Microwave Instrument Advances
About this Article
This paper by Ian Instone was first
presented at an IEE colloquium in January 1996, "The Use of
Software and Firmware in Measuring Instruments".
Network Analyzers
Until comparatively recently characterizing passive microwave components
could be a tedious task if sufficient measurements are performed
across the frequency band to ensure that the device has a predictable
response. An apparently simple device such as a fixed attenuator
needs three parameters measured involving two different sets of
test equipment. These are the attenuation and the mismatch that
must be measured at each port. Another complication arises because
the measured values will be influenced by effects related to the
measuring equipment and its interaction with the device being calibrated.
Consequently these devices would often have only been measured at
one or two frequencies and their response at other frequencies either
predicted or interpolated. With the initial development of network
analyzers during the early 1970s the situation changed dramatically.
Network analyzers are instruments designed to measure all the necessary
parameters of "two port" networks simultaneously. Measuring
in this way makes it possible to assess the dependency of the measurements
upon each other and on the measuring equipment. Early analyzers
took the form of several independent measuring instruments interconnected
together and controlled by an external computer. These systems had
many limitations over present day instruments, some of these being:
- Space:
A typical network analyzer system required a vast amount of space,
often occupying around three two metre high instrument racks.
- Stability: The high frequency
sources used in these systems were thermionic devices. Synthesized
sources were normally only constructed using solid-state components
but unfortunately these tended to be limited to UHF frequencies
and below. Non-sythesized sources have a much higher residual
frequency modulation and phase modulation content which results
in a larger IF (Intermediate frequency) bandwidth being necessary
in the receiver. Larger bandwidths produce higher noise that reduces
the useful dynamic range over which stable measurements can be
performed.
- Speed: These systems comprised
of many interconnected instruments distributed over several racks
which limited the operating speed to that governed by the slowest
instrument within the system. In addition computers were very
expensive so appreciable cost savings with no adverse effects
upon measurement accuracy were possible by compromising the computer.
Of course, the network analyzers were a big improvement on the
traditional measurement techniques in use at that time. These usually
involved a receiver to measure loss and a slotted line to perform
the impedance measurements in the form of standing wave ratio (SWR).
Using these instruments it was possible to measure phase angle but
this was often beyond the scope of many people due to the mathematics
involved. Measurements performed using these instruments were time
consuming and had to be performed individually at each frequency.
Wide-band signal generators were not available until a few years
later so a separate signal generator was required for each frequency
band where measurements were performed. To fully characterize a
"simple" device such as a resistive attenuator would need
a large amount of very expensive test equipment and would take several
hours. There was a temptation to perform only a few of the easier
measurements and make assumptions (or interpolate) about the values
at other frequencies. Fortunately, the instruments used at this
time had one saving grace, their accuracy. Slotted lines were carefully
made exploiting the best conditions and manufacturing equipment
available at the time whilst the receivers performance was usually
enhanced by employing a physical standard such as a piston attenuator
as a reference. Network analyzers on the other hand, had mediocre
accuracy which is improved by using a technique known as error correction.
In modern instruments this is performed in "real time"
using the analyzers internal processor, where as the earlier generation
of systems would employ external computers.
What Is Error Correction ?
Consider this simplified circuit diagram of a typical two-port device
(a fixed attenuator). 
Measurements normally made on these devices are the loss from input
to output and the impedance at the input and output ports. These
measurements are most useful if they are performed in the transmission
medium in which the device is designed to operate, in most cases
this will be 50 ohm. With the device shown above when the input
is driven by a source it will have a loading effect which will affect
the amplitude of the signal at the input port. Similarly the load
of the measuring equipment on the output port will affect the amplitude
of the measured signal. Generators and receivers are rarely a good
match to the 50 ohm system, i.e. their impedance isn't 50 ohm. Before
network analyzers were in common use, tuners would be fitted to
the output port of the signal generator and to the input port of
the receiver so that they could be tuned, at that particular frequency,
to be as close as possible to the characteristic impedance of 50
ohm as possible. This process would have to be repeated at every
frequency where measurements were made. On the other hand, the network
analyzer measures the device without the benefit of tuners. Prior
to measurement the network analyzer is characterized using several
known but simple devices. The device is then measured, the effects
of the network analyzer upon the measurements calculated and subtracted
from the measured values. This leaves a set of "error corrected"
values for the device under test.
Network analyzers rely on good quality standards and precise mathematical
modeling of them for their accuracy. These standards are usually
an open circuit, a short circuit and a load. The load must be as
close as possible to the characteristic impedance (50 ohm or 75
ohm) across the frequency range to be measured. The open and short
circuits are the easiest to achieve and verify using traditional
measurement methods where possible. The load has the greatest effect
upon the measurements and is also the hardest item to manufacture.
Various methods can be used to eliminate the load from the measurement,
the most common being the use of sliding load. The sliding load
consists of a length of precision transmission line along which
the load is moved. When the sliding load is moved it will produce
a circle when its impedance is observed on a polar diagram. The
center of this circle represents the impedance of the airline, thereby
removing the load from the measurement process. A modern development
of this technique involves the use of a short piece of precision
transmission line fitted to the front of the load. Since the length
of the line is known it is possible to calculate the phase shift
introduced when using and not using the line. Having established
the phase shift it is now possible to calculate the center of the
circle and therefore remove the load from the equation leaving just
the precision transmission line as the reference standard. The open
circuit is the second hardest to manufacture. It can be replaced
by the short circuit and the same short precision transmission line
described above. For these test methods to be effective we must
have the following:
- measurements of both phase and magnitude
- known and characterized standards
- a means of calculating the results
Network analyzers have become possible in the last 25 years because
computers and microprocessors have made the necessary mathematics
invisible to the user. For instance the correction equation for
the impedance of a two port device takes the form: 
The terms preceded with "S" are the measured parameters
of the item being calibrated and those preceded with "E"
relate to the network analyzer and are assessed during the initial
error correction procedure. Each of the quantities is complex, and
a similar expression will be required for each of the remaining
parameters. In addition, the equations are required at each frequency
where measurements are required. Before computers or calculators
the mathematics would have been very time consuming which made this
method of measurement impractical. With the first generation of
network analyzer, the computer often performed the calculations
in less time than was required for the measurements to be performed.
It became normal practice to measure attenuators, cables, couplers
etc. at frequencies up to 12.4 GHz (defined by the equipment available)
at 500 MHz intervals in a little under an hour, including the initial
error-correction.
The basic network analyzer changed a little during the 1970s taking
advantage of the emerging technologies. As solid state sources became
available at higher frequencies they replaced the previous thermionic
devices. These advances quickly produced an extended frequency range
and up to 18 GHz became available as an option. In addition the
emerging technologies allowed the computer to take control of the
amplitude of the source making active device characterization more
of a reality. The rapid advances in computer technology allowed
a more friendly and flexible user interface coupled with the ability
to perform measurements at many more frequencies. The biggest single
advance occurred around 1983 when the computer was placed inside
the network analyzer The entire system was redesigned producing
a complete self contained system occupying one instrument rack approximately
1.5 metres high. With the computer located inside the instrument
the limitations posed by transmitting data through long bus cables
was eliminated so a "real time" display of the corrected
measurements was provided on the internal CRT display. In addition
an option was included where the internal computer could display
the data in the time domain, allowing the operator to investigate
discontinuities etc. in his structure. The time domain display was
further enhanced using a feature known as "de-embedding".
De-embedding is a process where markers are positioned around the
time domain display of an item which is buried within a structure.
The display between the markers is then computed back into the frequency
domain so that the parameters of the device can be assessed whilst
it is still connected and active within the circuit. This process
is used to great effect when the network analyzer develops a fault
in one of the microwave components, it is possible to look inside
the analyzer and locate the faulty component without taking any
of the covers off the instrument.
During the 1980s improvements were made to the display, operating
frequency range, size, functionality and price of these instruments.
Most of these instruments now carry a substantial amount of memory
which enables them to record the results of several hundred measurements.
Most instruments now have an internal disk drive which doubles as
memory for recording results, and also provides them in a format
which can be processed away from the instrument at a later date.
Modern instruments can act as instrument controllers in their own
right, controlling such instruments as power meters. This enables
precise power levels to be set-up prior to device measurement so
parameters such as gain compression can be measured more accurately.
In the latter half of the decade the "economy" range of
network analyzers were released. These provided all of the functionality
of the "full price" analyzers at under half the cost.
The only area where performance has been compromised is in the design
of the microwave source, it produces substantially less power than
that in the other instruments therefore limiting the dynamic range
slightly. Couplers have been used have been employed in the lower
cost instruments instead of "tri-axial" bridges. These
reduce the cost of the instrument by several thousand pounds at
the expense of the low frequency performance which is compromised
slightly.
The recession of the 1990s brought about customers who specified
the instruments they required much more precisely. It was no longer
acceptable in the market place to produce an instrument which could
measure all known parameters and expect it to sell regardless of
cost. Manufacturers had to listen to their prospective clients and
build instruments which met the specification. Customers are not
willing to pay for features or performance which will not be utilized.
At the same time many customers are demanding that their instruments
might be upgraded in the future to meet their unforeseen needs at
that time. The use of "upgradeable" firmware within the
instrument helps to meet this requirement. For instance upgrading
a standard "economy" network analyzer to perform measurements
in the time domain is can be performed by the user with a 3.5 inch
disk supplied by the manufacturer. The process takes a couple of
minutes. To prevent the disk being used to upgrade more than one
network analyzer it is encrypted with some data relating specifically
to the analyzer submitted for upgrade. Software and firmware is
available for many of the measurement tasks performed using network
analyzers, including:
- Antenna Measurement
- Dielectric Measurement
- Transistor Parameters
- Amplifier Characterization
- Power Sensor Calibration
- Fault Location (in cables etc.)
Spectrum Analyzers
Spectrum analyzers are another instrument which have benefited
greatly from the inclusion of microprocessors. Early microwave spectrum
analyzers were more akin to low frequency instruments with an input
mixer and local oscillator added. Unfortunately, as described earlier,
wide band oscillators were not available therefore to keep the spectrum
analyzer to a manageable size and cost an available oscillator coupled
to a harmonic mixer was employed. By the nature of the way that
they work, harmonic mixers will produce some undesirable effects.
These include an image of the signal being measured at every harmonic
generated within the mixer. Also, because the mixer was designed
to produce harmonics, it will also create harmonics using the signal
under test. These effects can be either compensated for, or can
be identified in some way and ignored by a suitably qualified operator.
To reduce these effects a pre-selector was added as an option. This
is a filter placed before the RF mixer which ensures that only the
frequency which the spectrum analyzer is tuned to is applied to
the mixer. The preselector helps to prevent the mixer from becoming
overloaded and therefore reduces many of the "harmonic and
spurious" signals generated. A further complication arises
because the preselector must track the input frequency of the analyzer
exactly. This was not easily achievable so a tracking control was
provided. This further complicated the operation of the instrument.
Microprocessor controlled spectrum analyzers started to evolve around
1980. The major benefit was their ease of use over earlier instruments.
The choice of scan time and resolution bandwidth settings could
now be left to the microprocessor to optimize. If the operator chose
a some instrument settings which would have produced an out of calibration
display the microprocessor would reset the instrument to the most
appropriate settings closest to those desired so that measurements
would not be compromised. The operator was still left with the choice
of overriding the microprocessor, and would be reminded with a display
on the screen if their combination compromised the measurements.
Adding the microprocessor reduced the quantity of manually performed
internal adjustments on the analyzer For instance, instead of adjusting
the preselector to track the analyzer at all combinations of sweep
speed and scan width using a multitude of potentiometers, inductors
and capacitors it was now possible to perform a few fairly crude
adjustments and then let the microprocessor perform the remaining
adjustments whilst the instrument was performing measurements. The
microprocessor also brought with it the ability to control every
instrument setting remotely using a computer. Data is transmitted
in either direction, so the instrument can be set-up and then interrogated
to provide the measured values. Signal identification is provided
using the marker, as is the facility to lock the analyzer to a signal
which is drifting.
Through the 1980s the spectrum analyzers performance and features
improved as microprocessors became cheaper. The perceived accuracy
was always a problem with these instruments and the microprocessor
was again able to help. Data is stored in ROM relating to the performance
of various components in the RF signal path. These values are implemented
as corrections when necessary. Normally corrections are applied
for frequency response, IF and RF linearity and attenuator errors
and errors introduced when changing resolution bandwidth.
Another significant enhancement was realized when digital filters
were introduced. Making measurements using the spectrum analyzer
always involved compromises between speed, bandwidth and sweep time.
With digital filters these compromises were no longer significant.
With a digital filter the signal is measured using a wider bandwidth
and the response is mathematically interpreted using the internal
microprocessor. Digital resolution bandwidths are 3 to 600 times
faster than comparable analogue filters. The reduced bandwidths
provide the spectrum analyzer with a full 100dB on-screen calibrated
display. Resolution bandwidths of 1Hz on spectrum analyzers operating
up to 50GHz are now available.
As customers became more selective in their choice of instruments
it became necessary to tailor them exactly to their requirements.
The familiar "smart card" was the vehicle chosen since
it was fast and cheep. Various programs, measurement algorithms
etc. were stored in the memory of the smart card and used to configure
and control the analyzer In this way one Spectrum Analyzer could
be used for several different functions by simply changing the smart
card, for instance without any smart card it remains a general purpose
laboratory analyzer, an EMC smart card will adapt it to perform
EMC pre-compliance measurements and another smart card could be
used to perform some of the measurements necessary when checking
mobile telephones. The modern spectrum analyzers are able to act
as instrument controllers in their own right so the smart card can
be configured to drive instruments, printers, plotters, disk drives,
etc. which are external to the analyzer The guide for writing code
on smart cards is made available to customers and third parties
so that instruments can be customized on a "one off" basis
if necessary.
When the spectrum analyzer is connected to an external harmonic
mixer to perform measurements at very high frequencies the microprocessor
is able to calculate where spurious signals generated within the
mixer are likely to occur and identify them on the display. This
process reduces the skill level which was necessary when operating
earlier instruments.
The all important subject of routine maintenance and calibration
is addressed for some analyzers in the form of a module which fits
to the rear of the instrument which performs high level diagnostics,
various self tests, and key adjustment procedures. The module controls
both the spectrum analyzer and the necessary external test equipment.
Adjustments are performed using the analyzers microprocessor to
generate corrections for the errors determined in the calibration
routine. The errors are stored in the analyzers memory. Smart cards
are available for the following spectrum analyzer applications:
- Cable TV Measurements
- EMC Measurements
- GSM Measurements
- Digital Radio Measurement
- Phase Noise Measurement
- TV Broadcast Measurements
- Scalar Network Measurements
- Noise Figure Measurements
- Link Measurements
and many others.
Signal Generators
Microwave signal generators have benefited from the introduction
of microprocessors into their design. Many of the adjustments necessary
to produce a wide band, stable, constant amplitude source are now
performed by the microprocessor each time the output frequency or
amplitude is altered. In order that clean, harmonic free signals
are produced many modern signal generators include a YIG filter
which tracks the output signal. Since these are a magnetic structure
there will inevitably be some hysteresis which must be compensated
for. Earlier designs of signal generators fitted with YIG filters
accepted that it would not be possible to tune the YIG to the output
frequency at every point by compensating for the additional loss
in a degraded maximum output amplitude specification. The addition
of a microprocessor results in the filter being optimized every
frequency allowing the signal generator to produce higher amplitude
signals at its output socket.
Many signal generators now have very tight specifications regarding
the output amplitude accuracy despite the hardware used in the output
stages of them having changed little over the last 20 years. These
tight specifications are only possible because the complete signal
generator is calibrated after manufacture and the measured errors
stored in a ROM to be used as correction values by the microprocessor
when the instrument is used. The use of microprocessors and their
respective firmware enables instrument designers the freedom to
use established designs of hardware and to compensate for the inadequacies
using correction values stored in the instruments memory.
Oscilloscopes
Oscilloscopes have had a "bumpy ride" as microprocessors
have been incorporated into their design. The first generation of
programmable oscilloscopes had a bland front panel with several
pushbuttons and one control knob. This was an excellent design concept
if the instrument was only ever going to be operated using a computer,
but if it was to be operated manually the multilayer menus were
so different from the traditional analogue instrument that few engineers
were able to operate them efficiently. As digital oscilloscopes
have evolved design engineers have tried to recapture the traditional
analogue feel without compromising performance. To a large extent
this has been very successful with the modern 50GHz oscilloscopes
being easier to operate than the familiar analogue instruments .
Much of this is due to the short cuts built into the firmware, for
instance in order to initially display a signal the microprocessor
sets the timebase and amplitude to display two complete cycles over
the center eight divisions of the display. Very often there will
be a "stats" (statistics) button which instantly provides
a display of repetition rate, peak to peak amplitude, peak amplitude,
average and RMS values. Another "stats" control might
provide values for rise and fall time, duty cycle, overshoot and
undershoot. All of this information is also available over the instrument
bus.
Many of the microwave oscilloscopes also include a very fast pulse
generator for use as a Time Domain Reflectometer. With the modern
instruments having a 50GHz bandwidth, the location of faults in
fairly lengthy pieces of cable can be located very accurately.
It goes without saying that modern oscilloscopes are able to produce
hard copies of their measurement data using a wide range of printers,
plotters and disk drives. This has made the Polaroid cameras used
during the 1960s and 1970s redundant. The modern inkjet or laser
printer produces a much clearer image which can be enhanced with
markers to show measured values at certain points. If desired, all
of the measurement data used to create the display can be printed
or saved on a disk.
Summary
The use of microprocessors and firmware in measuring instruments
has revolutionized metrology. Measurement techniques, which a few
years ago would have been considered uneconomical to perform are
now executed routinely, and often by relatively unskilled operators.
The ability to monitor a components parameters whilst it is connected
in a circuit, without introducing additional loading effects helps
to speed up the development cycle of many devices.
The quantity of measurements performed on any individual item has
increased to the extent that we have a much better knowledge of
the how the items behave with changes in amplitude, frequency etc..
For points where measured values are not provided we can interpolate
with more confidence. These benefits ultimately help to raise product
quality.
|