical phenomena govern system behavior.  Generally, I find it helpful to track down some reference material to help navigate concepts found in the RF world.  The following articles provided helpful background on the concepts of inter-modulation distortion and third order intercepts.  I recommend you read at least the first one from R&S.

 

 

All of the experiments that follow examine how non-linear electronic systems affect a signal that is composed of two equal amplitude sine waves at different frequencies, f1 and f2.   First, I provide some exploration of the types of signals that are generated by two closely separated sine waves using the Tektronix MDO4000 series Mixed Domain Oscilloscope with a 3 GHz spectrum analyzer.  Following this overview I discuss how the R&S FPC1000 automated Third Order Intercept (TOI) measurement feature performed for me.

 

To generate a two frequency signal I used the dual channel combining features of the Keysight 33622A waveform generator.  This generator allows the outputs from its two channels to be electronically combined into a single output.  Signal amplitude, DC offset, and frequency can be coupled so that changes to the combined output channel affect both channels.  In addition a fixed frequency offset can be programmed between the channels, which is a great feature for conducting intermodulation experiments.  In the first video clip below I demonstrate how third order products are produced as signal amplitude is increased and how these products produce signals very near the original f1 and f2 signals (at 2f2 - f1 and at 2f1 - f2) and other undesired products well away from the original signals.

 

There is no external Device Under Test (DUT) in the demonstrations that follow.  The Keysight 33622A is fed directly into the MDO4000, and in subsequent demonstrations, directly into the FPC1000.  So where are the non-linearities coming from that create the intermodulation products?  This question is discussed in some detail in the R&S application note referenced above.  Essentially the non-linearities are coming from the generator, the spectrum analyzer, or both.  What is important is that you get to see that they are real, they are problematic, and that the FPC1000 can make automated third order intercept measurements for you. It took a lot of experimenting to find a sweet spot on the MDO4000 where everything responded as dictated by the nonlinear mathematics that describe intermodulation products.  What I looked for was 3rd order intermodulation products appearing at the expected frequencies (that part was easy) and changes in the amplitude of the products that agreed with theory (that was more difficult).  According to the math, when signals are subjected to non-linear processes, a 1 dB change in f1 and f2 amplitude should produce a 3 dB change in the adjacent (in-band) third order intermodulation products.

 

The 33622A was set up to combine a 10.000 MHz sine wave with an 11.000 MHz sine wave.  Initially the amplitude was set to -40 dBm.  At this amplitude, no non-linear effects are observed on the MDO4000.  With f1 = 10.000 MHz and f2 = 11.000 MHz, the math tells us that once amplitudes of f1 and f2 are sufficient to cause non-linear responses, we should see 3rd order intermodulation products at the following frequencies:

 

Fundamentals and

3rd order intermodulation products

Frequency (MHz)

f110.000
f211.000
2f1 - f29.000
2f2 - f112.000
3f130.000
2f1 + f231.000
2f2 + f132.000
f2 - f11.000
3f233.000

 

As a quick aside, I also provide a demonstration of the effects of RBW here.  In the screen capture below, the RBW is set to 1.75 MHz.  The two sine waves are only 1 MHz apart.  Notice that a 1.75 MHz RBW is too wide to resolve signals that are separated by only 1 MHz.  The result is an envelope that encompasses both signals.

Reducing RBW to 500 kHz begins to distinguish two separate signals.

 

 

Reducing RBW to 10.0 kHz distinctly reveals two signals separated by 1.0 MHz.  Reducing RBW further doesn't provide much better resolution at this span and it slows down the processing speed.

 

Notice that at this amplitude (-40 dBm), all that can be seen are the two combined sine waves.  There are no intermodulation products detected.  This suggests that at this amplitude all of the electronics in the signal path are operating in a linear fashion.  When the signal level is increased to -35 dBm, some 2nd order intermodulation products appear around 21.00 MHz, as well as the beginning of third order products at 30 and 33 MHz.  Still nothing significant near f1 and f2.

 

By the time the amplitude reaches -32 dBm, the full assortment of 3rd order intermodulation products from the table above are appearing in the spectrum (along with a few other products).  Note that every time I change the signal amplitude I activate an Auto Level function that optimizes the attenuation/gain settings for the MDO4000.  This is why the amplitude scale changes in the screen captures.

 

At this point it is beneficial to reduce the span so we can concentrate on the troublesome intermodulation products near f1 and f2.  These are the products that are most likely to fall in-band and cause distortion of a transmitted or received signal.  In the two captures below the span has been reduced to 10 MHz, centered at 10.5 MHz.  Four markers have been activated to show f1, f2, and the close proximity 3rd order intermodulation products at 9.00 MHz and 12.00 MHz.  f1 and f2 are at approximately -13 dBm in the capture on the left.  Note the amplitudes of the intermodulation products: One at -76.5 dBm (9.00 MHz) and the other at -79.3 dBm (12.00 MHz).  According to the math, a 1 dB increase in f1 and f2 (from -13 dBm to -12 dBm) should produce a 3 dB increase at 9 and 12 MHz.  In the capture taken at -12 dBm, the 3rd order intermodulation products are measured at -73.6 dBm (9.00 MHz) and -75.7 dBm (12.00 MHz).  The differences are 2.9 dB and 3.6 dB.

 

                   

The signals are behaving according to theory at this sweet spot around -12 dBm.  Moving just a few dB away from this sweet spot allows magical and mysterious factors to take over, causing the measured spectrum to abandon allegiance to the math presented in the reference material.  I'm confident there are rational explanations for the odd behaviors I witnessed outside the sweet spot, but I would rather spend some time with a Cypress PSoC or an Arduino right now, just to help me get back to my happy place.

 

TOI on the R&S FPC1000

The R&S FPC1000 evaluated in this review came equipped with a license for the FPC-K55 Advanced Measurement option.  This option provides automated measurements for Occupied Bandwidth, AM Modulation Depth, Harmonic Distortion, TDMA Power, Channel Power, and Third Order Intermodulation.  My experience using the last of these capabilities is described next.  As in the previous discussion, a Keysight 33622A waveform generator was used to combine equal amplitude 10.000 MHz and 11.000 MHz sine waves which were then fed directly into the RF input on the R&S FPC1000.  The preamplifier and electronic attenuator were turned off to remove any non-linear effects they might contribute to the signal path.

 

The centre frequency was set to 10.5 MHz, half way between the two fundamental signal frequencies.  Span was set to 6 MHz to include the upper and lower 3rd order intermodulation products that would be created with a 1 MHz separation between the fundamentals.  RBW and VBW were set to 1 kHz and sweep speed was slowed down until the measured amplitudes of the fundamentals agreed with the amplitudes set on the 33622A waveform generator.  Amplitude agreement occurred at around 20 second sweep time with the Trace Detector set to Max Peak.

 

The screen capture below shows the TOI automated measurement taken under the conditions described above.  You will see that the amplitudes of M1 and M2 are equal and measured at -25.0 dBm in agreement with the value set on the 33622A waveform generator.  However, a problem arises in the measured frequencies of at M1 and M2.  As you can see in the screen capture, M1 is reported to be 10.002538 MHz which is fully 2.538 KHz higher than the 10.000000 MHz set point on the 33622A (confirmed to be within 1.5 Hz of 10.000000 MHz by an Agilent 53230A counter).  Furthermore, the reported frequency of M2 is 10.997462 MHz, which is 2.538 kHz below the 11.000000 MHz set on the waveform generator (also verified to be within 1.5 Hz on the counter).

 

 

TOI measurement at -25 dBm (M3 and M4 misplaced)

 

These discrepancies in measured frequencies are problematic enough, but they have a serious impact on the accuracy of the automated TOI measurement.  From my observations it would appear that the FPC1000 in TOI mode seeks out the two highest amplitude signals within the span of the spectrum and places M1 and M2 at these peaks.  It then appears to calculate the expected location of the 3rd order intermodulation products at 2f1 - f2 and at 2f2 - f1.  Here is the problem:  If the measured frequencies for f1 anf f2 are inaccurate (and my experiments suggest they are), then the derived locations for the interfering side frequencies will also be inaccurate.  The consequence is that markers M3 and M4 get misplaced.  If you do the math using the displayed marker frequencies for M1 and M2, you will see that the FPC1000 did the math correctly, but with faulty numbers.  The result is that M3 gets dropped 7.614 kHz higher than it should be and M4 gets dropped 7.614 kHz lower than it should be.  Look carefully at the M3 and M4 marker locations in the screen capture above.  They are close to, but not on the signal peaks.  The amplitudes measured at the misplaced markers are then fed into the TOI calculation (a straight line interpolation) to produce the TOI value shown just above the spectrum on the left (15.9 dBm in this case).

 

As I did for the MDO4000 experiments, I changed the output amplitude of the waveform generator by 1 dB to see if the intermodulation products changed by the expected 3 dB.  The screen capture below shows what happened.

 

TOI measurement at -24 dBm (M3 and M4 misplaced)

 

The amplitude the fundamentals (f1 and f2) changed by the expected 1 dBm, but the amplitude of the 3rd order product at M3 changed by 0.3 dBm (-107 dBm to -106.7 dBm) and the amplitude of the 3rd order product at M4 changed by 3.1 dBm.  However, the numbers are meaningless because the markers are not on the 3rd order signals, they are beside them, in the noise.  This means the calculated TOI value is also meaningless.

 

I nudged markers M3 and M4 to the signal peaks and reran the experiment at -25.0 dBm and -24.0 dBm.  The results are shown below.

 

                             

TOI measurement at -25 dBm with M3 and M4 adjusted                                                                                TOI measurement at -24 dBm with M3 and M4 adjusted

 

You can see that the M3 and M4 markers have been moved to the signal peaks in both cases.  The amplitude measurements are more reasonable and the differences in intermodulation product amplitudes when the fundamentals are changed by 1 dBm is better (M3 changed by 3.8 dBm and M4 changed by 2 dBm).

 

The automated TOI measurement is a nice feature, but there may be a need to augment the measurement algorithm to search near the calculated IM product frequencies to find and measure the actual peak signals, rather than assume the peaks will be where the measured values of f1 and f2 suggest they should be. Of course, if the measured values of f1 and f2 were more accurate, there would be less of a need to search for the peaks.  I tried several variations of f1 and f2 frequencies with no better results.  There seems to be an issue right at 10.000 MHz, so I tried using an external reference signal fed into the rear panel BNC, but to no avail.

 

Addendum to the Road Test:  Using h-field probes with the FPC1000

 

In this addendum two element14 members ( and Instructorman) team up to use the FPC1000 in a real-world study of a Watkins-Johnson receiver.  James provided the receiver and Instructorman provided the FPC1000 spectrum analyzer.  James is an RF and mm wave IC design engineer in California and Instructorman works in applied research at a polytechnic institute in Canada. We worked on this addendum via e-mail and Skype and had one opportunity to get together in a lab setting to run a series of probe tests on the physical receiver. Our plan was to learn about the Watkins-Johnson receiver by examining its h-field emissions.  To sense h-field emissions from various circuits in the receiver we experimented with two h-field probes.  The first probe we tried is a commercially available product from Bee Hive Electronics, the second probe was designed and fabricated by element14 member @shabaz.   Attaching h-field probes to the FPC1000 allowed us to gather interesting spectrum information from emissions produced by circuitry in the Watkins-Johnson receiver.  With thoughtful analysis based on knowledge of superheterodyne principles, explained below, we gleaned useful insights into the operation and layout of circuitry in the Watkins-Johnson receiver.  A description of the receiver investigation, and a couple of additional side experiments on other RF sources, follows.

Experiments on the Watkins-Johnson receiver using the BeeHive Electronics 100A h-field probe

 

Mark was kind enough to let me spend some time playing with the FPC1000, and I had some HF/VHF/UHF receivers lying around that I thought would make for perfect subjects for testing this spectrum analyzer with near-field RF probes. We used a Beehive Electronics 100A H-field probe intended for electromagnetic compatibility (EMC) testing – in other words, sniffing around for sources of radiation that may lead to interference. When paired with a sensitive spectrum analyzer (especially one with a preamplifier, like the FPC1000), these nifty probes can also be useful for debugging RF electronics.  

 

The receivers at hand are some older (90s vintage) Watkins Johnson receivers that typically found their home in three letter government agencies. 

 

Pictured above is a model 8712A-3 with a tunable frequency range from 5kHz to 30MHz. It uses a triple-conversion superheterodyne architecture as shown in the block diagram below. Without getting too bogged down in the details, we can see that the input signal moves from the antenna input on the left through a series of three mixer stages.  Each mixer stage decreases the frequency of the input signal to an Intermediate Frequency (IF) before passing the signal to the next mixer stage. There are three independent local oscillator (LO) signals to produce three intermediate frequencies (IFs).  Notice that the 1st LO is tuneable over a 30 Mhz band from 40.455 MHz to 70.455 Mhz.  Also notice that the 2nd and 3rd local oscillators are fixed frequency.

 

For those unfamiliar with RF electronics, a mixer is a circuit element that multiplies one signal with another. This has a unique implication – it is an inherently nonlinear element. Suppose you take two sinuosoids -- sin(f1t) and sin(f2t) -- as inputs to a mixer; at the output you get the desired sine waves with the frequencies f2+f1 (useful for “mixing-up”), and f2-f1 (useful for “mixing down”). However, because you are dealing with a nonlinear element, you also get additional intermodulation products (eg.  2f1-f2, 2f1-f2, …) as well as harmonics (2f1, 2f2, 3f1,...) as shown below. The bandpass filters ahead of each mixer stage are used to attenuate intermodulation and harmonic components.  In a superheterodyne design like this, only the down mixed carrier (f2-f1) and associated modulation content is of interest.  So each bandpass filter following each mixer in the block diagram is designed to attenuate everything outside a band around f2-f1.

 

Suppose the receiver operator was interested in listening to a modulated signal at a carrier frequency of 10 MHz.  To down convert a 10 MHz carrier and its modulation signal, a control signal would be sent to the 1st LO causing it to oscillate at 50.455 MHz (f2).  A 50.455 MHz sine wave mixed with a 10 MHz modulated signal (f1) will produce a mixed down carrier at 50.455 MHz – 10 MHz = 40.455 MHz, which is the center of the first IF.  Knowing this will help us understand some of the signals we came across while looking at this receiver. 

 

The final IF of the receiver at 25kHz is digitized and sent to the digital signal processor for digital filtering and demodulation. 

 

Treating this as a real-world debugging exercise (in part because this was my first-time powering on this receiver), I thought it would be a good idea use the FPC1000 to start looking for signs of life in the RF-front end. With the RF shielding cans removed, the top of the RF board provided easy access to use the H-field probe, as shown below. 

 

 

 

It is worth noting that the frequency response of the 100A probe limits practical usefulness to the MHz range, meaning we should be able to spot everything up to the second IF. It would actually be a nice feature if the FPC1000 could take frequency compensation data for near field probes for more accurate field measurements (assuming a specific orientation for H-field probes).

 

 

 

With no input signal connected, we began speculatively tracing through the signal path. The first thing we stumbled upon was a strong signal at ~52MHz, as shown.

 

 

 

Knowing the general architecture of the receiver quickly led us to believe that this was the first LO section. If this is indeed the 1st LO, the strongest frequency we will see in this section will be the tuneable local oscillator signal.  When powered up, these receivers go through a quick self test, then they restore the 1st LO frequency in stored in memory.  The brief video below illustrates this power up behavior.

 

 

Because there is no RF signal being fed into the input, all we see in terms of h-field emissions from the 1st LO is the local oscillator signal being restored from memory following power up.

 

Recalling the “down-mixing” product (fLO-fRF= fIF), we can deduce that the receiver was last tuned to receive ~11.493 MHz. How?  Because to produce a down converted IF signal at 40.455 MHz, a LO signal at 51.948 MHz would have to mix with a modulated signal at 11.493 MHz (51.948 MHz – 11.493 MHz = 40.455 MHz).  Without having the control software at hand we could not adjust the first LO frequency (note the absence of a tuning knob on the front panel), so this ended up being useful information for when we fed in some test signals later on. 

 

At this point, I was using a fairly large sweep and a fine resolution bandwidth, which of course limited how fast I could physically move the probe around and expect to see something on the screen. Using a second trace and having it set to max-hold was helpful to avoid missing signals, and it also helped to play with the orientation of the probe to get the maximum response. 

 

The next large signal we discovered in another section of circuitry was a sinusoid at 40 MHz – the LO of the second conversion stage. 

 

 

Moving to the other end of this can, we see a large number of harmonics. Using the measurement markers, it is clear that they are spaced at 10 MHz intervals – hinting that we are seeing the frequency quadrupler that is responsible for taking the 10 MHz reference and turning it into the 40 MHz LO for the second conversion stage. Marker 6 also shows some evidence of the first LO, which makes sense as the PLL circuit is immediately adjacent. 

 

 

One difficulty I ran into while using the FPC1000 occurred when I attempted to have the FPC1000 automatically place multiple markers on peaks. From reading the literature beforehand, I knew it was possible, but I couldn’t figure out how to do it for the life of me, even after playing around with it all evening. 

 

(Note:  After we documented the experiments described here and after encountering the problem with the markers described by James, I reran some of the same experiments and had no trouble getting the markers to auto-locate the peaks. Under normal operating conditions, to activate all 6 markers and have them identify the 6 highest peaks, the user presses “Mkr”, then presses the down arrow at the bottom the menu selections, then presses “All Markers On”. The FPC1000 can, however, enter a quirky operating mode where pressing “All Markers On” places all markers at the same frequency, usually near the center of the displayed span, regardless of the number or location of peaks displayed on the screen.  This is what James and I observed when we ran the experiments described here.  This quirky operating mode does not happen often and I have not recorded a sequence of button presses that causes the FP1000 to enter this mode, but I have discovered that pressing Preset will take the FPC1000 back to normal operation.)

 

At this point, the second IF and following signals are below the usable frequency for our nearfield probe. Luckily, the second IF is available directly as an output on the rear panel, meaning we could connect it directly to the FPC1000 using coax. Using a signal generator, we fed in a AM signal with a carrier frequency of ~11MHz -- corresponding to the receiver’s assumed tuning based on our 1st LO measurements – and a modulation frequency of 5kHz. Setting the FPC1000 in center-span mode corresponding to the 455kHz IF output, we observed the expected carrier tone and corresponding sidebands. 

 

 

 

 

This rear-panel IF output is actually meant to be used as a signal monitor to see what is on the airwaves when tuning the receiver. With that in mind, we put the FPC1000 into spectrogram mode and simulated adjacent stations broadcasting 10 kHz apart.

 

Here, I ran into another quirk with the FPC1000 that Mark had previously mentioned. The accuracy of the signal was weirdly impacted by the scan speed – to the point where we began questioning the function of both the receiver and the signal generator we were using.

 

To test the FPC1000 and H-field probe in a true EMI scenario, we went back to the 40MHz LO section and re-measured the emitted signal amplitude with the RF shielding in place.  

 

 

The corresponding maximum signal dropped by over 50 dB from the previously recorded signal strength. This is really where the pre-amp of the FPC1000 shines. Some could argue that some of what we have done until now could have been accomplished by a fairly inexpensive USB spectrum analyzer (meant for software defined radio). However, the preamplifier as well as the very linear front-end provided by the FPC1000 provide the user with an astonishing dynamic range for an instrument of its price. 

 

 

 

Experiments using Shabaz's L0 h-field probe

 

I’d like to thank James for walking us through the Watkins Johnson 8712A-3 using the Bee Hive Electronics 100A h-field probe.  In this part I reproduce some of James’s measurements using a custom h-field probe designed and fabricated by element14 community member Shabaz.

 

Shabaz designed and fabricated the probes shown below and graciously shipped the photographed samples to me for testing on the FPC1000. 

 

It is no surprise that these probes are well made, just like every other contribution Shabaz has made to the element14 community.  They are composed of two layers of PCB material sandwiched together with a small exposed loop structure at one end and a female SMA connector at the other end. The loop end of the probe is sensitive to near-by magnetic fields.  These probes can be used to sense fugitive electromagnetic fields that are invisibly leaking from electronic circuits.  For comparison purposes, the photograph below shows the BeeHive Electronics 100A magnetic field probe next to Shabaz’s L0 (long probe model 0) and S0 (short probe model 0) probes.

 

 

Using a second Watkins-Johnson WJ-8712 that James left with me, I reproduced the first local oscillator probe measurement using Shabaz’s L0 probe as shown below.
When this particular receiver was taken out of service and put on the surplus market it appears the last frequency that was being observed was 50.425491 MHz - 40.455 MHz or 9.970 MHz, plus or minus the accuracy of the spectrum analyzer measurement and the LO accuracy.  Or stated another way, what input frequency, when mixed with a local oscillator frequency of 50.425 MHz, will produce a difference frequency of 40.455 MHz that will pass through to the 2nd IF mixer? 50.425 MHz - 9.970 MHz = 40.455 MHz.

 

 

I also reproduced the 2nd LO emission measurements using Shabaz's L0 probe as illustrated below.  I'd like to point out a display issue in the FPC1000 screen capture below.  All 6 available markers have been activated and they automatically positioned themselves over the 6 highest signal peaks, which is very helpful.  However, please locate the marker for M5.  It is the red maker which is half obscured on the right edge of the spectrum by the menu overlay for the soft buttons. Now, the marker is still visible and the peak is also just barely visible, but wouldn't it be better if the display size was adjusted so that the whole maker and peak were clearly visible?  That would be my preference.

 

 

To illustrate the importance of fully seating RF shields I placed the RF shield on top of the grounding wall, but did not seat it. Then I took two measurements in the same corner, one with Shabaz's L0 probe, and the other with the BeeHive Electronics 100A probe.  This test also allowed me to roughly compare the performance of the BeeHive probe against Shabaz's probe. See the results below.

 

 

In the photo you can see that the RF shield is just sitting on top of the grounding wall.  A properly seated shield can be seen to the left of the 2nd LO circuit.  Compared to no RF shield at all the emissions are attenuated, but the 40 MHz signal is still pouring out at -60.9 dBm.  Now lets see what Shabaz's probe measures under the same conditions.

 

 

As you can see in the photo I bumped the shield and moved it a bit, so the measurement won't be a perfect match, but the numbers are remarkably close.  Shabaz's probe measures the 40 MHz emission at -60.8 dBm.  The greatest discrepancy between the two probes is at 90 MHz (M3) where the difference is 3 dB.  With the RF shield properly seated I remeasured emissions with both probes. The screen captures below show the emissions measured with the BeeHive probe on the left and with Shabaz's probe on the right.  Only the emissions at 40 MHz are detectable at about -93 dBm.

 

 

That concludes the experiments with the Watkins Johnson receiver.  For fun I used Shabaz's probe to explore a few RF emission sources on my lab bench. On my laptop I discovered fugitive emissions from a USB memory stick.

 

 

There were also broadband intermittent emissions near the Ethernet connector and on the touchpad.  Finally, I placed Shabaz's L0 probe between my Samsung Galaxy S6 and a Qi wireless charging pad to see if I could capture some of the interplay between the charger and the phone.  This is what I discovered.

 

 

There is lot going on in there, but unpacking that will wait for another time.

Thank you all for reading and do let us know if this approach was useful.

 

Kind regards,

James and Mark

Anonymous
  • Hi Jon,

     

    Indeed the solenoid would solve that problem..  as you say even a single turn could be sufficient by having two probes close. The solenoid technique could work into the MHz range with less turns I reckon.. and also work into the hundreds of MHz if not higher, if designed as a transformer I think.. e.g. with a ferrite toroid and just a single turn. only thing with that is it is hard to get the PCB through : ( unless it was designed as two halves that are soldered together, and could be hard to accurately calculate flux density, because ferrite is a more imprecise material and varies with temperature etc. But it would work I reckon, i.e. just changing from air core to ferrite.

  • Can the box have a picture of Tesla, done as marquetry, on the lid with the mysterious words 'Level 8' just below and, perhaps, some decorative lightning flashes in the corners?

     

    Right. I was being far too optimistic and a bit stupid (let's hope no-one is reading this). As the frequency went up, the solenoid coil's impedance would increase steadily and we'd no longer have a simple way to know what the current was because we'd no longer be working into the 50 ohms that the analyser expects. I think you've possibly answered why they go for a transmission line. The big problem with a trace, that I was trying to get away from, is that you don't capture the total flux, so need some arbitrary correction, but it looks like that's what you'll need to try, unless anyone else has any ideas (or you drop the idea of calibration altogether - though it might be good to be doing at least something, even if roughly, as a quality-control measure on the board construction).

     

    Thanks for humouring me. I'll let you know if I come up with any other ideas.

  • Hi Jon,

     

    I think you're not wrong, basically you're suggesting a reference solenoid, using equations to work out the magnetic flux density (since we know what the current will be, number of turns, etc). That would totally work, however only at the low frequencies (it would present high impedance to RF). However the principle you mention could also be used for a straight transmission line, it's just that we wouldn't know the flux density without equations that I am unfamiliar with : ( since the flux density will now vary.

    (there is the equation for a straight wire, perhaps that is sufficient.. or maybe it is easy to do with simulation tools ). However, there are methods to measure flux density at the low frequency end fairly cheaply, e.g. using the hall sensors. It's a nice challenge..

  • According to LibreOffice I wrote 1057 words, so almost exactly right.

  • How does this sound as an idea? A solenoid coil with the same cross-section as the aperture of your probe. If you present the probe aperture to the end of the coil, all the flux goes through the probe. Irrespective of where it goes though, you can calculate an average field strength from the total and the area (I think the probe respose would be the same as for an equivalent even field). The coil flux I think you can get from the current and permittivity of free space, number of turns, and dimensions of things. The current you can get from the tracking generator voltage if you put a 50 ohm resistor in circuit, and you're there. Not sure how far up it would go frequency-wise, but then it's probably something like comparable with the probe. Someone put me out of my misery if I'm hopelessly wrong with this (I've resolved to dust off my books and see if I can relearn some basic, long-forgotten electricity and magnetism theory).

     

    Other idea I had was two probes back to back (one driving, one receiving) but then you probably wouldn't be able to easily calculate the flux generated.

  • Hi Jon,

     

    The test jig was just a one-off, but maybe a 3D printed version could be possible. I didn't have that, so used what I had. I suppose any trick to keep the probe in a fixed position would work. I love the idea of a CNC-machined case with probes. I'll really consider turning it into a product.

    I think you're right, it won't be an even field. I'm not sure how they did an absolute calibration, I bet they calculated or simulated what the field strength ought to be at different locations (i.e. based on input signal strength, and using magnetic field formulas), and then put their probe into place (i.e. in a defined position) and observed the signal level and plotted that versus frequency.

    I'm wondering if  maybe 90% of the time the use-cases for the probe are either to explore a circuit, or to just have relative levels, for making circuit adjustments and seeing the relative level rise or decrease.

    I actually could try to measure the field at different heights, by applying DC current or at a very low frequency to the trace, and using a hall probe. I've got maybe 2-3 ways to try to measure that, although they all require some setup. I've still got some crazy-sensitive flux-gate magnetometer TI dev-board ( Building a Fluxgate Magnetometer based Current Probe  ) , but also this cool thing: SS94A2 HONEYWELL, Hall Effect Sensor, Linear, 1 mA

    it should I think easily be able to detect just a few mA of current. I've not tried it yet, I ordered it recently for something unrelated, but was then on a work trip, and also working on the Bluetooth Design Challenge : ( so it got pushed to the back for a while..

  • You didn't think it would be interesting? Of course it isn't - get back to flashing the led on that Arduino immediately.

     

    That's a nice test jig - it certainly looks the part. Are you going to sell that separately or will it be a complete kit with the two probes? Perhaps you could have a CNC-machined hardwood case (with brass fittings) to store them in.

     

    If they suggest a microstrip then there are reasons for it - they'll know what they're doing (more than I do, anyway). Perhaps it's something to do with it not radiating - the magnetic field will be predictable even though it isn't even.

     

    I can see how you'd do a relative calibration with it - comparing one probe to another - but how would you do an absolute calibration with it? The Beehive curves say they're at 1 microtesla field strength. That implies an even field everywhere within the aperture of the probe's loop but the magnetic field around a track won't give that, will it?

  • Wow.  Thank you Jon for that very enlightening and enjoyable response!

    I actually learned a few things, and they are likely to be useful things, at least peripherally, in my work going forward.

  • I'm not an expert and I'm pretty much self-taught (the EMC stuff came along long after I left university and we all had to puzzle it out for ourselves from whatever sources we could find - no internet back then) so my knowledge is patchy and I probably still habour odd notions that aren't even correct. If it helps, this is how I understand it, but don't take me as your primary reference on the matter, question everything I write and, if it's for anything that really matters, check elsewhere.

     

    Short answer: a sixth of a wavelength [I looked it up in Henry Ott's book just to make sure I got it right].

     

    Long-winded answer follows here:

     

    The term 'near field' comes from rf engineers doing antenna design. It's the area close-in to an aerial where the magnetic and electric storage fields that develop around the aerial's conducting parts are still significant and will therefore affect measurements.

     

    Those fields decline as you move away from the antenna. The way that happens depends on the physical, 3-D, form of the antenna, however one thing you can be sure of, if the structure radiates, is that at some point the ratio between the voltage field and the magnetic field will equal the impedance of free space (approx 377 ohms). With our engineer's model of things, that's the point at which the propagating wave (the radio wave that is independent of the transmitter and will fly through space until it gets absorbed by something) is considered to come into existence. (That's a sort of fiction. The photons that make up the propagating radio wave don't suddenly come into existence in free space, they emanate from electrons accelerating in the antenna conductors, but the way the electrons behave depends on the physical structure and the fields derive from what the electrons are doing, so it all comes out in the wash and our engineering models are equivalent to the physics models and 'work', even if our ideas are a bit back-to-front at times. If a physicist wants to do electronics, generally they'll take off their physics hat and put on an electrical engineer's hat just like us - but, unlike some of us, they're very aware of which are good models to do physics with and which aren't.)

     

    How do those fields decline in space? That's an interesting question. It's surprising how many engineers will automatically say it's an inverse square law for both fields (I think that's by analogy with the propagating wave - a situation where it's actually due simply to geometry, because of the way the wavefront expands). Both the electric field and the magnetic field can be somewhere between an inverse square and an inverse cube and which they tend towards depends on the form of the antenna. A loop antenna is good at generating a magnetic field so, for that field to get down to the right ratio with the much lower electric field, it needs to decline much faster and you'll see something like an inverse-cube on the magnetic field. At the other extreme, a dipole is good at generating an electric field, so in that case it will be the electric field that diminishes quickly.

     

    Other antenna structures will result in situations somewhere in between.

     

    So where does that point come, where the ratio is the impedance of free space? That might seem like an almost impossible question to answer - modelling the fields and solving everything - but it actually turns out to be very simple. Because of the way the maths works (don't ask me to prove it though), it's about a sixth of the wavelength (the wavelength divided by 2 x pi). However, it's not an abrupt transition with storage fields one side and a propagating wave the other, instead there's a transition as the propagating wave becomes more and more dominant and the storage fields diminish further, so for measurement purposes (of the far-field radio wave) you'd do it further out at several times that range.

     

    At 30MHz, which is the lowest frequency that radiated emissions have normally been measured for emc testing, the wavelength is 10m. The transition from near to far then comes at 1.6m. That means measurements done at 10m are safely in the far field. Measurements at 3m (which is an option in the standards, by applying a suitable correction to make them equivalent to the standard at 10m) are ok, but perhaps slightly iffy down at the 30MHz end of the range. Measurement at 1m is going to include some near-field stuff at the lower frequencies, which is why it isn't an option in the standards, but it can be useful for testing because of the increased signal levels, particularly higher up where the transition is much closer to the antenna.

     

    Conversely, if you want to be sure of measuring storage fields and not the radiated wave, you need to be well inside that sixth of a wavelength distance, but what you're measuring is very difficult to quantify because of the rapid way the storage fields fall off and you need to take a guess at what sort of antenna you have in order to know which falls off more rapidly.

     

    Finally, all that I've written above is for an antenna. If we don't have an antenna that can radiate, then we probably shouldn't used the term 'near field' and instead should probably refer to storage fields (but that horse bolted long ago). Storage fields can exist without radiating. In terms of the above engineering model, that happens when the structure doesn't give fields that can achieve that impedance of free space, and that particularly tends to happen when the structure is small compared to the wavelength of the signal being created by the circuit - once you're down to a tenth of the wavelength in any dimension, you've probably got a poor radiator. [I say particularly there, because for all I know it may be possible to engineer larger structures that don't radiate effectively either, and I say probably because you can see with a ceramic chip antenna a way to make a physically small structure appear to be electrically much larger.]