Wednesday, August 31, 2011

Radar Analysis with the Tektronix MDO4104-6

This week, Tektronix introduced the "scope revolution", the new Mixed Domain Oscilloscope.  There is a ton of material already available describing the architecture of the oscilloscope and how it can be used for some common applications, such as Zigbee, power supply design, PLL settling, etc.  I won't repeat today what you'll find on the Tektronix site.  I want to blog on some less common applications.  In this case, I am using the MDO4104-6 to analyze a wideband chirped radar.
What are some common measurements I need to make in radar design, what tools are used today, and why would I use an MDO to analyze my radar?

At its simplest, a radar is a single tone that is sent out as an RF burst.  A classic problem in radar design is the tradeoff between range and resolution.  In short, a signal with a wide pulse width contains enough RF energy that it is able to travel a long distance to a target.  The problem is that if there are multiple targets, their individual reflections will be obscured by the long pulse.  Two airplanes will look like one.  In other words, long range, poor resolution. 

If you narrow the pulse width, the resolution will improve, but the range will get shorter.  In other words, you can discern two airplanes next to each other, but you might miss an airplane too far away.

The solution is pulse compression.  Make the pulse wide, but do something unique in the pulse that enables you to discern multiple reflections.  One of the most common techniques is linear frequency modulation.  If the pulse is centered at 4GHz, instead of playing out a pure 4 GHz, start the signal at 3.97GHz and sweep to 4.03GHz.  Many radar systems use narrow chirps, on the order of 10MHz to 50MHz wide.  But to get better spatial resolution, you chirp wider.  Modern radars can have chirp bandwidths of 500MHz and beyond!

How do you test a radar?  Often individual components (transmit/receive modules, antenna components, etc) can be tested as part of a golden radar, installed with known good components and evaluated.  However, using a device to test itself is not often ideal, so off-the-shelf tools have radar personalities for doing performance evaluation.

For radars with bandwidths under 100MHz, a spectrum analyzer can be used to measure the characteristics.    Tektronix, Rohde, and Agilent all make digitizing spectrum analyzers that can downconvert an RF pulse to baseband and store it as digital I and Q.  From that data, measurements can be made.  The Tektronix RSA6000 and RSA5000 series can perform 27 different scalar and vector pulse measurements.  Scalar means amplitude-vs-time based measurements, such as the pulse risetime, width, duty cycle, ripple, droop, etc.  Vector means measurements like frequency deviation, frequency error, impulse response, phase deviation, and pulse-to-pulse phase error.

For radars with wider bandwidths, a real-time digital oscilloscope is typically used to acquire the signal.  There are oscilloscopes with enough sample rate to digitize the RF carrier from vendors like Tektronix, Agilent, and LeCroy.  You just buy enough bandwidth to digitize your signal.   With oscilloscopes going in real-time bandwidth beyond 30GHz today, just about any signal can be acquired and analyzed.  Combined with specialized software packages (such as Tektronix SignalVu), the same 27 vector and scalar measurements can be made.

The world seems happy with golden radars, spectrum analyzers, and oscilloscopes, so where does the Mixed Domain Oscilloscope fit for radar diagnostics?
  1. Portability and size.  At just 5" deep and 11lbs, it is far easier to carry than larger instruments.  The aforementioned spectrum analyzers and oscilloscopes can weight over 50 lbs and can be the size of a small piece of luggage.
  2. Cost is considerably less.  Often a lab can only afford one high-end spectrum analyzer or high-bandwidth scope.  Each of these may cost over $100k.  But at less than $30k, every bench can have an MDO.
  3. Most importantly, the MDO gives users the unique ability to not only analyze their radar signal, but also the other numerous digital gates and timing signals present in a radar system.

The MDO has 4 analog channels with bandwidths up to 1GHz, 16 digital channels, and 1 RF input.  The RF input can be tuned up to 6GHz with a minimum real-time bandwidth of 1GHz.  It can also decode up to 4 buses, such as MIL-STD-1553 and RS-232.

Bear in mind that the MDO does not replace an oscilloscope with a dedicated radar analysis package like SignalVu.  Examples of this type of oscilloscope include the Tektronix DPO/MSO5000, 7000, and 70000.  SignalVu can do significantly more measurements specifically related to pulsed radar, including pulse-to-pulse frequency difference, pulse-to-pulse phase difference, frequency error over the pulse, impulse response, and can analyze up to 10,000 consecutive pulses.  This powerful, detailed analysis is not replicated on the MDO, so both tools really have a different role to play in an RF tool bench.  Below is an example from a scope running SignalVu, analyzing consecutive pulse parameters, showing a histogram of those parameters across up to 10,000 pulses, and an impulse-response spectrum from the chirp.

SignalVu on a High Bandwidth Scope - Tremendous Single Channel Analysis of a Radar Signal

The MDO does a subset of the analysis you can do with a scope running SignalVu, but an important subset that often covers the basic need, while adding the functionality of multiple channel time correlation.

Tektronix MDO4104-6
In the following display, I have quite a few traces.  Starting at the top, I have the amplitude vs time of the radar pulse centered at 4GHz with a 500MHz bandwidth.  This would be similar to "zero span" on a conventional spectrum analyzer, although no spectrum analyzer could do a 500MHz zero span display.  It is labeled "A" for "Amplitude vs Time".  The next trace down is labeled "f" for "Frequency Vs Time".  It shows the frequency deviation during the pulse.  The trace is blank in spots because I have squelched the frequency display when the pulse is not on.

Below that signal is the trigger for the transmit pulse (channel 1), a digital gate that is valid during the pulse (channel 2), a receive trigger for the return pulse (channel 3), and a system clock for all of my digital gates (channel 4).  In the interest of saving space, I am not showing them, but there are 16 more digital channels I can assign to gates in my system.

I have triggered on the RF burst (see the RF trigger at -9dBm in the lower right), and made a series of automatic measurements.  I have measured the risetime (4.957us), pulse width (31.35us), duty cycle (18.8%), and PRF (5.99kHz) of the radar.  I have also measured the delay from the TX trigger rising edge to the start of the envelope (15.15us), and the delay from the falling edge of the RF pulse to the rising edge of the receive gate (7.773us).  There is also a Peak-Peak measurement on the Frequency Vs Time, showing a 504MHz frequency deviation on the chirp.  Finally, I have measured my system clock which is running at 120MHz.

Also, note that while I triggered on the rising edge of the RF burst, the MDO can also trigger on other pulse characteristics as well.  It can be setup to trigger on narrow pulses, wide pulses, runt pulses (pulses that are not full amplitude), and even missing pulses.  You could even setup to trigger when the RF pulse is present but the On-Gate was not (using a logic trigger).  No regular oscilloscope or spectrum analyzer can trigger on all of these conditions.

Below, I am showing the same data but I have split the screen to also show the RF spectrum of a pulse.  You can see the orange bar under the second pulse indicates where I am doing my FFT.  To get a good looking FFT of a chirped radar, you need to set the window function to Rectangular as Kaiser will just look like a series of spikes in the frequency domain.  With this display up, I am using manual markers to measure occupied bandwidth.  From 4.235 to 3.735, I have my 500MHz.  I can also move the spectrum analysis window to show the pulse off time, seeing what the spectrum looks like when the pulse is not active.

The ability to capture high frequency RF and slower control signals is unique to an MDO.  Suppose I was analyzing a narrowband radar at 4GHz.  A spectrum analyzer could do a tremendous amount of analysis, but could not acquire other related signals like the digital gates.  A real-time oscilloscope could acquire my RF and additional signals, but would be limited in capture depth.  If my real-time oscilloscope digitizes at 50GS/s and comes standard with 10M points of memory, it can only capture 200us of data.  Even at max memory, it can still only capture 5ms of RF data.  If I lower the sample rate, I can capture more time, but I won't be sampling high enough to get my RF data.  How do I see a 4GHz signal and a 100MHz clock?

While the MDO4104-6 time correlates RF and analog data, the digitizers are independent.  So while the RF section captures a pulse at 4GHz (storing up to 75ms of RF Data), the analog section can run at 500MS/s and with just 20M of memory store about the same record.

Mixed Domain Oscilloscopes are a pretty new and exciting category.  They certainly do not replace spectrum analyzers or conventional digital oscilloscopes for radar analysis.  In my mind, the MDO is a tool that overlaps some key functions at a smaller size/lower price point, while simultaneously adding key functionality not available on any other platform.


  1. Great explanation! Thanks

  2. At its simplest, a radar is a single tone that is sent out as an RF burst. A classic problem in radar design is the tradeoff between range and resolution.
    electrical services perth