New instruments excel at constructing super-complicated waveforms that once took a lot of time to create.

CHRISTOPHER SKACH, SAHAND NOORIZADEH
TEKTRONIX

IT’S hard to beat arbitrary waveform generators (AWG) for convenience and flexibility when it comes to generating signals for test applications. If a waveform can be defined or captured, chances are good that an AWG can generate it. What’s more, the last few years have seen significant advances in digital-to-analog converter (DAC) technology, making it possible to directly generate high-bandwidth, high-frequency signals with high fidelity.

As DAC technology and AWG instruments have become more capable, so too have the techniques for designing the various waveform types necessary for radar and electronic warfare (EW) or for a variety of applications in verification, characterization, and stress/margin testing. The traditional approach has involved defining waveforms in software programs such as Excel or Matlab or even in simple .txt text files and importing them into the AWG. This approach offers flexibility but also requires in-depth knowledge about the waveforms in question, and it can be tedious and time-consuming. To improve efficiency and productivity, AWG vendors are now adopting a modular software-based plug-in architecture that allows for point-and-click waveform design.

AWG trends

Fundamentally, an AWG is a sophisticated playback system that delivers waveforms based on stored digital data. The digital data describes the constantly changing voltage levels of an arbitrary signal. An AWG starts with a digital representation of waveforms that are either defined mathematically or derived from measurements made on actual signals and then stored as waveform files. These files are then played back from memory. As you might expect, this practice makes memory a valuable commodity for AWGs. The more memory available, the longer the run times for generated waveforms.

Sample rate, usually specified in terms of megasamples (MS/sec) or gigasamples per second (GS/sec), denotes the maximum clock or sample rate at which the instrument can operate. The sample rate affects the frequency and fidelity of the main output signal. The Nyquist criterion states that the sampling frequency, or clock rate, must be more than twice that of the highest spectral frequency component of the generated signal to ensure accurate signal reproduction. To generate a 1-MHz sine wave signal, for instance, it is necessary to produce sample points at a frequency of more than 2 MS/sec. Although Nyquist is usually cited as a guideline for acquisition, as with an oscilloscope, its pertinence to signal generators is clear: Stored waveforms must have enough points to faithfully retrace the details of the desired signal.

Modern, high-performance AWGs offer deep memory depth and high sample rates. These instruments can store and reproduce complex waveforms such as pseudo-random bit streams. Similarly, these fast sources with deep memory can generate extremely brief digital pulses and transients. Memory depth plays an important role in signal fidelity at many frequencies because it determines how many points of data can be stored to define a waveform. Particularly in the case of complex waveforms, memory depth is critical to reproducing signal details accurately. Expressed by the number of binary word samples, memory can range from 2 Gsamples for mid-range AWGs to 16 Gsamples for high-end instruments. For reference, 2 Gsamples of waveform memory can play 400 msec of data at 5 GS/sec, which is sufficient for most applications.

An instrument’s bandwidth is independent of the sample rate. The analog bandwidth of a signal generator’s output circuitry must be sufficient to handle the maximum frequency that its sample rate will support. In other words, there must be enough bandwidth to pass the highest frequencies and transition times that can be clocked out of the memory, without degrading the signal qualities.

Another important AWG consideration is vertical resolution. It pertains to the binary word size, in bits, of the instrument’s DAC. The vertical resolution of the DAC defines the amplitude accuracy and distortion of the reproduced waveform. A DAC with inadequate resolution contributes to quantization errors, causing imperfect waveform generation. While more is better, in the case of AWGs, higher-frequency instruments usually have lower resolution. Taking advantage of the latest advances in DAC technology, AWGs are now appearing on the market with an impressive combination of 16-bit resolution and a 10 GS/sec sample rate. AWGs with higher sample rates in the 50 GS/sec range or so typically have 8- or 10-bit vertical resolution.

A growing number of applications require multiple output channels that must be precisely synchronized. For obvious reasons, anti-lock braking systems in cars require four stimulus signals. Biophysical research applications call for multiple generators to simulate various electrical signals the human body produces. Quantum computing applications push multi-channel output requirements to new levels. These applications involve sending dozens or even hundreds of synchronized signals to manipulate q-bits. Another extreme example is that of occupied spectrum measurements where designers must try to simulate an environment filled with RF/microwave signals from military and commercial radar and radios as well as countless consumer devices. For these measurements, the more test sources the better.

To meet multi-channel signal generation requirements, the latest AWGs offer up to eight analog output channels per instrument, and multiple instruments can be synchronized to create large systems with 32+ channels. In the case of the Tektronix AWG5200 introduced earlier this year, each of the instrument’s eight independent channels provide less than 10 psec channel-to-channel skew. Each of the channels have independent paths out, individual amplification, separate sequencing, up-conversion, dedicated memory, and can be controlled independently with minimized cross talk or limitations on performance. The only common factor is that all channels share a common clock or, if the user chooses, can share an externally supplied reference clock.

Advancing DACs

The push to reduce the size and cost of telecommunication and military systems is driving manufacturers to integrate more functions into a single DAC chip. Some advanced high-speed DACs also incorporate digital signal processing and conditioning functions such as digital interpolation, complex modulation, and numerically controlled oscillators (NCO). These features enable direct generation of complex RF and other signals in an efficient and compact way.

AWGs that incorporate these highly-integrated DACs offer such features as a digital complex modulator and multi-rate interpolation. With internal quadrature (also called IQ) modulation, IQ mismatches often attributed to external modulators and mixers are eliminated. (IQ imbalances arise from mismatches between the parallel sections of the receiver chain dealing with the in-phase (I) and quadrature (Q) signal paths.) Also, there is no in-band carrier feed-through, and there are no images. Built-in interpolators also support the ability to create waveforms more efficiently. This feature reduces waveform size and compilation times and extends playback time as well.

DACs offer a variety of familiar modes including non-return-to-zero (NRZ) mode, return-to-zero (RZ) and mix-mode. AWGs provide access to these modes to ensure signals are output at the cleanest portion of the DAC BW and frequency roll-off positions.

Creating Waveforms

It has always been a significant design challenge to create stimulus signals that can fully exercise a prototype. In the case of RF applications, complex high-frequency modulated signals with jitter, spread-spectrum clocking, and other time-variant effects have traditionally required a benchtop of pulse, function, modulation and RF generators.

Compared to these alternatives, an AWG provides finer control, granularity and repeatability for the signals and stresses being generated. The wide variety of different waveforms that AWGs create can ultimately lead to thousands of different test scenarios. Additionally, the AWG’s sequencing abilities can be helpful. Sequencing means iterating over several segments with each segment comprised of several samples. AWGs may support several stored sequences and may also allow programmatic control of the sequences using external triggers. Sequencing allows users to sweep through a wide range of voltage values and stresses. The AWG generates multiple waveforms and sequences them so they emulate the effect of turning the knob to dial-in jitter.

Another advantage of AWGs is their ability to save waveforms, edit them offline, then share them with teams globally. This can help to isolate bugs early in the development cycle for globally dispersed teams, for example.

For many users, the default approach for generating signals is to use MathWorks’ Matlab and the Signal Processing Toolbox, which provides a broad set of functions and apps to generate, measure, transform, filter and visualize signals. The Instrument Control Toolbox lets users then configure and control AWGs in Matlab.

For users who fall short of Matlab guru status, however, it can be a challenge to create the signals needed for complex applications, particularly if there are many subtleties and changes that must be made quickly for what-if analysis. It can be time consuming, for example, to use Matlab in applications such as the testing of receivers for compliance to complex standards such as MIPI D-PHY or PCIe Express. The process will involve painstakingly walking through the standards’ test documentation to add appropriate stresses and impairments needed to simulate real-world conditions.

To speed the process of creating waveforms for specific applications or to simulate complex real-world environments, AWGs are now available with a special plug-in architecture. It integrates into the AWG’s GUI, or can be run on a PC and be moved to the AWG over an Ethernet LAN. The library of available plug-ins continues to grow. If plug-ins are available for a particular application area of interest, the plug-in should be considered the go-to starting point. The plug-in can always be complemented with waveforms created externally or imported from a test instrument such as an oscilloscope.

Here’s a rundown on some of the more common plug-ins and their capabilities:

Generic precompensation – Users today need the cleanest signals and the lowest EVMs possible. The precompensation plug-in simplifies the process of generating correction factors and applying them to get the best signals and cleanest performance from an AWG.

For instance, when creating waveforms that test wideband receivers, it is important that the AWG signals have a flat frequency and linear phase response. Users can compensate for the first and second Nyquist zones of the AWG. Users can define the LO frequency and choose to get correction coefficients for either the lower or upper side bands as well as define the carrier frequency. Users can also define the bandwidth of compensation either by specifying start and end frequencies (RF & IF) or bandwidth (in IQ/IQ with modulator).

High-speed serial – High-speed serial data signals continue to grow more complex. The plug-in for high-speed serial simplifies signal creation and jitter simulations to reduce overall development and test time. It can be used to create the waveforms required for thorough and repeatable design validation, margin/characterization and conformance testing.

Specific capabilities include generation of jitter (random, periodic (sinusoidal), inter-symbol interference (ISI), and duty cycle distortion (DCD) as well as spread-spectrum clocking (SSC), pre-emphasis, and noise addition. A combination of various impairments can be created simultaneously to stress the receiver. The input data pattern can be scrambled by defining a polynomial. The pattern duty cycle can also be defined using pulse-width modulation (PWM), which allows for alternatively encoding the bit stream at up to 16-PAM.

Multi-tone and chirp – This plug-in is useful for military, aerospace, and RF applications where creating and generating tones are part of a successful mission. Tones can be created for various applications, including noise power ratio (NPR), with a set of desired start and end frequencies, spacing or the number of tones. Frequencies can be notched out by setting the start and end frequency of choice. When generating chirps, the user can decide between high-to-low or low-to-high frequency sweeps and define chirp qualities by sweep time or by sweep rate.

While creating waveforms for testing wideband receivers, it is important that the test equipment generate signals with flat frequency and linear phase response. In this regard, correction files can be directly applied to tones or chirp waveforms while they are being compiled.

Optical – High-speed telecommunications networks are moving toward faster and more complex modulated signals. The optical plug-in simplifies waveform creation to reduce design iteration intervals. Single or dual-polarization modulation schemes can be defined with a variety of preloaded modulation formats including BPSK, QPSK, OQPSK, OOK, NRZ, up to 8 PAM and QAM8. Baseband data can be selected from several predefined patterns, user-defined patterns, or from a pseudorandom binary sequence (PRBS) 31 generator, and each N-bit word can be defined with symbols from N unique data streams.

Whether you use one of the above plug-ins standalone or design waveforms using other software tools like Matlab or Excel, it has never been easier to generate high-fidelity signals for test and measurement applications in conjunction with a modern AWG.

REFERENCES
TEKTRONIX INC., AWG5200