1. Home
  2. Featured Articles
  3. Radar/EW Signal Analysis Using Ultra High Sample Rate Oscilloscopes

Radar/EW Signal Analysis Using Ultra High Sample Rate Oscilloscopes


by Brad Frieden & Philip Gresock, Aerospace and Defense group, Keysight Technologies Inc.


Modern electronic warfare (EW) trends are demanding an increase in capability to keep pace with the rapid evolvement of radar systems and techniques. Electronic protect (EP) measures such as frequency agility, complex modulation on pulse, and multi-mode capability require advanced capture and analysis tools to validate operation and fidelity.  Typical measurement capability results in the inevitable tradeoff between frequency range, bandwidth, and capture length.  However, recent advancements in digital signal processing implemented in ASICs, as well as ultra high sample rate oscilloscopes, enable useful capture length across multi-gigahertz wide bandwidths at frequencies from RF to millimeterWave.  This article will explore how these enabling technologies provide meaningful characterization of radar and threat emulation systems.

The purpose of this article is to discuss why ultra high sample rate real-time oscilloscopes are useful for modern RF analysis, and also to dive into the techniques used to efficiently measure signals over time.


Modern day oscilloscopes are a far cry from their original implementation years ago.  No longer are a few hundred-megahertz sample rates enough to measure modern day signals of interest.  Recent advancements in test and measurement capability and related architecture have allowed oscilloscopes to span the use case from time domain measurements, predominately for digital applications, to the radio frequency (RF) domain and RF analysis. 

Figure 1: Seven 20 GHz RF pulses captured over 6.25 ms with 128 Gsa/s sample rate

The Keysight UXR oscilloscope can simultaneously sample up to 256 GSa/s on 4 channels.  This allows for a functional RF frequency range/bandwidth from DC to 110 GHz.  This exceptional capability combined with the performance of a 10-bit ADC allows the UXR to be considered for RF measurements across the spectrum. 

However, with such high sample rates, one must pay attention to the memory usage since most oscilloscopes have limited onboard memory.  Techniques like segmented memory and digital down conversion play an important role in memory efficiency to insure a meaningful measurement.

Memory Efficiency Techniques

Segmented Memory

For the last several years, if not decades, real-time oscilloscopes have made repetitive measurements with the use of an external trigger signal to define an event of interest.  Triggering is an integral capability of an oscilloscope to make meaningful visual measurements to allow the user to see an event of interest based on the trigger criteria.  For example, one common trigger criteria is voltage level crossings.  

However, this visualization technique alone does not provide enough memory efficiency for such applications as radar pulse analysis, although the use of an external trigger for a segmented capture does.  Segmented capture allows the user to define the length of a capture interval in terms of samples or seconds along with the number of segments to capture.  This enables the user to easily capture short events of interest of similar length and therefore only uses the precious onboard capture memory for meaningful information that meets the user’s criteria.

Figure 2: Fourteen 20 GHz RF pulses captured over 12.5 ms with 64 Gsa/s sample rate
Figure 3: 6,712 20 GHz RF pulses captured over 6.7s of target activity time with segmented capture and 64 Gsa/s sample rate

Digital Down Conversion

The ultra high sampling rate of modern real-time oscilloscopes is simultaneously a great strength and weakness.  Aside from capturing only meaningful segments vs. superfluous samples, exceptionally high sample rates consume the onboard memory blazingly fast.  Sub-sampling the data can effectively reduce the functional data rate, while maintaining the frequency range and bandwidth needed to make a meaningful measurement.

Digital down conversion (DDC) is a relatively common technique used in digitizers that has now made its way into modern digital oscilloscopes.  Within the FPGAs and ASICs of the UXR oscilloscope, users can selectively define a center frequency and bandwidth for a measurement, instead of having to directly sample at 2.5 times the highest frequency component of a signal of interest to satisfy Nyquist.

DDC allows the functional reduction of sample rate from 10s-100s of GSa/s to 100s -1000s of MSa/s.  This reduction of sample rate enables the user to capture a longer signal of interest.

The combination of segmented capture and digital down conversion allows users to go from capturing a signal over mere milliseconds of time, in hope that there is enough information to determine the correct operation of a device, to instead acquiring 10s-100s of seconds of data to truly show operation of a system under test.

Figure 4: 73264 20 GHz RF pulses captured over 70 seconds of target activity time with segmented capture and 800 Msa/s I/Q sample rate

Measuring Challenging Signals

Modern Radar/EW signals are more diverse than signals that resulted from techniques of the past.  Frequency agility, low probability of intercept (LPI), wider bandwidth, and staggered pulse repetition intervals (PRIs) present measurement challenges because one must now capture longer segments of a radar mode or Electronic Attack (EA) SUT technique to ensure that they are operating correctly.  With signals from a technique lasting 30 seconds or more, combined with wider bandwidths, the challenge to effectively capture, analyze, and report results is even more critical.

Theory in Practice

Using the Keysight UXR0334A Oscilloscope and the 89600 PathWave Vector Signal Analysis Software, we can visualize the efficiency gains from segmented memory and digital down conversion.  In the case provided below, we will capture and analyze a pulse radar signal of interest with the following parameters:  

  • Center frequency (CF)= 20 GHz
  • Pulse repetition interval (PRI) = 1ms
  • Pulse width (PW) = 1us
  • Duty cycle = 0.1%
  • N7660C Multiple Emitter Scenario Generation (MESG) application scenario

Measurement Setup 1:  Direct Full Sample Rate

We first start the test by running the scope at a full 128 GSa/s sample rate.  This sample rate is more than adequate to satisfy Nyquist given the center frequency of 20 GHz (i.e. 20e9 *2 = 40 e9 samples/second).  At this sample rate, we can utilize the 2 GSa of onboard memory to capture 7 pulses of our signal or a 6.25ms acquisition as shown in Figure 1.

 This length of capture certainly provides some insight into our example signal; however, it would be difficult to provide insights into long term pulse stability or proper operation over the course of part or all of a scenario.

Measurement Setup 2:  Reduce Direct Sample Rate

The user can manually reduce the sample rate of the UXR to something that still satisfies Nyquist based on the known center frequency of 20 GHz.  This requires a minimum of at least 40 GSa/s sample rate; however, in practice, a ratio of 2.5 is used. Ideally, a 50 GSa/s sample rate could be selected for the UXR; however, the native sampling rate of the instrument operates on a base 2, so we must select 64 GSa/s.

64 GSa/s should logically allow twice the capture length as compared to our 128 GSa/s use case.  Such a capture can be seen in Figure 2.

We can now see that we have captured 12.5 ms of time for our acquisition and 13 pulses from our signal of interest.  This is a 100% increase in capture time, however, it’s still relatively short with respect to modern radars and EW signals and does not provide a view of the signal over the course of the scenario.

Measurement Setup 3:  Introducing Segmented Capture

With the user setup sample rate of 64 GSa/s, we can now attempt to become more efficient with the oscilloscope memory by utilizing the segmented capture feature of the UXR.  We will now define an adequate segment length based on the prior knowledge of the expected signal.  In this case, we choose a 1.2us segment length since our pulses of interest are 1us long.  Given the known pulse repetition interval (PRI) of 1ms and the pulse width of 1us, by capturing a segment length of 1.2us and not saving samples during the off time of the pulsed RF signal, we can increase our memory efficiency by over 600% from our reduced sample rate example.

The segmented capture capability given this use case allows for an acquisition of up to 6712 segments/pulses, which results in RF pulse capture over >6.7s of target activity time. The 6.7s capture, as compared to the paltry 6.25ms that we originally started out with, is over a 1000% increase in capture length, as shown in Figure 3.

Measurement Setup 4: Segmented Capture + Digital Down Conversion  

Finally, we can incorporate digital down conversion combined with segmented capture to allow for the longest acquisition.  As described above, instead of requiring a sample rate based on the center frequency of the signal, we only need to account for the capture bandwidth desired; in this case, about 400 MHz to capture the signal modulation on the carrier. This functionally allows us to set a sample rate of 800 MSa/s on the UXR.  These settings enable us to capture over 70,000 segments/pulses, for a total logical capture length of over 70 seconds.  Now, we can view RF pulses over the course of a typical scenario.  Let’s now consider a practical example.

Electronic Attack (EA) Range Gate Pull Off (RGPO) Jamming Example


One fundamental technique of jamming against a radar to prevent it from successfully tracking a target is called range gate pull off (RGPO).  Consider a situation where an electronic attack (EA) system aboard an aircraft needs to effectively jam a surface-to-air missile radar with the goal of preventing the missile from launching or being guided to the aircraft.  In this example, we’ll say the distance from the missile launch point to the aircraft is a 10 km distance.  A reasonable assumption would be that it would take approximately 20 seconds for the missile to reach that distance.  That will end up being an important period to capture pulses.  

Figure 5: Radar echo signal and larger RGPO jammer signal as seen on UXR oscilloscope

With the RGPO technique, the aircraft EA system “listens” to the missile radar, and then exploits the range gates in the radar tracking system by creating “false” radar return signals.  The signals that are received at the radar receiver are larger than the real radar RF pulse reflections, possibly 10 to 20 dB larger, and slowly move in time away from the location of the real reflections, as shown in Figure 5.  This causes the range gates in the radar receiver to get “pulled off” from tracking the real reflections and being “fooled” to track the false echo jamming signals.

There is a cycle of:

  • Range gate pull off, radar breaks track
  • Radar back to search mode
  • Radar reacquires
  • Radar tracks again and then the cycle repeats

 RGPO Example

A reasonable scenario example would include 1 usec wide radar pulses and a jammer that creates 1 usec wide false echo RF pulses that pull off from real radar echo signals by around 10 usec in time over a 10 second period, as shown in Figure 6.  To increase range resolution without having to increase transmitted power in a radar system, we can also employ linear FM chirping combined with a matched compression filter in the receiver.  With increased modulation bandwidth, one realizes better range resolution.  In this example, we’ll say that the chirp is 400 MHz wide, the carrier frequency is 30 GHz, and the pulse repetition interval (PRI) is 1 msec.

RF Pulse Capture Requirements and Use of an Oscilloscope with Segmented Memory

What could be very helpful to the EA designer when verifying the EA system would be to capture radar and jammer signals as the RGPO technique is applied, but in a lab environment with coaxial signal connections and emulated radar signals, as seen by the EA receiver and as seen at the radar receiver, plus the injected EA false signals.  Can the UXR oscilloscope be used for such a test?

To answer that question, several important factors for such a measurement must be considered.  They include the rearm time of the oscilloscope between triggers, the scope sample rate/bandwidth, the scope sensitivity, and the scope memory depth.  Let’s first consider the oscilloscope sample rate/bandwidth and the memory depth.  

Figure 6: View of range gate pull off profile of 10 usec pull off over 10 seconds and then vanishing

The UXR oscilloscope series offers a 4-channel, 33 GHz bandwidth model with 128 Gsa/s sample rate that is more than sufficient to capture 30 GHz signals. However, if triggering once and making one continuous capture, the 2G sample point deep memory would only allow for a 15.625 ms record length:

Record length = (# samples / sample rate) = 2E9 samples / 128E9 samples/s = 15.625 ms

Only 15 radar echo and 15 jammer pulses would be captured and this would not give a view of the 20 second missile flight and RGPO engagement.  Alternatively, oscilloscope segmented memory capture could be used, where a trigger condition, such as an RF pulse rising envelope, causes a segment of memory to be filled.  However, the rearm time must be considered.

If an oscilloscope triggers on an event, once the related signal is digitized and samples are placed into memory, there is a dead time before the trigger can “rearm” to be ready to trigger on an event again.  The UXR oscilloscope has a trigger rearm time of 5 usec if the high bandwidth trigger is enabled, and 3.5 usec if it is disabled.  Those rearm dead times can be important when trying to capture RF pulses such as those in this 10-second-long RGPO engagement where jammer pulses pull away up to 10 usec from radar reflection pulses.  In this example, where radar reflection and jammer pulses are 1 usec wide each, one might think that 1.2 usec wide memory segments would be ideal for the capture.  But consider a point in the engagement where the jammer pulses are a couple usec away from the radar reflection pulses.  The scope would trigger on a radar reflection pulse, put it into the 1.2 usec wide memory segment, have a 5 usec rearm dead time, and miss the 2 usec away jammer pulse altogether.  A solution is to pick 11 usec wide segments so that with each trigger, both radar reflection and jammer pulses are captured into each segment at all points of the RGPO engagement.

Each 11 usec segment requires around 1.4M samples:

# samples = (record length x sample rate) = 11 usec x 128 Gsa/s = 1.408e6 samples

The total number of 11 usec wide segments that can be captured is around 1400:

# segments = (total number of samples / samples per segment) = (2e9 / 1.408e6) = 1420 segments

This would correspond to 1.4 seconds of missile flying time:

Portion of missile flying time = (number of segments x PRI) = 1420 x 1 msec = 1.4 seconds

This is still not enough missile flying time to see the RGPO engagement of the radar with the jammer.  

Variable Width Segmented Capture and the Use of Real-Time Digital Down Conversion

A better approach is to use variable width segmented capture, together with real-time digital down conversion, where an IF trigger senses when a signal is present and only stores samples into segments when the signal is present.  This eliminates the “dead time” in between pulses and maximizes the use of memory.

Figure 7: Capture of radar echo and jammer RF pulses over 50 seconds of target activity time

However, if real-time digital down conversion (RTDDC) was used together with segmented memory in the UXR, then instead of using up the 2G point scope memory for 128 Gsa/s capture, a much slower I and Q sample rate can be used that will still support a bandwidth span wide enough to capture the 400 MHz wide LFM chirp modulation on the RF pulses.  The UXR uses discrete spans, and to measure a 400 MHz wide signal, a 640 MHz wide span is available that uses 800 Msa/sec I and Q data.

That sample rate would be:

I and Q sample rate = modulation bandwidth x 1.25 = 640 MHz x 1.25 = 800 Msa/s

The combined I and Q sample rate of 1600 Msa/s is 80 times less than the original 128 Gsa/s sample rate for the oscilloscope front end.  This allows for a much more efficient use of the UXR oscilloscope memory for the RGPO capture.  Now the total amount of missile flying time where RF pulses can be captured has increased significantly.  The VSA software, with the BHQ Radar pulse option, can capture 83,000 pulses in “Record” mode which equates to around 50 seconds of scenario time, more than enough to analyze the 20 second missile flying time as well as multiple 10-second-long cycles of the RGPO engagement, including interaction prior to a launch.  This capture and analysis is shown in Figure 7.

 The upper left trace shows a Log Magnitude view of the pulse amplitudes over the 50 second period.  The top center trace shows a linear magnitude view of six pulses including the analysis region for each pulse shown in green.  The right top trace shows the 400 MHz wide linear FM chirp on each RF pulse.  The bottom left trace is a spectral view of the 400 MHz wide pulse modulation.  The bottom center trace is a plot of the pulse repetition interval (PRI) time, zoomed in at a 1 msec value, so the RGPO process can be seen between the radar pulses and the jammer pulses as the jammer pulses pull away over multiple cycles.  Finally, the bottom right table shows the pulse parameter measurements for each of the 83k captured pulses.  These values can be further analyzed or even exported to be played out through a signal source such as the N5193A/94A.

Receiver Sensitivity

A final factor in making successful measurements is whether the oscilloscope has enough receiver sensitivity.  The UXR oscilloscope has a very low noise front end (~ -164 dBm/Hz noise density at the most sensitive setting of 32 mV full scale), but consider that at the full 33 GHz bandwidth, the actual noise level seen would be 266 uV (rms) which corresponds to -58.5 dBm.  So, weak jammer signals less than around -60 dBm would not be seen.  However, with RTDDC, reducing the measurement span reduces the noise by:

Noise reduction = -10 log (RTDDC span / Full scope BW) = -10 log (640 MHz / 33 GHz) = 17.1 dB 

With RTDDC, the weak jammer signals can still be measured to as small as -58.5 dBm – 17.1 dB = -75.6 dBm.


Modern day real-time oscilloscopes like the Keysight UXR allow for exceptional performance with the direct sampling of signals up to 110 GHz.  The combination of this raw performance combined with segmented capture and digital down conversion allow for greater insights into signals of interest and their long-term trends.

About the Authors

Brad Frieden

Brad Frieden is a factory application engineer working for the Aerospace and Defense group of Keysight Technologies and is focused on the calibration and use of UXG signal generators in multi-port electronic warfare threat emulation systems.  He has been with HP/Agilent/Keysight for 37 years with experience in business development, product marketing, and product planning.  Brad began his career as a radar system engineer at General Dynamics Fort Worth in 1982.  Brad received his BSEE from Texas Tech University in 1982 and his MSEE from The University of Texas at Austin in 1991.  

Phil Gresock

Philip Gresock is a solution planner for the Aerospace and Defense group of Keysight Technologies, focusing on high performance Radar and EW signal analysis platforms and measurement techniques. He has been with Agilent/Keysight for 11 years in a variety of roles, including application engineering, marketing, and technical sales. He received his BSEE from Michigan State University in 2007 and his MSEE from Lawrence Technological University in 2009.