[For a copy of this article in PDF format, which displays figures and equations, click here. Requires Adobe Acrobat Reader, free download.]

Aerospace/Defense systems are often developed using design software and test instrumentation. As test instruments have become more computer-based, these two areas in the product development process can both be enhanced by a tighter integration.

The application that will be analyzed is an X-band chirp radar system. The system will initially include only simulation models (such as virtual hardware) as might be done early in any leading edge design process.

Data from these analyses can be compared with the design specifications and the system can be modified or optimized to meet the desired performance goals. Data from these simulation models can be collected, displayed and reused by connecting the simulation environment with the test instrumentation, such as a vector spectrum analyzer (VSA).

This collected data can then be passed onto other test equipment such as an electronic signal generator (ESG) or be used by the VSA itself as source data. This test equipment source can then act as a signal source, either within the simulation software or as virtual signal source to check for the proper operation of various parts of the system as they mature toward physical realization.

For example, the simulated chirp signal can be passed into a source and then this “real” signal can be applied to various transmitter components to see what unexpected effects may occur. It could be seen if there might be problems with spurious or adjacent channel products in the transmitter.

As hardware becomes available, it could be incorporated into the simulation to further see the effects. Initially, the bandpass filters in the transmitter might be modeled using a simulation model. As the manufacturer starts to make the physical components available, the measured S-parameters, from a network analyzer, for the filter could be embedded into the system in place of the model. Effects that may not be obvious, such as in-band and out-of-band voltage standing wave ratio (VSWR) effects of the filter on a preceding mixer, could be analyzed. In addition, the group delay response of the filter and its effect with the “real signals,” mentioned previously, passing through it and the rest of this hybrid transmitter could be measured.

The measured, real data, results from the intermediate frequency (IF) or baseband of the receiver collected by the VSA could also be sent to the baseband designer. They could apply this time-domain data to the correlator, which compresses the chirped signal, to help insure that their module can properly handle the desired and undesired signals it may have to accommodate. This could also help to mature the design specifications.

For example, maybe the low noise amplifier (LNA) noise figure doesn't need to be quite so low, or the phase-noise requirements of the local oscillator (LO) synthesizer needs to be more stringent. The earlier one can find out about any particular problem in the system, the easier it will be to fix it.

The initial system uses simulation models for the system components such as in the chirp source, transmitter, radar link model, receiver and correlator (compressor). The radar frequency, in this example, is 10 GHz and the IF frequency is 2 GHz.

Each of the models blocks shown in Figure 1 incorporate a complete subcircuit. For example, the chirp generator, as shown in Figure 2, will generate the complex chirped pulse and a trigger to activate the receiver protection network during the pulse transmission.

The time-domain real part of the complex 2 GHz signal envelope at the IF output of the transmitter and receiver is shown in Figure 3. Likewise, the other blocks in the chirp radar system consist of various components, such as filters, mixers, and amplifiers. These can be modeled at the behavioral level, as a block or at the component level with lumped and distributed components and non-linear devices.

Finally, the baseband signal is passed into a correlator that compares a copy of what is coming out of the receiver with a copy of the reference pulses coming from the chirp source. It then computes a correlation function at each offset. The result, as shown in Figure 4, is a response that shows the time offset of the received signal.

The next step is to connect a VSA at the transmitter and receiver IF to collect and display the results and compare them with the simulator results as shown in Figure 5.

A VSA allows the user to not only collect signal magnitude information, as in a conventional spectrum analyzer, but also phase information. Thus, it can collect and display in-phase and quadrature (I/Q) data for a signal, which is critical for most modern communications and radar systems. The recorded signal can also pass to other instruments and directly interface with design software. The VSA used here has no front panel display of its own. Its user interface is run from software on a PC, which makes its front panel very flexible and allows easy verification of the system simulator results that are running on the same PC.

Connecting the vector spectrum analyzer in the virtual system

A virtual spectrum analyzer is placed at the transmitter IF in the design's schematic. The simulation response from the spectrum analyzer matches very closely to the simulator response shown in Figure 6.

Using the spectrum analyzer results as a new source

Now that the results from the transmitter IF have been recorded, it can be reused as a virtual source for several useful applications.

Replacing the chirp source model with the spectrum analyzer results

The chirp generator simulation model can now be replaced by the VSA using either the data recorded previously or a real-time measurement of actual hardware as it becomes available. This source could then be applied to the rest of the system as shown in Figure 7.

The source can provide a single pulse period or have the data continuously repeat. The transmitter designers, those doing the link modeling and designing the receiver and its protective circuitry, can all utilize this source as a representation of what their part of the system might face in the real world. As the system design matures, this signal model will continue to improve, thus, different departments can get access to the latest data to see how any changes might affect their particular system component.

Utilizing the receiver IF data to test for proper correlator operation

In the same way, the data at the receiver IF port can be supplied to the baseband designers building the correlator. This can help insure its proper operation, not only with ideal signal inputs but also with all the noise and impairments that a real system might be presented with. A good example of this is to determine as early as possible in the game what the minimum detectable signal level is necessary at the receiver front end. The receiver IF signal could be recorded and supplied to the designer of the signal correlator (compressor) with more realistic signals. The output of the correlator, along with other data, could be provided to the various baseband circuits for further evaluation of signal parameters, such as to calculate target range.

The effect of multiple signals at the receiver input

Now that the system is running, other factors come into play, such as what the system's response would be to having multiple microwave signals of various types applied to its input. In the example shown in Figure 8, the same signal is modeled as shown before but also has a second target return from a object further away, but with a higher radar cross section (RCS). In addition to that, there is a strong continuous wave (CW) carrier not far off the radar frequency and a broadband jamming signal.

How do these various signals affect the operation of the radar? Can the system resolve the two targets? How will the strong interference signals effect the receiver operation? Other things that could be done would be to collect data from various sources such as other collocated radars, other high power transmitters, and jamming sources to analyze their effects on your systems performance.

In Figure 9 the radar systems response to these various microwave signals can be seen. The plot on the left is the signal magnitude at the receiver IF. The return from target 1 is totally hidden by the jamming signal. The target 2 return can be seen because it's between the jamming bursts. After pulse compression, the two returns can be easily identified.

Either of the two interference sources shown in Figure 8 or any other signal could be replaced with measured signals from the VSA or an electronic signal generator. Signal data for these sources can be uploaded directly from the simulation software.

For broader band signals, a system such as the one shown in Figure 10 could be used to generate and collect signals.

The signal generator can generate a signal with a bandwidth of up to 100 MHz with its internal arbitrary waveform generator. By using externally generated I/Q data, the bandwidth can be increased to 160 MHz. The RF carrier frequency can be up to 20 GHz, depending on the type of source. The spectrum analyzer can measure RF frequencies directly up to 26.5 GHz. By using the IF directly, instead of passing it through the preselector, it can output measured signals with bandwidths beyond 250 MHz. This signal can be passed directly into one of the Agilent Technologies Inc. (developed by Infiniium) series oscilloscopes. The results from this can be read in and manipulated directly from your design software.

Initially the system design may consist of mostly behavioral models. That is, models that are described by their behavior, not their specific component parts. For example the filters at the mixer input and output are initially modeled as behavioral models as in Figure 11.

The next step might be to design or buy a filter that meets the requirements. The designed filter may initially consist of ideal component, possibly Ls and Cs. The design could then be matured by including realistic component quality factors (Q), coupling effects and parasitics. If this is to be part of an integrated circuit, the foundry models will be the key to getting accurate results.

Using measured network analyzer data in the simulation

Whether the filter is being built or bought from a vendor, the next essential step is to measure the filter to insure that it meets the design requirements. The component being tested, a 400 MHz lowpass filter in this case, can be measured on any of a number of network analyzers. The results can be evaluated based upon the stand-alone response versus the design specifications. These results can also be read directly into the schematic design from the network analyzer for use as a model.

The design goals were to have less than 1 dB of insertion loss out to 400 MHz and at least 60 dB of rejection beyond 800 MHz. Other important parameters for a radar system are group delay and group delay deviation, the second derivative of phase with respect to frequency, as shown in Figure 12.

The group delay for the measured filter is actually better than the behavioral filter, but the group delay deviation is somewhat worse though still within the design goals. Other parameters that are of interest are the passband insertion loss (<1 dB to 400 MHz) and return loss (>10 dB to 400 MHz) and shown in Figure 13.

Simulating the entire receiver

The measurements in Figure 11, Figure 12 and Figure 13 were of the lowpass filter by itself. The next step would be to incorporate the measured S-parameters into the simulation of the entire receiver. The goal of the receiver is to have 0 dBm ± 1dB output power with -50 dBm at the input port, and have all out-of-band spurious products down at least 60 dB.

The receiver system is shown in Figure 14.

The second mixer is a radio frequency integrated circuit (RFIC) Gilbert Cell mixer modeled at the transistor level. With the behavioral filter in the system output, the spur product table appears as shown in Figure 15.

The output power (0.092 dBm) meets the specification (0 dBm ±1 dB). The point of concern might be the 0 × 0 × 2 spur product that is only down 60.782 dBc. The specification is that all the spurs must be down at least 60 dB. It might be necessary to go back and redesign the system to give a comfortable margin for this specification. This would delay the product development and add expense.

In this case, we can implement the actual filter that will be placed at the mixer output and see what the spurious response will be. This is very critical because this filter will have a significant effect on the operation of the mixer and thus the system performance.

In Figure 15 the product of most concern is the 0 × 0 × 2 product, which is the second harmonic of the second LO. Mix(1) is the RF signal products, Mix(2) is the 1^st LO product and Mix(3) is the 2^nd LO product. There is a high risk that if the system is implemented with this narrow margin it may fail to qualify.

Since the real filter measured response is available, it can be quickly placed in our system to see how it will perform before investigating a costly redesign of the receiver. The results with the measured data filter are shown in Figure 16.

It can be clearly seen that the margin with this filter is much better than with the behavioral filter. In addition, the next highest spur level, at 1.7 GHz, has also improved by about 5 dB. This means that a system redesign is unnecessary.

To design complex systems, it is beneficial to combine all available resources to a design that will reduce the chances of serious problems late in the product cycle. Combining measured data with simulation models can result in much more mature products earlier in the design cycle.

This hybrid approach allows the sharing of simulation information throughout the typically widely dispersed design teams on a large project.

The transmitter designer can get a real signal to work to verify that the system will meet the specifications instead of having to work exclusively with theoretical or ideal signals until the later stages of the project.

The receiver designer can also have an image of what a real signal will probably look like along with the possibility of being able to quickly test for the effects of new threats that may come up. If changes are necessary anywhere in the system, the new data can be quickly sent to those areas that may be affected so the changes can be evaluated for their impact.

As design software and test equipment become more closely linked in the future the opportunities to combine simulation and measurement into a seamless process will only increase.