Function Generators vs. Waveform Generators? The Automotive Ethernet case

Introduction

Modern signal generators are built using two different HW architectures that also result in a different user interface, giving rise to two types of products: arbitrary function generators, or AFGs, and arbitrary waveform generators, or AWGs.

Both types of tools are very flexible and equally popular on the market. However, it is important to know their differences in order to choose the most suitable test equipment for the specifi c application, as we will discuss in the following paragraphs. As a test case, we will consider the characterization of a receiver for a fast serial bus, such as Automotive Ethernet.

Direct Digital Synthesis (DDS) Function Generators

Direct digital synthesis, better known as DDS, is a signal generation technique that uses digital devices to generate a frequency-variable and phase-variable signals from a fixed frequency reference clock signal.

This is the method typically used by less expensive signal generators guaranteeing excellent results, but it has an important limitation arising from the use of a fixed sampling rate.

It is worth remembering that in an arbitrary generator, the desired custom waveform is ‘drawn’ using a certain number of samples.

Figure 1: DDS Schematic Function Generator

The output frequency of the generated signal is closely related to the sampling rate of the generator and the number of samples that define the waveform according to this simple relationship of fundamental importance:

In practice, a DDS-type generator generates a waveform at a certain frequency by taking data from a lookup table at a rate determined by the sampling rate, which, remember, in this case is fixed.

Thus, when we wish to generate the same waveform, but at a higher frequency, the number of samples taken into account by the DDS generator will decrease, with the result that the actual waveform generated will be somewhat less accurate than our sample waveform description.

Similarly, if we wish to generate the same original waveform at a lower frequency, the sampling rate being fixed, the DDS generator will replicate some of the points of the original sample waveform description.

In essence, for high-resolution waveform generation, direct digital synthesis can lead to the generation of waveforms that are essentially too approximate for applications where complex and detailed waveforms are needed.

Depending on the output frequency, some small variations and details present in the sample waveform may be completely absent in the actual generated waveform. In the best case, this contributes to increased jitter, while in the worst case this excessive approximation may result in very distorted output waveforms, or even lack of some details.

On the other hand, the great advantage of the DDS technique, in addition to its ease of implementation that helps to lower the cost of generators, is due to the ability to easily and quickly vary the frequency of the output signal avoiding phase discontinuities, facilitating for example the creation of signals with frequency sweeps in real time, even at high speed.

True Arbitrary Waveform Generators

Instruments that are often referred to as true arbitrary waveform generators, or ‘True AWGs’, have a conceptually very simple operating principle, corresponding to the inverse of a common digitizer or digital oscilloscope.

This type of arbitrary waveform generator is sometimes also referred to as ‘PPC’, or Point Per Clock, for a reason that will become apparent in a moment.

Figure 2: True ARB variable clock

Each point of the sampled waveform representation is stored within a memory, which is read sequentially via a variable frequency reference clock signal. The values of the memory cells sequentially addressed by the clock signal are forwarded to the digital-to-analog conversion (DAC) circuit, which is followed by a low-pass filter.

Since the number of samples used to describe the waveform to be generated is fixed, to vary its output frequency it is sufficient to change the frequency of the clock generator, which will result in faster forwarding of samples to the DAC.

In other words, the frequency of the output signal depends only on the frequency of the clock signal and the number of samples in memory that describe the waveform, according to the relationship already illustrated above.

Unlike DDS-type generators, True Arbitrary waveform generators always use all the individual samples describing the waveform, whatever the output frequency.

As a result, the desired waveform can always be generated with the same accuracy, whatever the frequency of the output signal. Samples are never skipped or repeated, so even the smallest details stored in the sample waveform are always reproduced in the output waveform.

Therefore, the accuracy depends only on the resolution of the waveform's description and the stability of the generator components, while it is essentially independent of the output frequency of the generated waveform.

Accuracy that remains independent of output frequency and waveform complexity is the main advantage of true arbitrary waveform generators over direct digital synthesis function generators.

On the other hand, it is more difficult to perform a real-time frequency scan of the generated signal, since all the definition points of the waveform have to be recalculated as a function of frequency.

Pros and cons of DDS and True Arbitrary signal generators

To get the best of both worlds, there are also arbitrary function generators that allow you to optionally use both generation methods in the same instrument, depending on your specific needs.

Teledyne Test Tools' T3AWG3K series models are examples of arbitrary waveform generators that integrate both DDS and True Arbitrary modes of operation in the same instrument.

Depending on the application requirement of the moment, the generator switches from one mode to the other by simply pressing a button on the front panel.

Figure 3: T3AWG3358 front

How to stress an Automotive Ethernet receiver

Suppose we want to test a receiver for a fast serial bus, such as Automotive Ethernet in its 100Base-T1 version.

We would not only like to see if the receiver works properly when receiving a well-formed and undistorted signal, but also how the receiver behaves when faced with signals affected by various imperfections, such as jitter, noise, incorrect voltage levels or other types of defects.

To perform these tasks, we therefore need a stimulus signal generator that is flexible enough, accurate enough, and easy enough to use to create a sequence of test signals that allows us to understand what are the true operating limits of our receiver.

The Automotive Ethernet bus works with PAM3 modulated signals, which is a ternary amplitude modulation capable of transferring a ternary symbol (+1, 0, -1) for each clock period.

The version of the 100Base-T1 standard uses a 66.667 Mb/s clock and signals with a nominal voltage of 1 V, or 0 V and -1 V corresponding to the symbol to be transmitted, so the receiver works with two levels of decision threshold to distinguish between the three allowed symbols.

It is precisely in these cases that an arbitrary waveform generator excels, because the stimulus waveforms needed can be articulated and not easy to obtain with a classic function generator

Figure 4: Clean PAM3 signal with two levels of decision threshold

In fact, as a starting point we need a clean PAM3 signal of adequate frequency and amplitude, which is not easily available in all function generators.

But, above all, we also need a whole series of ‘distorted’ PAM3 signals, so that we can stress our receiver under test, to understand to what extent it continues to work satisfactorily.

Let’s see how to generate such a sequence of stimulus signals with a modern arbitrary function generator, the T3AWG3K model from Teledyne Test Tools.

Let’s assume that we want to generate a sequence of stimulus signals consisting of a clean pseudo-random PAM3 waveform, followed by a similar but noise-affected waveform, a next one affected by even louder noise, and so on, for a total of five waveforms.

Segments and components

Using the T3AWG3K generator in AWG mode we can very easily create a pseudo-random PAM3 stimulus signal using the Waveform Editor.

In this environment we can create waveforms of any type and complexity by combining various segments and components.

Segments correspond to successive portions of the waveform in the time domain. Each segment of the waveform, in turn, can be composed of the superposition of several components.

. Figure 5: Clean PAM3 100Base-T1 clean signal created using Waveform Editor Utility

Let’s start by creating a PAM3 signal for Automotive Ethernet. In our case we will only need one segment. We will construct the segment as a sum of two components.

As the first component we choose from the predefi ned waveforms available in the Waveform Editor the pseudo-random sequence PRBS7, which will become the fi rst component of our segment. As second component of our segment we choose the pseudo-random sequence PRBS8, which added together create a clean PAM3 signal that we can use as a stimulus to test our Automotive Ethernet receiver

Figure 6: PAM3 100Base-T1 signal with added noise

To construct a sequence of stimulus signals increasingly affected by noise, we duplicate the waveform created earlier and add a new component to the segment. This time, from the predefined signals available in the Waveform Editor we choose white noise and set a level of 100 mV. That’s it: we now have our noisy stimulus shape.

Figure 7: Sequence of PAM3 signals with increasing noise contribution

With the same procedure we can create the other waveforms affected by an increasing noise level.

At this point we can take advantage of the Sequencer capability of the T3AWG3K generator to create a complete sequence of stimulus signals to send to our receiver.

We load in each entry of the Sequencer the sample waveforms previously created, starting with the clean one followed by those with increasing noise. For each entry we set the infinite repetition mode and the trigger transition to the next waveform to pressing a button.

Figure 8: PAM3 signal with increasing noise as acquired and analyzed by an oscilloscope

Now we are ready. Starting the sequence will first generate the clean PAM3 signal. Each time we press a button, we will add noise to increasingly stress our receiver and evaluate its operating limits.

Just as we have created waveforms affected by noise, we could just as easily create other waveforms affected by different imperfections. For example, we could alter one or more symbols by directly intervening on their voltage level, or we could add jitter or voltage offset, or we could create stimulus signals with multiple overlapping defects.

The deep memory available and the high resolution in the T3AWG3K generator does not impose any practical limit on the complexity and articulation of the stimulus signal.

Moreover, in the case of offset, the T3AWG3K generator offers an additional advantage. Since the offset voltage level on the output signal is added by hardware, this allows us to always take full advantage of the high vertical resolution of the instrument (16 bits), even when the output signal is strongly unbalanced.

Conclusion

As is often the case, there is no one-size-fits-all answer to select best tool between an arbitrary function generator and an arbitrary waveform generator.

When standard waveforms are needed whose frequency must be varied rapidly, a direct digital synthesis function generator is often the most convenient solution.

However, when more complex and articulated signals are needed, into which defects must be artificially inserted in a controlled manner, as in the case of debugging an Automotive Ethernet receiver, then an arbitrary waveform generator is the most rational choice.

If you have a Teledyne Test Tools T3AWG3K signal generator in your laboratory, this is no problem: the instrument can work in both modes and guarantees high vertical resolution (16 bits), large memory (1 Gp), high sampling rate (1.2 Gs/s), synchronous digital channels (up to 32) and high dynamic range (±12 V) with additional hardware offset (±12 V).