Joan
Mercade, Application Engineer at Tektronix
Introduction
Arbitrary
Waveform Generators (AWG or arbs) are among the richest and most
flexible instrument categories in the instrumentation world. In
theory, they can make create any signal in the mind of the most
imaginative test engineer. Reality is, however, not so rosy. Selecting
the right instrument for a given job may be a challenging task as many
of the features and specifications required may be difficult to
evaluate.
The
Form Factor factor
AWGs come in all shapes and sizes. From card-based architectures to
full-featured self-sufficient bench top instruments, there are a
variety of form factors to fit any application environment. Arbitrary
generators are a very flexible stimulus solution for ATE systems.
After all, their flexibility and versatility match one of the desired
characteristics of any instrument oriented to be used in any automatic
test system: universality.
The same device can generate, say, a serial digital system to emulate
a communication interface while testing a telecom card and a video
signal. While testing a
video processing card, an AWG can be downloading the right sample
sequences and changing the signal generation set-up. All the AWGs in
the market are potentially usable in ATE systems as external computers
can control them and signals can be downloaded to them depending on
the actual needs of the test procedure. In any case, there are
different form factors that can make integration simpler and cheaper.
Card-based architectures such as PXI or VXI are typically seen as more
appropriate for automatic measurement systems. Although transfer speed
and multiple device integration, through specialized built-in timing
and trigger buses, may simplify system integration and speed, the lack
of powerful user interfaces, internal mass storage (such as hard
disks) can result in lower than expected performance levels, specially
when compared to the most powerful, multi-channel bench top solutions
available. As in any measurement situation, the most important factor
should be the instrument performance and accuracy.
Reading
the Specs
A very useful way to understand an AWG is by thinking of an inverted
digital storage oscilloscope (DSO). See
Figure 1. The block diagram is composed of a waveform memory, a
converter (a DAC instead of an ADC), and an amplifier/conditioning
block at the output. Just like any DSO, the main specs are the maximum
sampling rate, the vertical resolution, and the waveform memory
length. Generally
speaking, the faster the sampling rate, the higher the vertical
resolution (expressed as the number of bits for the DAC), and the
larger the waveform memory, the better. There are, however, many
additional issues connected to these basic specs.
Maximum sampling rate defines the frequency band that a given
generator can re-create. According to the sampling theorem, the
maximum frequency band that could be covered is limited to half the
sampling rate. It is important, though, to look at the analog
bandwidth spec as this could be the limiting factor for signal
bandwidth. Does this mean that any extra sampling rate is a waste of
capacity? Not at all. Oversampling can be used to improve the signal
to quantizing noise ratio (as it is spread over a larger band) and the
timing granularity of the resulting waveform. In other words, sampling
faster is like increasing the number of bits of the DAC. In some
application, such as wireless test, it is even possible to reach
higher than expected carrier frequencies by using one of the signal
images of the DAC. This is
possible in those cases where analog bandwidth allows us to do so. It
is important to understand, though, that edges can be positioned with
a much higher resolution and accuracy than the sampling period for any
bandwidth-limited signal so, generally speaking, sampling rate is not
the most relevant spec to properly define timing in a given signal for
applications such as jitter tolerance testing.
Vertical
resolution gives a measure of the amplitude detail of the resulting
signal. This spec is typically connected to the sampling rate, so in
general, the faster the sampling rate is, the lower the vertical
resolution we can get. Current state-of-the-art DACs can offer up to
16 bit at several hundred MS/s and up to 8 bits at several GS/s. The
latest generation of low-cost high-performance instruments are even
offering 14 bits @ 2 GS/s. Again, this spec does not tell us
everything. Theoretically, a higher resolution makes it possible to
create finer details on the final signal. However, there are other
factors that may limit this capability. Analog noise is one of them,
especially for low level signals. The circuits at the output can also
influence the actual quality of the signal. The availability of
different amplitude ranges, switchable low-pass filters, and offset
controls may help to greatly reduce the noise and adjust the available
dynamic range to the requirements of the signal. Some applications,
such as ADC or wireless test, may require some quality level in the
frequency domain. For these cases, linearity (both from the DAC and
the output circuits) may be the limiting factor. Specs such as the
SFDR (
Spurious
Free
Dynamic
Range
) or Spectral Purity (typically for sine waves) may be more
descriptive than just the number of bits of the DAC.
Although
the block diagram of an AWG is similar to a DSO, there is a major
difference between both types of instruments: in DSOs signals are not
a problem as they come from the real world while in AWGs every sample
in the waveform memory is the result of a synthesis process. The
capability to replicate a real-world signal is limited by the
available record length. The longest available time window can be
calculated by dividing the record length by the sampling rate. The
only way to create continuous signals is by repeating the available
time window over and over. This issue is important in applications
such as serial data or wireless signals emulation where a perfect
match between both ends of the signal is necessary to avoid the
classic wrap-around artefacts that can destroy the signal integrity
characteristics. High-performance arbs may implement an elegant
solution to simulate extremely long time-windows based in real-time
signal sequencing. Using this scheme, the available generation memory
can hold different signal segments that are sequenced through a
sequence list. The most advanced instruments can even use external
signals to control branching at the sequence so real-world intelligent
behaviours may be emulated.
Usability
Issues
As in any instrument, the availability of the raw performance does not
guarantee success. This is especially true in AWGs where creating the
signals is an important part of the process. Signals can come from a
variety of sources. Some may be external to the instrument environment
such as simulation tools or digital oscilloscopes. In this case the
capability of importing these signals is crucial. In other cases
signals must be synthesized from scratch. Almost any AWG manufacturer
offers general-purpose software packages running on external computers
that allow creating, editing, and processing signals in different
ways, ranging from drawing them with a mouse up to defining them by
writing complex formulae. Some advanced instruments offer the same
capabilities implemented in their user interface. Although these tools
may be enough for a big portion of users, others will require
specialized packages for vertical applications. A good example of this
may be the creation of wireless signals. Analysing the available
tools, the import capabilities, and the interfacing options for any
AWG may be as critical for the final success as the raw specs.
Many applications may require more than one simultaneous stimulus
channel. Although most AWGs in the market can be externally
synchronized, multiple channel arbs make life much easier for the
users as they guarantee signal alignment and greatly simplifies
control and integration. Some instruments even offer additional
digital signals that may be used for DUT synchronization. In some
cases the waveform memory is connected to digital outputs so it is
possible to use these devices as digital pattern generators or in
mixed-signal application.
Many low-range arbs include function generator functionality. These
instruments are known as Arbitrary Function Generators (AFG). The
function generation block may be implemented as an independent
internal device but most modern devices use a DDS (Direct Digital
Synthesis) architecture, which reuses most of the AWG circuits such as
the DAC. The continuous improvement of AFGs is greatly extending the
number of potential users.
The AFG3000 series from Tektronix is a good example of the latest
generation AWG/AFG, showing very high performance at an affordable
price. See
Figure 2.
Conclusion
As
digital scopes replaced analog scopes in the past, AWGs will replace
many of the generators in laboratories and production lines. Probably,
selecting your next generator means selecting your next AWG.
References
1.
“XYZ of Signal Sources”, Tektronix application note 76W-16672-3,
available at http://www.tektronix.com
2.
“Replicating Real World Signals with an Arbitrary/Function
Generator”, Tektronix application note 76W-18661-0, available at http://www.tektronix.com
3.
“Waveform Sequencing with the AWG2000 Series”, Tektronix
application note 76W-10373-0, available at http://www.tektronix.com
|
| Memory
Management in Arbitrary Waveform Generators |
Arlene
Meadows,
Director of Sales, Tabor Electronics Ltd.
True
Arbitrary Waveform Generators offer the ability to simulate an
infinite variety of electronic and mechanical signals. In order
to properly exercise many Devices Under Test, the test waveform
may need to go on for long periods of time and to contain many
different amplitudes, frequencies, shapes, timing relationships,
or combinations of these parameters. To facilitate the creation
and output of these signals, from simple sine waves to complex
pulse trains and modulated waveforms, the generating instrument
must have the ability to store a large amount of data.
To
create complex waveforms with precise timing characteristics,
true Arbitrary Waveform Generators have increased their sample
rates so that 100MS/s and greater generators are readily
available. True AWGs with sample rates over 1GS/s are also being
produced, such as the Tabor
Model 1281. The need to
control the programming of each of these high-speed points
increases the requirements for large memories in true AWGs.
But
memory chips are expensive and can drive up the cost of a true
AWG, making it too expensive for many users who really need
their flexibility, requiring them to settle for less-expensive,
but much less capable, Arbitrary Function Generators. To
overcome this limitation, true AWG designers have developed a
number of techniques to maximize control of all the points
without requiring massive amounts of expensive memory
components. These techniques, including segmenting, sequencing
and triggering also facilitate the ease of creation and
implementation of complex waveforms.
Sample
Rate Selection
In
selecting an arbitrary waveform generator, the first criterion
will typically be the sample speed required to create the
users’ waveforms. Sample speed in true AWGs is also a
considerable cost factor, so the ability to make use of every
point in designing a waveform is a key budgetary consideration.
The instrument and its control tools should allow precise
control over each point. By cleverly controlling how each point
can be defined, and then concatenated to form segments that can
be sequenced and stored, will make it possible to select a
slower speed generator yet retain control over the necessary
parameters, and ultimately save money.
Memory
Management
Once
a sample rate to define a user’s complex signal has been
determinied it is necessary to determine how to control and
store the points, at an economical price. Try answering the
following questions, which will help in figuring out how much
memory will really be needed:
-
What
is the smallest common denominator of your waveform that you
can draw?
-
Does
it repeat itself? How many times?
-
How
will it know when to advance to the next signal that is
defined as its smallest common denominator?
Determining the answers to these queries helps to give us clues
for selecting the best combination of sample rate and memory
size. A true AWG should have the ability to segment its memory,
then cycle the segments, sequence them and offer flexible
methods of advancing through the sequence. This segmenting and
sequencing is the key to good and economical memory management.
For
example, the Tabor
Wonder Wave series, of true AWGs offers standard memory
sizes from 512kB to 8MB per channel, depending on the sample
rate, with options for each model to double that level. In order
to make these levels of memory work for sample rates starting at
50MS/s to 1.2GS/s, they offer repetitions of up to 1 million
loops, with storage of from 2048 to 16k segments.
Flexible
triggering means that the user can choose from many different
ways to move from one defined segment to another. For example, a
simple clock waveform can be programmed to run indefinitely
until it receives a software or hardware trigger caused by some
other event in the test system such as the cessation of a motor
or an instruction from a complex communication bit stream. The
clock signal then changes to the waveform defined in the
sequence table, perhaps a modulating signal that is telling the
Device Under Test to change speed, update an axis, or output a
new audio or video pattern.
Alternatively,
some instruments allow very complex waveforms to be stored, and
then sequenced by an external signal to simulate a
frequency-hopping radio. In all of these cases, the memory
requirements are kept to a minimum. If the Arbitrary Waveform
Generator did not have the segment memory, it would have had to
store a complete waveform with a finite length and would not
have been able to output the signal indefinitely. The radio
hopping sequence would have to be completely defined before the
test was initiated, meaning that the sequence of hops would be
known in advance and not allowed to be random, or changed on the
fly.
Summary
The
true Arbitrary Waveform Generator that offers separate segment
memories, large number of repetitions, various methods of
sequence storage and flexible triggering modes for advancement
maximizes the generator’s sample speed. This allows the true
AWG to create and output long, complex waveforms, yet stay
within the lean test equipment budgets that management demands
today from R&D labs, production floors, quality departments
and service. |
|
Newsletter
