Saturday, 25 March 2017
  Home arrow Microwave Projects arrow Spectrum Analyzer  
Latest Articles
Main Menu
GMG News
World Microwave News
Job Opportunities
People at the GMG
Microwave Articles
Microwave Projects
Online Software
Discussion Board
Users QSOs
Users Links
Contact us
Login Form

Lost Password?
No account yet? Register
GMG RSS feed
Embed to your site!
Spectrum Analyzer

Spectrum Analyzer 0...1750MHz

Matjaz Vidmar, S53MV

(please wait for the images to load)

1. Oscillators for spectrum analyzers

An important piece of radio-frequency or microwave test equipment is certainly the RF spectrum analyzer. Spectrum analyzers can roughly be divided in two groups: professional and low-cost. Although there are many differences between these two groups of instruments, the most important difference is in the type of (sweep) oscillator used for the first frequency conversion.

Professional spectrum analyzers use YIG (Yttrium-Iron Garnet) oscillators. YIG resonators can be tuned over wide frequency ranges (more than an octave in the microwave frequency range) with an external DC magnetic field. YIG resonators also have a high Q allowing a low phase noise when used in an oscillator. Finally, the tuning characteristic of a YIG oscillator is linear, since the frequency is directly proportional to the applied DC magnetic field.

Low-cost spectrum analyzers use varactor (varicap) tuned oscillators. The Q of varactor diodes is rather low and is inversely proportional to the operating frequency. Silicon varactors usually have the Q less than 30 at 1GHz. GaAs varactors are somewhat better, but they are not easily available and are much more expensive. The frequency coverage of low-cost spectrum analyzers is therefore limited below 1GHz and the phase-noise performance is usually 20-30dB worse than that of YIG oscillators.

A spectrum analyzer can also be built by a skilled radio amateur. While most circuits of professional spectrum analyzers can be reproduced in amateur conditions, the major problem is building a wide-band, low-noise VCO for the first (swept) conversion. YIG oscillators probably can not be built in amateur conditions. The price of a new commercially-available YIG oscillator is comparable to the price of a surplus professional spectrum analyzer.

A varactor-tuned VCO covering the frequency band 2-4GHz will be presented in this article. Such a VCO allows the design of a spectrum analyzer with the first IF in the 2GHz range, similar to professional instruments. The phase noise of the described VCO is reasonably low, within 20dB of a free-running YIG oscillator. Finally, the VCO design is fully reproducible using standard SMD parts installed on a conventional FR4 (0.8mm thick) printed-circuit board.

2. Some oscillator fundamentals

The design of an amateur RF spectrum analyzer therefore depends strictly on the type of VCO that is available for the first conversion. In order to explain the design of a wide-band varactor-tuned VCO, some oscillator basics have to be discussed first. Any oscillator must contain an active device (gain) and a feedback network, as shown on Fig.1. There are two conditions for oscillation: enough gain (including feedback loss) and correct phase of the feedback.

Fig.1 - Oscillator block diagram.

The frequency of oscillation is determined by both conditions as shown on Fig.2. However, close to the actual oscillation frequency, the gain curve has a broad and flat peak. Therefore, the exact frequency, the stability of the oscillator and the phase noise are all determined by the phase response. The steeper the phase slope, the better the stability of the oscillator and the lower the phase noise.

Fig.2 - Frequency of oscillation.

Low-frequency oscillators are usually designed for operation at a total phase shift of 2*PI radians. PI radians are usually provided by the active device (transistor) itself, while the remaining PI radians are provided by the feedback network. A steep phase slope is obtained by a high-Q LC tuned circuit or quartz-crystal resonator.

At frequencies above 1GHz the phase shift of all known active devices is much larger than PI radians due to chip and package parasitics. If an oscillator is designed for a total phase shift of 2*PI radians, then only a small fraction is left to the feedback network. The phase slope of the latter is certainly not very steep resulting in poor stability and high phase noise.

In the case of a variable-frequency oscillator, the frequency coverage is rather restricted since the influence of the feedback network is small compared to the active device itself. Conventional oscillator designs (with a LC circuit or transmission-line equivalent coupled to a negative-resistance active device will only provide a restricted frequency coverage and poor stability. Most microwave oscillators are designed in this way, since a negative resistance can easily be obtained from most microwave transistors when considering chip and package parasitics.

Replacing a negative-resistance device with a true two-port, unidirectional amplifier provides the oscillator designer with some more degrees of freedom. In particular, the feedback network can be tailored for the desired amplitude and phase response. The feedback network should both match the impedances and compensate the phase shift of the active device as well as introduce its own, frequency-dependent amplitude and phase response.

A successful wide-band microwave VCO design is shown on Fig.3, covering more than an octave with conventional silicon varactors. Although I developed this circuit for my first spectrum analyzer built in 1985, I published the circuit diagram one year later as part of a satellite-TV receiver indoor unit [1], [2]. Many other amateurs used this circuit in their own spectrum analyzers and other RF test equipment, but only few acknowledge the original source [3].

Fig.3 - Wide-band microwave VCO.



The major drawback of the VCO design from Fig.3 is that its operation is still based on lumped components: capacitors (varactors) and inductors. Its upper frequency limit is therefore defined by the parasitic inductance of available (packaged!) varactors to about 2-2.5GHz. The phase noise can be reduced by a carefully-designed bias regulator, to stabilize the current through the bipolar transistor, so that the impedances and phase shifts do not change.

3. Interdigital-filter microstrip oscillator

Active-device phase shifts become much larger at higher frequencies. For example, the phase shift inside a helium-neon laser tube may reach one million radians, although its theory of operation is the same as for other electrical oscillators. The total amplifier plus feedback phase shift still has to be an integer multiple "m" of 2*PI radians, however "m" is not restricted to unity and may become very large at light-wave frequencies.

Lumped-component oscillator designs become useless at microwave frequencies, since all available components behave as sections of transmission lines. On the other hand, additional phase shifts can be readily implemented as sections of transmission lines. The total phase shift of a microwave oscillator may be an integer multiple "m" of 2*PI radians. "m" may be larger than unity, but still relatively small at microwave frequencies.

Most microwave circuits are built in microstrip technology, since the latter is compatible with inexpensive manufacturing techniques like printed-circuit boards and surface-mount components. An example of a microstrip oscillator is shown on Fig.4. An interdigital band-pass filter is used as part of the feedback network. In order to bring the total phase shift to an integer multiple of 2*PI radians, additional delay lines may be required to obtain the correct feedback phase.

Fig.4 - Interdigital-filter oscillator.

Although the described oscillator may look complicated, it includes some advantages when compared to conventional low-frequency designs. Although the Q of microstrip resonators is not very high (in the range of 50-100), the phase slope may be made high thanks to the large total phase shift (increasing the multiple of 2*PI radians). On the other hand, oscillation at unwanted multiples of 2*PI radians can be suppressed by tailoring the amplitude response of the feedback network.

A fixed-frequency oscillator can be modified into a VCO by tuning the band-pass filter. For narrow-band operation it is sufficient to tune one of the quarter-wavelength fingers of the interdigital filter. A varactor is therefore inserted in the central finger, since the latter has the highest loaded Q and provides the highest tuning sensitivity.

The varactor should be inserted either in parallel with a voltage maximum or in series in a current maximum along the length of a resonator. Since the capacitance of most varactors is quite high for microwave frequencies, varactors are usually installed in series in current maxima. In the case of a quarter-wavelength resonator, the "cold end" of the latter is grounded through a varactor diode.

The circuit diagram of a narrow-band low-noise VCO is shown on Fig.5. The circuit diagram includes an output buffer (INA10386) to isolate the oscillator from load variations. Some supply voltage and tuning voltage filtering is included for the same purpose, as well as an output coupler for an auxilliary output.

Fig.5 - Narrowband low-noise VCO.



The tuning range of an interdigital oscillator with one single varactor is limited to about 10-20% of the central frequency. The frequency range of the oscillator on Fig.5 is about 1850-2200MHz. Outside this frequency range oscillation is not possible, since the gain maximum does not match the correct phase of the feedback.

A narrow-band low-noise VCO has many applications in frequency synthesizers. The phase noise is sufficient for both analog (SSB) communications [4], [5] as well as digital (coherent PSK) communications [8], [9], [10]. Used in a fast PLL it even allows the correct demodulation of complex radionavigation signals [6], [7].

The tuning range can be increased by increasing the coupling in the interdigital filter. This decreases the loaded Q of the resonators, degrading the phase-noise performance. Further, such an oscillator may also oscillate on higher-order resonances of the interdigital filter. This effect can be observed as "kinks" in the voltage/frequency curve, which is no longer monotonic.

4. Wideband low-noise microstrip VCO

To obtain a true wide-band microstrip VCO, all fingers of the interdigital filter have to be tuned. For example, inserting one BB833 varactor in each of the three fingers of the interdigital band-pass allows a tuning range up to 50% of the central frequency. For example, a VCO with three BB833 and a BFP183 (ft=8GHz) as the active device operates reliably in the frequency range 2.0-3.2GHz. Using a better transistor like the BFP420 (ft=25GHz) allows to shift the frequency range up to 2.6-3.8GHz.


A VCO with a contiguous frequency coverage of 1200MHz may look as the upper limit for BB833 varactors (minimum capacitance 0.75pF, series resistance 1.8ohm). Once again, even better results can only be obtained by changing our way of thinking. Varactors are usually considered as discrete lumped components while our oscillator is built with microwave transmission lines with distributed parameters.

In order to shift the frequency of a microstrip filter, one should preferably change the phase velocity on the microstrip transmission lines. The phase velocity of a microstrip line can be made variable if the line is periodically loaded with discrete variable reactances (varactors). Therefore, several varactors have to be distributed along the transmission lines to obtain the widest possible frequency coverage of a microstrip VCO. Since silicon varactors are inexpensive, 6 or even 9 diodes can be used in a single VCO circuit.

The circuit diagram of a wide-band, low-noise VCO is shown on Fig.6. Each microstrip resonator is tuned by two BB833 varactors, installed in different positions along the same resonator. This circuit configuration allows a very wide tuning range of about 2GHz around a center frequency of about 3GHz. In other words, the tuning range of the described varactor VCO matches the tuning range of YIG VCOs.

Fig.6 - Wide-band low-noise VCO.



On the other hand, a large number of varactors also increases the feedback losses. Further losses are introduced by the complicated varactor bias network (nine 22kohm resistors). All these losses must be recovered with the gain of the active device. A high ft transistor like the BFP420 is required for such an oscillator. Maybe the described oscillator could also be modified for other active devices, like GaAs FETs, HEMTs, HBTs o MMICs.

Just like its narrow-band counterpart the circuit on Fig.6 includes an output buffer (another BFP420 and an ATF35176 HEMT) to isolate the oscillator from load variations. The buffer frequency response is selected to partially compensate the output level variation across the octave frequency band. Some supply voltage and tuning voltage filtering is also included. An output coupler provides an auxiliary output, for example to feed a PLL prescaler, a frequency counter or a tracking generator for the spectrum analyzer.

The measured tuning characteristics of three wide-band VCOs are shown on Fig.7 and Fig.8. The main difference between VCO#1 and VCO#2 is in the printed-circuit board. VCO#2 uses lower impedance lines and the coupling between resonators is weaker. There are also differences in the BB833 varactors used: VCO#1 has rather old BB833s providing a coverage of only 1.8GHz while VCO#2 has new BB833s providing a coverage of 2GHz. Finally, VCO#3 is built on the same microstrip board as VCO#1, but uses the new BB857 varactors resulting in a frequency coverage of 2.2GHz.

Fig.7 - VCO tuning characteristics table.




Fig.8 - VCO tuning characteristics diagram.

The tuning curves of all three VCOs are monotonic without kinks or jumps. The curves are quite non-linear as shown on Fig.8. All three curves are the steepest at band center around 3GHz (or at tuning voltages in the 5-10V range). Both below and above the tuning slope decreases, falling at the upper end (30V) down to less than 1/10 of the maximum slope. The frequency coverage of all three VCOs can be extended on the lower end by about 50MHz by applying a small negative voltage (about -0.7V) to the varactors.

The phase noise was accurately measured for VCO#1 and VCO#2 and the averaged results are shown on Fig.9. The phase noise is about 5dB worse at band center than at band edges. The phase noise peak coincides with the maximum tuning slope, indicating that at least part of this noise is caused by the varactors used and in particular by the 22kohm bias resistors.

Fig.9 - VCO phase noise.

In order to decrease the phase noise, the 22kohm bias resistors should be replaced by suitable RF chokes. Unfortunately suitable chokes are not easy to find. The resistor values can not be decreased much without introducing additional insertion loss in the feedback network. Finally, a more sophisticated bias regulator for the BFP420 oscillator transistor should also bring some improvement in the phase noise.

5. Practical microstrip VCO construction

Both the narrow-band and the wide-band VCOs are built as microstrip circuits on conventional, double-sided, 0.8mm thick glassfiber-epoxy FR4 laminate. The upper sides are shown on Fig.10 while the bottom sides are not etched to act as groundplanes. Although the FR4 is quite lossy at microwave frequencies, the losses in the BB833 or BB857 varactors are even higher, so using inexpensive and easy-to-handle FR4 laminate is not a drawback.

Fig.10 - Printed-circuit boards (150dpi).

The dielectric constant and losses of 0.8mm thick FR4 laminate were found quite similar even for materials obtained from different suppliers. One should only be careful about the thickness, since the same microstrip VCO circuit may not oscillate at all on 0.7mm thick laminate (too weak coupling between microstrip lines) or cover a restricted frequency band on 0.9mm thick laminate (coupling too strong). An additional problem is the large temperature coefficient of the FR4 laminate, shifting the frequency of the VCO downwards with increasing temperature.

All three printed-circuit boards have the same dimensions: 20mm (width) X 80mm (length). The whole surface of the groundplane may be tinned, while one should try to avoid tin-plating the microstrip lines on the upper surface except for the areas where SMD components are installed. The boards do not have plated-through holes. All ground connections are made through 2.5mm diameter holes. The latter are first covered on the ground-plane side with thin (0.1mm) tinned copper foil and then filled with solder. The advantages of this grounding method are a low inductance to ground and easy removal of installed components, without damaging the board nor the component to be removed.

Supply and tuning voltages go trough several feed-through capacitors. Some feed-through capacitors (3 or 4) are installed in the printed-circuit board itself in 3.2mm diameter holes drilled at the marked positions. Wire-leaded 1/8W resistors are used to connect the feed-through capacitors to the microstrip circuit. Electrolytic capacitors, 220uH chokes and other filtering components are installed on the groundplane side and are supported directly by the feed-through capacitors.

Finally, two feed-through capacitors are also installed in the 0.5mm thick brass walls of the box housing the VCO. The box should be 30mm high, extending 20mm above the board surface and some 9mm below the board bottom. If the printed-circuit board is well soldered on all four sides to the brass frame, then no additional shielding is required on the bottom side with supply filtering components. On the upper side a shielding cover is required and some microwave absorber is recommended.

In the frequency range of interest only teflon-insulated coax cables should be used. Both flexible and semirigid teflon cables can be used. The cable end should be first prepared by tinning both the inner conductor and the shield. Then the inner conductor should reach the microstrip board through a 3.2mm diameter hole in the brass walls, while the shield is soldered all around the perimeter of the hole.

The microstrip boards should be checked before installation in the shielding enclosure. Both narrow-band and wide-band VCOs should be checked for frequency coverage and output power level. In particular the PCB#2 for the wide-band VCO may require some trimming of the central resonator length if the power drops or the oscillator stops at low tuning voltages. The output power of the buffer amplifier should not drop below +10dBm in any of the oscillators shown.

6. Applications of microwave varactor-tuned VCOs

Microwave VCOs have many applications. Narrow-band VCOs are suitable for many frequency-synthesizer projects. Wide-band VCOs are mainly intended for instrumentation, like CW and sweep generators, spectrum analyzers and corresponding tracking generators. In the following paragraph a simple spectrum analyzer using both described VCOs will be presented, although for space reasons the description will be limited to the block diagram as shown on Fig.11.

Fig.11 - Spectrum-analyzer block diagram.

The design of a spectrum analyzer is based mainly on the available wide-band VCO. Considering the frequency coverage of the described VCOs and the amateur microwave allocations (1.3GHz and 2.3GHz) and satellite frequency bands (GPS at 1.575GHz, weather satellites at 1.7GHz) it makes sense to select the first IF at 2.1GHz. This is far enough from 2.3GHz and allows a simple input low-pass filter for the band from 0 to 1750MHz.

On the other hand, the last IF is set down to 10MHz (10.7MHz) due to the design restrictions of LC and crystal filters and the logarithmic detector. To simplify image and spurious filtering, the first IF at 2.1GHz is first converted to a second IF of 70MHz and the latter is afterwards converted to the final third IF of 10MHz.

The IF filters offer six different bandwidths: 4MHz, 700kHz, 150kHz, 50kHz, 20kHz and 10kHz. The 4MHz bandwidth is required for full-band sweeps, considering the limited resolution of the CRT display. On the other side, the resolution of the spectrum analyzer is limited to 10kHz bandwidth and 50kHz/div display. Narrower IF filters would require additional stabilization circuits for all oscillators. Since narrow IF filters also require very slow scanning, they are practically seldom used and were omitted in this project.

The logarithmic detector is built with discrete components, since available integrated circuits will not handle a dynamic range of over 90dB (achieved already with the 150kHz IF filter) and many different IF bandwidths at the same time. The following video amplifier provides a 20dB/V output to drive the vertical deflection of the CRT display as well as an AF (earphone) output. Of course the latter is only useful at zero span.

On the other hand, the horizontal deflection signal (5Vpp) is provided by the built-in sawtooth oscillator. The latter may sweep the frequency of the first local oscillator (wide-band VCO) or the frequency of the second local oscillator (narrow-band VCO for 500kHz/div or less). Due to the nonlinear voltage/frequency response, the wide-band VCO requires a rather complex linearization circuit.


The spectrum analyzer is designed to use a standard XY oscilloscope display. If an external X deflection is not available, the spectrum-analyzer electronics also provides trigger pulses for the time base of the oscilloscope display. Z-axis or display blanking has two functions in a spectrum analyzer: retrace blanking and out-of-band blanking. If a Z-axis (intensity) input is not available on the CRT display, then the blanking output (open collector) may be wired in parallel with the Y output to deflect the trace outside of the visible screen when the blanking is active.

Original full-size drawings

Spectrum Analyzer


1. Spectrum-analyzer design

Wide-band (panoramic) receivers and RF spectrum analyzers are usually designed as multiple-conversion receivers with the first IF above the maximum input-signal frequency. Amateur designs are mainly limited by the performance of the VCO used for the first frequency conversion. The 2-4GHz VCO presented in [1] and [2] allows the design of a simple spectrum analyzer covering the frequency range 0...1750MHz in a single span.

The resulting spectrum-analyzer block diagram, shown on Fig.1, was already briefly described in [1] and [2], although only the microwave VCOs were described in detail. In this article the remaining building blocks of the above-mentioned spectrum analyzer will be described in detail, as well as the overall assembly and tuning of the completed instrument.

Fig.1 - Spectrum-analyzer block diagram.

As already mentioned in [1] and [2], the described spectrum analyzer is a triple-conversion receiver with the corresponding IFs around 2.1GHz, 70MHz and 10MHz. Since both the first LO and the second LO are VCOs, the first IF may be made variable. This may be useful to shift some spurious responses of the first mixer in some difficult measurements.

The described spectrum analyzer is designed to use a standard XY oscilloscope display. Additional control signals are provided to drive different displays as well as frame memories, storage normalizers, marker generators and/or trigger frequency counters. The outputs of both VCOs are also made available to drive a tracking generator or a frequency counter. Some of these additional circuits will be described in future articles.

2. Input attenuator

The input attenuator is a simple yet important part of a spectrum analyzer. The main function of the input attenuator is reducing the signal level to avoid overdriving the first mixer. The input attenuator should therefore only include very linear components like resistors and mechanical switches.

The basic attenuator circuit is a simple "PI" or "T" resistor network. The resistor values are selected both for the desired attenuation value and input/output impedance matching. Since a "PI" or "T" network contains three independent resistors, all three quantities can be adjusted independently within certain limits: attenuation, input impedance and output impedance.

Generally speaking, attenuators can be built for any frequency provided that the resistors (and switches) are sufficiently small with respect to the wavelength and/or are designed as parts of transmission lines with carefully controlled characteristic impedances. Professional attenuators, both fixed and adjustable, are available for frequencies up to 18GHz and beyond, depending mainly on the coaxial connectors.

On the other hand, it is much more difficult to build good microwave attenuators from standard electronic components. Resistors with wire leads are only useful up to about 500MHz. SMD resistors are much better due to their smaller size and can be used up to at least 5GHz. Finding suitable mechanical switches for microwave frequencies is even more difficult.

The circuit diagram of the attenuator, shown on Fig.2, therefore can not tell much about the microwave performance of the circuit. The attenuator includes four identical "PI" networks and four DPDT switches. This design allows a nominal attenuation of 0dB to -40dB in 10dB steps.

Fig.2 - Attenuator.



Standard SMD resistors of the size 0805 or smaller are certainly good enough for a 2GHz spectrum analyzer. It is much more difficult to find suitable DPDT switches with low parasitics. In the prototypes standard miniature DPDT toggle switches were used. The resulting frequency response was compensated with 1pF SMD capacitors soldered across the 68ohm resistors. Finally, the toggle switches were installed between two small printed-circuit boards to keep their impedance closer to 50ohms (to be described later in this article).

A careful construction together with some compensation capacitors allows to keep the frequency response of the attenuator within +/-1dB for frequencies up to 1GHz and within +/-2dB for frequencies up to 2GHz from the nominal attenuation value. Since the maximum power dissipation of SMD resistors is 1/8W, the maximum input power to the described attenuator is about 250mW (+24dBm). At the maximum attenuation value of -40dB this means -16dBm at the first mixer input.

Of course additional (external) high-power attenuators or couplers are required for transmitter measurements. The input attenuator of a spectrum analyzer is only intended for fine adjustments of the input signal level and/or finding the source of some signals: real signals or unwanted mixing products in the spectrum analyzer?

Finally, the described attenuator does not include any protection against DC voltages that may be present in some RF circuits!

3. First mixer

The dynamic range of any receiver depends mainly on the performance of the first mixer. The low end of the dynamic range is defined by the mixer noise figure (insertion loss) while the high end of the dynamic range is defined by the mixer distortion (intermodulation). Further, the first mixer of a spectrum analyzer may be easily destroyed by high RF levels, DC voltages and even static discharges on the input connector.

The first mixer of a spectrum analyzer should therefore both provide the best possible dynamic range and at the same time allow quick and inexpensive repair in the case of damage. The latter requirement makes commercially-available, doubly-balanced mixers with schottky quads and ferrite balancing transformers unpractical.

The described spectrum analyzer is therefore using an inexpensive double schottky diode BAT14-099 in the first mixer. The balun for the local oscillator is built from UT-085 semirigid cable, as shown on Fig.3. A completely symmetrical construction allows a balancing of at least 30dB without any tuning and more than 45dB by adding small drops of solder in the circuit. Of course the symmetry of commercially-built mixers can not be improved since the latter are built in hermetically-sealed packages.

Fig.3 - First mixer.



A good mixer symmetry is required for many reasons. A balanced mixer prevents the oscillator noise from getting directly in the first IF. Further, some unwanted mixing products are suppressed in a balanced mixer. In the case of a spectrum analyzer it is especially important to suppress second-order distortion products, so that the spectrum analyzer can be used to accurately measure the suppression of the second harmonic in radio transmitters.

The first mixer module includes a microstrip low-pass filter with the cutoff frequency of about 1.75GHz. The latter should suppress unwanted responses of the spectrum analyzer in the microwave frequency range. Its insertion loss amounts to about 45dB at the first IF around 2.1GHz and becomes even higher at higher frequencies. The low-pass is followed by a RC network to slightly compensate the slow decay of the sensitivity at frequencies above 1GHz.

The mixer diode BAT14-099 is followed by several components to provide impedance matching and suppression of unwanted resonances. For example, the open end of the UT-085 balun is also used as a coupling capacitor. The two 51ohm resistors and corresponding quarter-wavelength lines provide a termination for the image frequency and other unwanted mixing products, reflected by the 2.8GHz low-pass and the following cavity bandpass filter.

4. Cavity bandpass filter

In all wide-band receivers, the first mixer should be followed by the best possible bandpass filter. In HF receivers covering 0...30MHz, a 15kHz wide crystal filter is usually used in the first IF around 45MHz or 70MHz. Considering the first IF the described spectrum analyzer at 2.1GHz, a cavity bandpass filter is the only technology that provides both high selectivity and low insertion loss.

Microwave cavities are electrically simple to describe, but usually require lots of precision mechanical work and special tools for manufacturing. Several efforts were spent in finding a reproducible cavity-filter design, made from standard materials using only simple mechanical tools.

The described bandpass filter is built inside a piece of standard aluminum tube of rectangular cross-section with the external dimensions of 40mmX20mm and 2mm wall thickness. Such rectangular aluminum tube can be found elsewhere in Europe. Of course, its internal dimensions of 16mmX36mm are the most important parameter while building a cavity filter.

The construction of the cavity bandpass filter is shown on Fig.4. The filter includes five quarter-wavelength resonators made from 8mm diameter aluminum rod. All five resonators are oriented in the same direction ("comb" filter) to decrease the coupling between adjacent resonators. In this way the overall dimensions of the filter are smaller than in the case of an "interdigital" arrangement of the resonators.

Fig.4 - Cavity BPF.



The input and output couplings are made by two small rod antennas, supported by the corresponding SMA connectors. The coupling is adjusted by the length (around 27mm) of the two antennas made of thin copper tube (UT-085 shield). The coupling between resonators is defined by the distance 25mm between resonator centers and sets the filter bandwidth to about 25MHz. Five M3X20mm tuning screws are used to bring all five resonators to the desired operating frequency. The tuning screws are inserted from the opposite narrow side of the cavity and secured with a counternut after tuning.

The cross-section of the cavity is small enough that the electromagnetic field exhibits a very fast exponential decay at both ends of the rectangular aluminum tube. Covers are therefore not required for the electrical performance of the filter. On the other hand, covers are useful to keep dust and dirt outside. Covers may extend up to 10mm inside the cavity or stay at least 25mm away from the coupling antennas without having any influence on the filter performance.

The described cavity bandpass filter provides over -100dB of suppression for the second-conversion image frequency around 1.94GHz. The insertion loss is only around -2dB at the nominal first IF frequency of 2.1GHz and over -100dB outside the passband anywhere between 0 and 4GHz. Spurious higher-order resonances appear above 4GHz, when the aluminum tube starts operating as a waveguide. A cavity filter alone is therefore not sufficient. Additional microstrip low-pass filters are therefore included in both the first and second mixer modules to suppress the spurious cavity responses above 4GHz.

The described cavity design allows narrowing the passband down to just a few MHz. A bandwidth of 25MHz was selected to allow a narrow sweep of the second LO, to avoid some spurios responses of the first mixer and finally to allow for some frequency drift of the second VCO.

5. Second mixer

The requirements for the second mixer are not as far as severe as for the first mixer, since most unwanted signals have already been removed by the cavity bandpass filter. Also the signal levels are about -10dB weaker due to the conversion loss of the first mixer and cavity bandpass insertion loss. Calculations and experiments show that no amplifier stages are required between the two mixers if the maximum dynamic range is desired.

The design of the second mixer is very similar to the first mixer, except that the input and output are interchanged, as shown on Fig.5. The second mixer is also using a BAT14-099 double schottky diode and a balun made from UT-085 semirigid cable. Since the second mixer operates in a narrow frequency band, damping resistors and other compensation components are not required.

Fig.5 - Second mixer.



The second mixer module includes low-pass filters both at the input and output. The input low-pass filter cuts above 2.8GHz to suppress the spurious cavity responses above 4GHz. The output low-pass cuts above about 800MHz to suppress unwanted mixing products and feed-through of the LO signal.

6. Third mixer

The first IF of the spectrum analyzer around 2.1GHz is far too high for the different IF filters and logarithmic detector. A more suitable choice for the final IF is 10MHz. The latter can be conveniently reached from 2.1GHz in two down-conversion steps. Due to the relatively low frequencies and low signal levels, the requirements for the third mixer are not particularly severe. An additional requirement is the maximum bandwidth B=4MHz that requires a carefully designed bandpass filter at 70MHz and wide-band impedance matching at the final IF of 10MHz.

The circuit diagram of the third mixer and related components is shown on Fig.6. The circuit includes a low-noise amplifier at 70MHz, followed by a LC bandpass filter for 70MHz and a dual-gate MOSFET mixer. The 70MHz low-noise amplifier (BF998) is the only true amplifier stage in the whole receiving chain of the spectrum analyzer. The only purpose of this stage is to compensate for the conversion loss in the mixers. Any gain increase or additional amplifier stages would just impair the dynamic range of the spectrum analyzer.

Fig.6 - Third mixer.



The LC bandpass filter at 70MHz has two functions. First, the image response of the third mixer at 50MHz has to be suppressed. Second, the widest IF bandwidth of the spectrum analyzer is defined mainly by the 70MHz bandpass filter. The bandwidth of the 70MHz LC bandpass filter itself is around 5MHz, limiting the overall bandwidth of the complete receiving chain to about 4MHz. The 70MHz LC filter is built with adjustable coils (about 500nH) wound on shielded supports for IF transformers. The input and output are terminated with 1.5kohm resistors.

The third mixer is built with a dual-gate MOSFET BF981. The input 70MHz signal is fed to the first gate while the 60MHz LO is applied to the second gate. The mixer is followed by a low-pass impedance-matching network. The latter should both remove the 60MHz LO signal and other unwanted mixing products as well as provide a wide-band transformation of the MOSFET high output impedance down to 50ohms.

The design of a suitable low-pass/matching network is complicated, since the required bandwidth is comparable to the center frequency 10MHz. Impedance matching is therefore performed in several steps with low-pass LC networks. The cicuit shown on Fig.6 allows reasonable impedance matching in the frequency band 6...15MHz and a high suppression of the 60MHz LO at the same time. The circuit is built with fixed inductors of the size and shape of 1/4W or 1/2W resistors.

The third LO includes an overtone crystal for 60MHz and a BF199 transistor. The 1.5uH inductor in the emitter of the BF199 forces the crystal to oscillate on the third overtone. The crystal is also used at the same time as a filter for the output signal fed to the mixer. The 1uH inductor in series with the crystal further reduces the amount of harmonics fed to the mixer. Finally, unwanted mixing products can be further suppressed by carefully setting the MOSFET bias with the trimmer "LINEARITY". The trimmer "GAIN" is set for the lowest practical gain of the BF998 that does not impair the noise figure of the whole spectrum analyzer.

7. LC filters

The output of the third mixer can be fed directly to the input of the logarithmic detector, setting the IF bandwidth to about 4MHz. If a narrower IF bandwidth is desired, additional bandpass filters are required between the third mixer and the logarithmic detector. LC filters can be used for bandwidths above 100kHz, while crystal filters are required for even narrower bandwidths.

A spectrum analyzer should include several different IF filters, to be included as required in the IF chain. Filter switching can hardly be performed by standard mechanical switches only, since a crosstalk better than -100dB is required between the input and output of a filter. A better solution is electronic filter switching with PIN diodes. Each filter should also include an amplifier to compensate for the insertion loss of the filter. In this way the measured signal strength will remain the same while switching among different IF filters.

Professional spectrum analyzers usually allow the selection of the IF bandwidth in steps of 1/3/10 etc. On the other hand, practical requirements show that just a few different bandwidths are required above 100kHz. Many different bandwidths are only required below 100kHz, where the IF bandwidth also defines the sweep time and limits the display update frequency.

The described spectrum analyzer includes two different LC filters with the bandwidths set to 700kHz and 150kHz. In addition, the crystal filter bandwidth can be adjusted in smaller steps to 50kHz, 20kHz or 10kHz. Finally, with no additional filters an IF bandwidth of 4MHz is obtained, providing a total selection of six possible IF bandwidths.

The two additional LC filters are shown on Fig.7. Each filter includes a switching network with four diodes BA423. The filter is inserted in the IF chain by simply applying the corresponding +8V supply. At the same time the BF199 transistor cuts the direct signal path and further improves the crosstalk attenuation between the input and output of the filter.

Fig.7 - LC filters.



The 700kHz wide LC filter includes four adjustable coils (about 10uH), wound on shielded supports for IF transformers. The BFR96 amplifier compensates the insertion loss of this filter. Of course the exact amount of gain depends on the loss in the coils and can be adjusted with the 220ohm trimmer.

The 150kHz wide LC filter includes two separate filters, each including two tuned circuits. The required inductances are in the range of 2.2uH. All four coils are wound on slightly larger shielded supports for IF transformers to achieve an unloaded Q of about 100. The insertion loss of both narrow LC filters is compensated by the BF981 amplifier. The gain of the latter is adjusted by the bias voltage (10kohm trimmer) on the second gate.

The LC filter module requires three different supply voltages. The +8V supply should be present at all times to bypass the filters while the latter are turned off. The +8V/700kHz supply inserts the 700kHz wide filter while the +8V/150kHz supply inserts the 150kHz wide filter. Both the input and output of the module should remain matched to an impedance of 50ohms at all times.

8. Crystal filter

Spectrum analyzers require somewhat different IF filters than those installed in communication receivers. Communication receivers usually require filters with a flat passband, to avoid modulation distortion, and a very steep increase of the insertion loss immediately outside the useful passband, to reject adjacent channels. Such filters are not suitable for spectrum analyzers, since their time response is rather slow (ringing!) compared to the filter bandwidth.

A slow filter response and/or ringing is especially harmful at small bandwidths, where the time response of the filter defines the sweep time and display update period. Commercial crystal and ceramic filters are therefore almost useless in spectrum analyzers. A suitable crystal filter or set of different filters has to be specially built for a spectrum analyzer.

A spectrum-analyzer IF filter should have a "triangular" frequency response with a sharp peak and smoothly and symmetrically increasing attenuation outside the passband. In practice this requires under-critically-coupled resonators or better a series connection of several single-resonator filters and buffer amplifiers, to avoid any interaction among the resonators.

The crystal filter shown on Fig.8 includes a series connection of four independent, single resonator filters. BF199 emitter followers are used to avoid unwanted coupling among the crystals. Each individual filter includes a single crystal in a balanced network. The capacitive trimmer is used to compensate the capacitance of the crystal. The 10uH center-tap coil resonates with the 68pF capacitor and compensates the remaining parasitic capacitances of the circuit.

Fig.8 - Crystal filter.



The bandwidth of a single-crystal filter depends mainly on the source and load impedances. The source impedance is kept low by the previous emitter follower. The load impedance is adjustable, since a PIN diode BA596 is connected in parallel to the input of the following emitter follower. The filter bandwidth is therefore adjustable with the DC current IBXTAL fed to the four BA596 PIN diodes.

The crystal filter module includes a switch with four diodes BA423 to insert the filter in the IF chain. The supply voltage +8V should be present at all times to bypass the crystal filter while the latter is turned off. The supply voltage +8V/XTAL inserts the crystal filter. The insertion loss of the latter is mainly compensated by the emitter followers. Some additional gain is provided by the 1:5 step-up transformer at the input, followed by the first emitter follower. The overall gain is adjusted by the 470ohm trimmer on the output.

9. Logarithmic detector and video amplifier

A spectrum analyzer differs from a communication receiver also in the type of detector used. Spectrum analyzers use a logarithmic amplitude display with a very wide span of 80...100dB while communication receivers use linear detectors. A spectrum-analyzer display is therefore more similar to the operation of the S-meter of a communication receiver.

Some inexpensive commercial spectrum analyzers (Hameg) and many amateur designs in fact use the S-meter output of popular FM-demodulator chips as a logarithmic detector. This is a very poor technical solution, since the S-meter output of most FM-demodulator chips is very inaccurate and has wide deviations from the ideal logarithmic curve. Further, the dynamic range of most single-chip logarithmic detectors hardly exceeds 70dB. Chips with a wider dynamic range require interstage bandpass filters to limit the wide-band noise.

A logarithmic detector for a serious instrument therefore has to be built from discrete transistors, resistors and capacitors as shown on Fig.9. The circuit includes a series connection of ten identical stages, operating as limiting amplifiers and as linear detectors at the same time. Each stage includes a balanced amplifier, allowing a very high overall gain without instabilities and a precisely-defined saturation mechanism. The logarithmic response is obtained as a sum of linear responses that saturate above a certain level, provided that the gain of each single stage does not exceed 10...12dB.

Fig.9 - Logarithmic detector.



Similar balanced amplifier chains are used as FM limiters in all known integrated circuits. The main difference between the FM chips and the discrete circuit on Fig.9 are the 470pF emitter-coupling capacitors. The latter can not be built inside a monolithic integrated circuit. The emitter-coupling capacitors allow a simple setting of the bias point and center frequency of the logarithmic detector.

Since the described logarithmic detector does not amplify DC or very low frequencies, "1/f" noise can be avoided resulting in an up to 20dB increase of the dynamic range. The noise level of the described detector is around -105dBm, while the logarithmic response is acceptable up to -10dBm, resulting in a dynamic range of at least 95dB. Another advantage of the described circuit is that the output does not saturate but becomes linear for an additional increase of the input signal level of 10...15dB beyond the nominal logarithmic response. The latter is very useful to clearly indicate excessive signal levels on a spectrum-analyzer display.

A drawback of the described circuit is a rather low output voltage, only around 100mV difference for a dynamic range of almost 100dB. The output video signal is therefore related to a reference voltage provided by a dummy BF199 amplifier, operated at the same current as the remaining twenty BF199s used in real amplifier stages. Without this compensation the output voltage would change by 2mV (meaning 2dB!) for each degree of temperature change.

The described logarithmic detector requires a video amplifier to boost the output signal level. A suitable circuit is shown on Fig.10. The overall voltage gain amounts to about 50 and is distributed among three operational amplifiers to improve the video bandwidth. The fourth operational amplifier from the MC33074 boosts the reference. The MC33074 includes four relatively fast operational amplifiers (unity-gain bandwidth 4MHz), so that the bandwidth of the overall video amplifier is about 500kHz.

Fig.10 - Video amplifier.



The video amplifier includes two trimmers to adjust the offset (DC component) and gain. The gain should be set for an output of 20dB/V. The output circuit includes a simple video low-pass filter to limit the video bandwidth to 20kHz or 1kHz. An additional audio output is provided to drive earphones with the emitter follower BD139. The video amplifier module also includes a uA723 regulator to supply the logarithmic detector.

10. Sawtooth generator and linearization

Almost any oscilloscope can be used as a display for the described spectrum analyzer. Although most oscilloscopes have built-in a sawtooth generator for the horizontal deflection, the generated sawtooth is usually an internal signal inside the oscilloscope and is not made available externally. A spectrum analyzer should therefore include its own sawtooth generator to drive the oscilloscope in the XY mode or synchronize the internal sawtooth generator in the oscilloscope.

The spectrum-analyzer sawtooth generator is shown on Fig.11. The shortest sweep time is set to about 20ms or in other words corresponding to a 2ms/div sweep on the oscilloscope. The sweep time can be increased up to 20 times with the 220klog potentiometer. On the other hand, sweep times shorter than 20ms usually make no sense in a RF spectrum analyzer.

Fig.11 - Sawtooth generator.



The sawtooth generator is built with a quad operational amplifier MC33074. The MC33074 performs different functions: constant-current source, sawtooth oscillator and two output amplifiers. The circuit provides two sawtooth signals for the VCO sweep and display sweep as well as trigger pulses. The supply voltage is stabilized to 30V with the integrated regulator uA723 and BD139 power transistor.

The sawtooth signal may sweep the first VCO (wide scans), the second VCO (narrow scans) or none of them in the zero-span mode. The sawtooth amplitude defines the span width. Besides the span width also the center frequencies of both VCOs have to be set to the desired values.

The described functions require several switches, selectors and potentiometers on the front panel of the spectrum analyzer. Their wiring is shown on Fig.12 together with the linearization circuit for the wide-band VCO. The center frequencies of both VCOs are set with 10-turn Helipots. The span width is set with a selector in steps 1/2/5/10 etc. At 500kHz/div and narrower spans the second VCO is swept, while the frequency of the first VCO can be additionally stabilized with the STAB VCO#1 switch. Finally, the zero-span function is activated by yet another switch that removes the sawtooth and connects 47uF capacitors in parallel with the control inputs of both VCOs.

Fig.12 - Linearization.



As already described in [1] and [2], the frequency response of the wide-band VCO with six BB833 varactors is quite nonlinear. The tuning slope is the steepest around the center frequency of the VCO coverage, where the tuning slope reaches 120MHz/V at a tuning voltage around 7V. At lower voltages the tuning slope decays slowly to about 90MHz/V on the lower end. The tuning slope is much more nonlinear in the upper part, where the tuning slope may decay below 10MHz/V at the maximum frequency.

The linearizer circuit includes two operational amplifiers. A negative feedback defines the minimum gain at an output voltage of about 7V. At lower or higher output voltages some positive feedback is switched in through resistive dividers and diodes. The positive feedback increases the gain of the amplifier to compensate for the decay of the tuning slope of the wide-band VCO.

A linearization of the narrow-band VCO is not required, since the latter is only swept over a small fraction of its frequency coverage. A single operational amplifier is required to drive the narrow-band VCO, since the polarity of the sawtooth has to be inverted considering the frequency conversions used in the spectrum analyzer. For the same reason it makes sense to interchange the connections to the Helipot that sets the center frequency of the narrow-band VCO (+30V on the beginning and ground on the end of the Helipot scale).

The blanking generator includes a quad voltage comparator LM339. Two comparators are used just as amplifiers for the trigger pulses to enable blanking during retrace. The other two comparators check the control voltage of the wide-band VCO, turning the display off when the VCO is tuned out of range. The LM339 comparators have open-collector outputs and the final blanking signal is obtained from a wired-or connection of three LM339 outputs.

The linearizer (another MC33074) and blanking (LM339) circuits receive the same +30V supply as the sawtooth generator. The +7V reference voltage is obtained directly from the uA723 regulator on the same printed-circuit board.

11. Power supply

The described spectrum analyzer is designed for a nominal DC supply voltage of 12VDC, negative grounded, with reasonable tolerances (10...15V). Of course, internally the spectrum analyzer requires many different supply voltages. There are also differences in the stability requirements. Some modules require very stable and well-filtered supply voltages while others are more tolerant.

Varactor diodes inside both VCOs require very stable voltages up to about +30V. Therefore the corresponding drivers require a very stable +30V supply. Considering the voltage drop in the uA723 regulator and BD139, a DC/DC converter with an output voltage of about +37V is required, as shown on Fig.13. The DC/DC converter includes a power oscillator with the transistors BC308 and BD139. Voltage spikes from the 100uH inductor are rectified by the 1N4148 diode to charge the 47uF electrolytic capacitor. When the voltage on the latter exceeds +37V, the two 18V zener diodes turn on the BC548 to stop the oscillator.

Fig.13 - DC/DC converter.



Of course, a DC/DC converter is a potential noise source to be built inside a sensitive piece of test equipment. Both input and output of the DC/DC converter are well filtered with the 39uH and 1mH chokes as well as several electrolytic capacitors. The DC/DC converter is further built on its own printed-circuit board to be installed far away from the more sensitive circuits of the spectrum analyzer.

The power supply of the remaining circuits of the spectrum analyzer is shown on Fig.14. The input voltage +12V is first filtered with the VK200 choke and then fed to the DC/DC converter, to the video amplifier and to the 7808 regulator. The latter provides a stabilized +8V supply for most circuits of the spectrum analyzer. The 7808 regulator uses the baseplate of the spectrum analyzer as a heatsink and is installed together with the VK200, 470uF electrolytic capacitors and protection diode close to the +12V supply connector.

Fig.14 - Power supply.


sa8v    sa8v

Fig.14 also shows the IF-bandwidth switching. The latter requires a two-section, six-way selector. One section of the selector is used to switch the +8V supply voltage to different filters. The two 1N5818 schottky diodes keep previous filters inserted while new filters are added in the IF chain. The second section of the selector sets the current in the PIN diodes that define the bandwidth of the crystal filter.

12. Construction tips

A spectrum analyzer is a sensitive receiver operating over a wide frequency range. The circuits of a spectrum analyzer therefore require considerably more and better shielding when compared to conventional communication equipment. Better shielding also requires a larger number of shielded modules that contain just a few electronic components each. Besides shielding, microwave absorbers are required outside and inside some shielded enclosures.

The spectrum-analyzer modules can be roughly divided in three groups: (1) microwave modules built as microstrip circuits inside shielded boxes, (2) IF modules built on single-sided printed-circuit boards inside shielded boxes and (3) video/supply modules that do not require special shielding.

The microstrip printed-circuit boards (except the VCOs published in [1] and [2]) are shown on Fig.15. All microstrip circuits are etched on double-sided, 0.8mm-thick FR4 laminate. Only the upper side is shown on Fig.15, since the bottom side is not etched to act as a groundplane.

Fig.15 - Microstrip circuit boards (150dpi).

Only SMD components of the size 0805 or smaller should be installed on the microstrip circuit boards. The SMD components are grounded through 2.5mm diameter holes, covered with 0.1mm-thick copper foil on the groundplane side and filled with solder before installing the SMD components. The completed microstrip boards are soldered in brass frames and covered with a shielding cover. A piece of 1cm-thick microwave absorber foam should be installed under the whole surface of the shielding cover to suppress parasitic resonances of the shielding enclosure.


The IF printed-circuit boards, shown on Fig.16, are all etched on single-sided, 0.8mm-thick FR4 laminate. SMD components should be soldered first: resistors, capacitors and semiconductors. The printed-circuit boards are designed for SMD-component sizes 0805 or 1206. Inductors, IF transformers, trimmers, BF199 transistors, BA423 diodes, crystals and electrolytic capacitors are installed as conventional "through-hole" components from the other side.

Fig.16 - IF circuit boards (150dpi).

The IF printed-circuit boards are also soldered in brass frames just like the microstrip boards. Of course, the IF modules do not require any microwave absorber foam under the cover nor any special shielding on the bottom. The supply voltages go through feed-through capacitors soldered in the narrow sides of the brass frames while the IF cable shields are soldered directly to the brass frames.

The video/supply printed-circuit boards, shown on Fig.17, are all etched on single-sided, 1.6mm-thick FR4 laminate. All of the components are conventional types with wire leads for "through-hole" installation. The resistors and diodes are installed vertically on the sawtooth and linearization module to save some space. The video/supply modules do not require any shielding and are installed on the baseplate with four M3 screws in the corners. All connections go through simple connectors made from quality IC sockets.

Fig.17 - Video/supply circuit boards (150dpi).

The shielded enclosures of the RF/IF modules are shown on Fig.18. The input attenuator includes four miniature DPDT toggle switches installed in a sandwich between two microstrip boards. The microstrip boards are soldered to an "U"-shaped piece of 0.5mm-thick brass sheet. The four toggle switches are installed in 5mm diameter holes, while two 3.2mm diameter holes are provided for the input and output coaxial cable. The input attenuator does not require a shielding cover.

Fig.18 - Shielding of RF/IF modules.

Both microwave mixers and all four IF modules are installed in a brass frame with the dimensions 120mmX30mm, while the two VCOs are installed in similar frames with the dimensions 80mmX20mm. All frames are made from 0.5mm-thick brass sheet. Covers are also made from brass, although somewhat thinner brass sheet (0.4mm or even 0.3mm) is sufficient.

Both microwave mixers include baluns made from UT-085 semirigid cable. The shield of the latter is first tinned for a length of about 35mm to let expand the teflon dielectric. Then the shield is cut 17mm from the open end and pulled away to obtain a 1mm-wide gap. Such a prepared cable end is then soldered on the microstrip circuit board and the board is then soldered in the brass frame. The balun cable pops out of the frame for about 2mm on one side, while a longer cable (20-25mm) is available on the other side to connect the VCO.

The overall spectrum-analyzer module location is shown on Fig.19 inside an aluminum box with the following dimensions: width 220mm, depth 240mm and height 65mm. A central baseplate (1mm-thick aluminum) holds all of the modules and divides the useful volume in two sections, each 32mm high. The top section may require a piece of microwave absorber between the second and third mixers.

Fig.19 - Spectrum-analyzer module location.





While positioning the different modules on the baseplate, care should be taken that the mounting screws do not interfere with the modules on the other side. The cavity is hold in place by a pair of screws on each "cold" end, so that no metal parts enter the "hot" central part of the filter. The front panel is connected to the baseplate with the TNC connector, toggle switches and Helipots, so no additional screws are necessary. On the other hand, the back panel is connected with four M3 screws in addition to the SMA connectors.


Finally, it is convenient that all display connections are brought to a single multi-pole connector on the back panel. The suggested 6-pole, 270-degree DIN connector is shown on Fig.20. Different displays therefore only require special cables. The same display connector also carries the +12V supply for a storage/normalizer unit or marker generator.

Fig.20 - Display connector.



13. Alignment and checkout

Before starting the construction of a complex piece of test equipment like a spectrum analyzer, some simpler test equipment should be available as well as enough practical experience to efficiently use the latter. Besides a conventional analog AVO-meter (digital AVO-meters are useless for tuning), a grid-dip meter, a microwave frequency counter up to at least 4GHz and a microwave power sensor should be available. Last but not least, the oscilloscope to be used as the display of the spectrum analyzer should also be available.

The construction of the spectrum analyzer should start with the most difficult part: the VCOs described in [1] and [2]. Both VCOs have to be checked for the frequency coverage and output power. In fact it makes sense to build two wide-band VCOs. The best VCO is installed in the spectrum analyzer. The leftover VCO can still be used in other test equipment, like a harmonic converter that extends the coverage of a spectrum analyzer, to be described in a later article.

When both VCOs are operating correctly, the remaining modules can be built except the LC and crystal filters. The third mixer and logarithmic detector can be aligned and checked with a grid-dip meter. The operation of the sawtooth generator, linearization circuit and several selectors and potentiometers can be checked with the oscilloscope that will be used for display. All the trimmers in the linearization should be set to their "cold" ends.

The cavity bandpass filter is the unit that is most difficult to align without complex test equipment. The tuning screws are initially set about 1...2mm above the "hot" ends of the resonators. The narrow-band VCO is set to 2.03GHz with a frequency counter while the wide-band VCO is sweeping the whole frequency band. The cavity filter can then be tuned to the characteristic "DC peak" of all RF spectrum analyzers.

When the "DC peak" is found and the cavity bandpass is roughly tuned to the correct frequency, an initial checkout of the spectrum analyzer can be performed. A short antenna connected to the RF input will provide many peaks around 100MHz (FM broadcast) and 900MHz (GSM). Even better is a simple 100MHz test source built with a 100MHz ECL crystal oscillator, as shown on Fig.21.

Fig.21 - 100MHz test source.



The adjustment of the linearizer requires several signals on known frequencies. The simplest solution is a comb generator (shown on Fig.22) that provides many equally-spaced harmonics of the same fundamental frequency. The usual image on the display is rather dense lines between 700MHz and 1000MHz, while the spacing of the lines gradually increases both towards lower frequencies and even more towards higher frequencies. The eight trimmers in the linearizer are then adjusted to obtain equally spaced lines across the whole display. If the adjustment range of the trimmers is not sufficient, other resistors in the circuit may be changed to match the response of the varactors in the wide-band VCO.

Fig.22 - Comb generator up to 2GHz.



When the spectrum analyzer is operating correctly with a 4MHz IF bandwidth, additional LC and crystal filters can be built. Finding suitable filter crystals is not simple. Cheap "clock" or "baud-rate" crystals were found useless, since they have many spurious resonances close to the main resonant frequency. Such crystals may still provide some useful performance in narrow SSB filters, but their response is badly distorted in wide (B=50kHz) filters for spectrum analyzers.

The only reliable crystals are those found in old 35kHz-wide filters for 50kHz FM-channel spacing. These crystal filters were usually built for a center frequency of 10.7MHz. One filter contains four different crystal pairs. Therefore two identical crystal filters have to be disassembled to obtain enough crystals for four spectrum analyzers...

Both crystal and LC filters are simply adjusted on the "DC peak", when the remaining parts of the spectrum analyzer are already operating. The alignment of LC filters is straightforward and the desired response is obtained easily.

On the other hand, the alignment of the crystal filter is very demanding. Each filter stage has to be aligned independently while the remaining three crystals are shorted with 100ohm resistors. A symmetrical response of a single filter stage is sought with the capacitive trimmer. Next the coil is adjusted to obtain the broadest peak, of course with the current in the PIN diode turned off. This procedure has to be repeated several times for each stage, since the two adjustments interact.

After the LC and crystal filters are roughly adjusted, the two trimmers in the third mixer module can be adjusted. The mixer linearity is adjusted with a strong (around -10dBm) signal on the first mixer input. Setting the resolution bandwidth to 150kHz, some spurious responses may be visible +/-2MHz around the desired signal. These spurious responses should be minimized by adjusting the mixer linearity. Finally, the gain of the 70MHz amplifier is set so that the front-end noise just overrides the detector noise with the 150kHz IF filter.

At the end, there are many fine adjustments to be performed. All IF filters should be aligned to the same frequency and same insertion loss. If the noise floor changes with frequency, then the symmetry of the first mixer should be improved by small drops of solder on the UT-085 balun. The remaining trimmers in the sawtooth and linearization module have to be adjusted so that the display matches the selected span width.

Of course, a number of fine adjustments are only performed after spotting problems in the practical use of the spectrum analyzer. Some adjustments, like the precise impedance matching of the wide-band VCO to the first mixer, can only be performed with a tracking generator, to be described in a future article. In any case it is assumed that the builder is familiar with the theory of operation of a spectrum analyzer, since in this case most troubleshooting can be performed with the same spectrum analyzer.

Original full-size drawings


Tracking Generator 100kHz...1750MHz


1. Tracking generator principles

A basic and very useful addition to a spectrum analyzer is a tracking generator. A tracking generator produces a RF signal on the exact frequency where the spectrum analyzer is receiving at the same time. A tracking generator therefore allows testing many passive and active RF circuits that do not produce any RF signals on their own, like filters, amplifiers etc. With an additional directional coupler or bridge, reflection measurements can be performed in the whole frequency range covered by the spectrum analyzer.

The simplest tracking generator is a wide-band noise source, since part of the noise power always falls in the instantaneous receiving bandwidth of the spectrum analyzer. Since the noise figure of typical spectrum analyzers is rather high, in the range around 15-20dB, the required ENR (excess noise ratio) of the noise source is very high. Further, the dynamic range of noise measurements is quite limited, since the device under test and the input of the spectrum analyzer are both loaded with much larger power levels than are actually being measured. Finally, noise measurements always require some averaging (video filtering) due to the random nature of noise signals.

To avoid the drawbacks of noise measurements, almost all manufacturers offer tracking generators that produce a coherent sinusoidal signal on the receiving frequency of the spectrum analyzer, either built-in or as separate units. A tracking generator usually does not include very expensive components. Nevertheless, commercial tracking generators may be quite expensive, up to one quarter of the cost of the spectrum analyzer. Therefore it makes sense to build companion tracking generators even for commercial spectrum analyzers.

A tracking generator designed to operate with the spectrum analyzer described in [1] or [2,3] will be described in this article. Since most commercial spectrum analyzers use similar frequencies for the first IF (usually in the range 2...3GHz), many of the described circuits are also useful to build a tracking generator for a commercial spectrum analyzer. Of course, the exact circuit diagram of a tracking generator should match the frequency conversions in the corresponding spectrum analyzer.

A tracking generator obtains its output signal exactly in the opposite way the different frequency conversions are made inside the spectrum analyzer. A tracking generator therefore requires the signals from all variable oscillators (VCOs) inside the spectrum analyzer. Most spectrum analyzers therefore have the signals of the first, second and sometimes even third local oscillator accessible on suitable RF connectors on the front panel or sometimes back panel.

The spectrum analyzer, described in [1] or [2,3], includes two variable oscillators (VCOs) for the first and second conversion. A suitable tracking generator requires the same conversion processes in the reverse direction, as shown on the block diagram on Fig.1. First, the tracking generator should add the frequency of the second IF (around 70MHz) to the second VCO at 2.03GHz (+/-10MHz). The sum is a signal at the first IF around 2.1GHz. The latter is afterwards subtracted from the frequency of the first VCO (2.1...3.85GHz). An automatic-gain-control (AGC) loop is used to stabilize the output-signal amplitude over the whole frequency range 100kHz...1750MHz.

Fig.1 - Tracking-generator block diagram.

Frequency summation and subtraction can be performed in different ways: using mixers and filters or with phase-locked loops (PLLs). All signals inside the tracking generator have relatively high levels. Unwanted mixing products should be monitored carefully. On the other hand, thermal noise is not very important if compared to the design of a spectrum analyzer or other receivers. Finally, suitable shielding should be provided to avoid unwanted signal paths as well as using buffer amplifiers to provide the required separation wherever necessary.

In the described tracking generator, the first frequency summation 70MHz and 2.03GHz is performed by a phase-locked loop. The second conversion (subtraction) of the 2.1GHz first IF from the first LO 2.1...3.85GHz is performed by a balanced mixer followed by a low-pass filter on the output. Two identical buffer amplifiers are used for both VCO signals to prevent any interference from the tracking generator back into the spectrum analyzer.

The AGC sets the output-signal level to about 1mW (+0dBm). The latter can be further attenuated down to -40dBm when required on the main tracking-generator output. An auxiliary, fixed -10dBm output is provided for a frequency counter.

The described tracking generator includes nine shielded RF modules and a power supply. Some modules are identical to those in the spectrum analyzer: the PLL mixer is identical to the second mixer in the spectrum analyzer while the output mixer is identical to the first mixer in the spectrum analyzer. The output attenuator with four 10dB steps is identical to the input attenuator in the spectrum analyzer. Of course, the description of all identical modules will not be repeated.


2. Buffer amplifiers

The tracking generator includes two identical buffer amplifiers for the signals of both VCOs in the spectrum analyzer. The buffer amplifiers have two main functions: amplify the signals of both VCOs to about +13dBm to drive both mixers and prevent any signals from the tracking generator returning back in the VCOs and mixers of the spectrum analyzer.

The circuit diagram of (one) buffer amplifier is shown on Fig.2. The buffer amplifier includes a -10dB attenuator and two amplification stages. The input attenuator allows a good input impedance match even with the power turned off. Even more important, the input impedance does not change much after turn-on, so that the frequency pulling of both VCOs inside the spectrum analyzer is kept small enough.

Fig.2 - Buffer amplifier.



The input attenuator is followed by a simple amplifier stage with a BFP420 transistor. The high-frequency gain decay of the latter is partially compensated by the 0.68pF capacitor in parallel to the 68ohm resistor in the attenuator. The output stage uses an ATF35176 HEMT to obtain the required output power. The output of the buffer amplifier is connected through a short piece of UT-085 semirigid cable, used at the same time as a balun inside the balanced mixer.

Both amplifier stages are supplied with +8V through suitable resistors and feed-through capacitors. The operating point of the output stage may change due to the Idss tolerances of the ATF35176 HEMT, but these changes have little effect on the performance of the amplifier. The drain voltage (Uds) may be anywhere in the range 2...3.5V.

3. PLL for 2.1GHz

The first signal-processing step inside the tracking generator is the addition of the second VCO frequency and the second IF. Of course, the tracking generator should contain its own oscillator to generate a signal on the second IF around 70MHz. The two frequencies could be simply added in a mixer. Unfortunately, the filtering of the output signal would require a complex cavity bandpass filter to remove the image at 1.96GHz as well as the second VCO leakage at 2.03GHz. Further, an AGC circuit would be required to avoid overdriving the mixer.

The same task can be performed by a phase-locked loop with its own VCO operating at the output frequency of 2.1GHz. A small fraction of the latter is mixed with the second VCO signal at 2.03GHz coming from the spectrum analyzer. The difference of the two frequencies is compared with the nominal second IF value and the result of this comparison is used to correct the frequency of the 2.1GHz VCO. The 2.1GHz signal does not require any further filtering. However, the PLL should be designed to achieve lock reliably and to track the 2.03GHz signal even when the frequency of the second VCO in the spectrum analyzer is swept.

The VCO for 2.1GHz (shown on Fig.3) is very similar to the second VCO in the spectrum analyzer operating at 2.03GHz. Since only a relatively narrow frequency range around 2.1GHz has to be covered, a single tuning varactor BB833 is sufficient in the central finger of the interdigital bandpass. A BFP183 transistor is used as the active device inside the VCO, while an INA10386 MMIC is used as the output buffer. A directional coupler takes a small part of the output signal (around -5dBm) for the PLL mixer.

Fig.3 - VCO for 2.1GHz.



The VCO for 2.1GHz is built on an identical printed-circuit board as the narrow-band VCO in the spectrum analyzer. Besides a different varactor, both the collector and base microstrips are shortened by about 2mm at their open ends to achieve a higher operating frequency. The control-voltage low-pass network has to be designed carefully to allow a fast response of the control loop.

The PLL mixer is identical to the second mixer in the spectrum analyzer. The -5dBm VCO output is connected to the 2.8GHz low-pass inside the mixer module, while the buffer amplifier is connected to the semirigid balun. The frequency difference (around 70MHz) is taken from the 800MHz low-pass.

The comparison between the nominal value 70MHz and the actual frequency difference is performed by the PLL logic shown on Fig.4. The PLL logic includes a reference crystal oscillator, two dividers for the reference frequency and the actual frequency difference and a charge-pump frequency/phase comparator.

Fig.4 - PLL logic.



While the PLL is unlocked, the actual frequency difference may deviate substantially from the nominal value around 70MHz. An upper limit of about 140MHz is set by the input divider 74F74 with the suggested BFP183 driver. The frequency difference is divided by 64, while the 8.8MHz reference is divided by 8. The 1.1MHz comparison frequency allows fast tracking of the narrow-band VCO in the spectrum analyzer even when the frequency of latter is swept.

The reference crystal-oscillator frequency is selected to 1/8 of the second IF of the spectrum analyzer. Of course, the exact value depends on the components used in the spectrum analyzer, in particular the crystal used in the third conversion (usually 60.000MHz) and crystal filter (usually 10.700MHz). For a nominal second IF value of 70.700MHz a reference crystal oscillator at 8837.5kHz is required.

The output-voltage (Vf) range of the charge-pump frequency/phase comparator (max 0...+5V) usually has to be further reduced by the two 1kohm trimmers. The frequency difference should never become too large (140MHz limit imposed by the 74F74 divider) nor should the PLL be allowed to lock onto the image response of the mixer. The charge pump with the two schottky diodes BAT62-03W is followed by a RC low-pass network, defining the settling time and stability of the feedback loop.

4. Output mixer, amplifier and AGC

The second signal-processing step is the subtraction of the 2.1GHz frequency (first IF of the spectrum analyzer) from the first LO frequency 2.1...3.85GHz. All unwanted spurious conversion products from a correctly-designed mixer can simply be removed by a low-pass filter on the output, leaving only the desired signal in the frequency range 0...1750MHz. On the other hand, it is very difficult to build a PLL covering such a wide frequency range.

The output mixer is identical to the first mixer in the spectrum analyzer. The latter module includes a 1.75GHz low-pass. The wide-band-VCO signal (2.1...3.85GHz) is fed through a buffer amplifier directly to the semirigid balun of the output mixer. No further processing is required, since the harmonics of the first spectrum-analyzer LO do not cause any harmful spurious frequencies in the tracking generator.

On the other hand, the tracking generator is sensitive to the harmonics of the 2.1GHz first-IF signal. The latter may cause spurious mixing products in the output-frequency range 0...1750MHz. Therefore, any harmonics of the 2.1GHz signal have to be removed. Further, the 2.1GHz signal level has to be controlled carefully to avoid mixer nonlinearities.

The 2.1GHz harmonics are removed by a 2.8GHz low-pass filter, followed by a PIN-diode attenuator to set the signal level for the mixer. Both circuits are included in the AGC module shown on Fig.5. The PIN attenuator is followed by another 2.8GHz low-pass in the output mixer module.

Fig.5 - AGC.



The PIN attenuator includes three PIN diodes BA596 connected in a PI network. The BC238 transistor feeds all three PIN diodes with suitable DC currents to keep the input and output impedances stable while adjusting the attenuation. The AGC control voltage spans from zero to +8V. Minimum attenuation is achieved at +8V while zero voltage provides maximum attenuation.

Since the output mixer operates in the linear region, the output signal is relatively weak, around 10uW or -20dBm. To increase the output signal level and drive the AGC detector, an output amplifier is required as shown on Fig.6. The latter is built with an INA10386 MMIC that provides 26dB gain and an output power of more than +10dBm in the whole frequency range up to 1750MHz.

Fig.6 - Output amplifier.



The INA10386 MMIC amplifier requires DC decoupling capacitors on both the input and output. The latter limit the lower end of the frequency range of the tracking generator to about 100kHz. Each coupling capacitor includes a parallel connection of two SMD capacitors: first a low-loss, 100pF NP0 0805 capacitor is soldered on the printed-circuit board and afterwards a 100nF, lossy Z5U 1206 capacitor is soldered across the 100pF capacitor.

The AGC detector (schottky diode BAT62-03W) is connected directly to the output of the INA10386 amplifier. The BC238 transistor operates as the AGC feedback amplifier, while the time constant of the feedback is set by the 680nF capacitor. Thanks to the AGC action, the output amplifier behaves as a voltage source. The main output (+0dBm) therefore requires a 51ohm series resistor. The 150ohm and 68ohm resistors allow an auxiliary -10dBm output and provide a DC path to ground for the AGC detector.

The main output of the tracking generator includes a step attenuator to decrease the output signal level down to -40dBm in four 10dB steps. The design of this attenuator is identical to the attenuator used on the input of the spectrum analyzer. Due to the non-ideal frequency response of the latter, better measurement accuracy can be achieved with the auxiliary -10dBm output, especially for wide-band measurements.

5. Construction tips

The described tracking generator is built in a similar way as the corresponding spectrum analyzer, described in [1] or [2,3]. All modules are installed in shielded boxes made from 0.5mm thick brass sheet. Most modules require a +8V supply voltage obtained from the 7808 regulator shown on Fig.7. The PLL logic has its own 7805 regulator inside the module and requires a +12V external supply. The tracking generator is equipped with an ON/OFF switch on the front panel to allow an immediate check of the signals shown on the spectrum-analyzer display.

Fig.7 - Power supply.



Except for the PLL logic, all other modules use microstrip circuit boards. Some of the latter were already described in the spectrum-analyzer article, while the new ones are shown on Fig.8. All microstrip boards are etched on one side of a 0.8mm thick FR4 glassfiber-epoxy laminate, while the other side is not etched to act as a groundplane. The PLL logic is built on a single-sided board shown on Fig.9 and etched on 0.8mm thick FR4 laminate.

Fig.8 - Microstrip circuit boards (150dpi).


Fig.9 - PLL logic PCB (30X120, 150dpi).

Microwave absorber foam is built under the cover of two modules: buffer amplifier for the wide-band (2.1...3.85GHz) VCO and output mixer. Other modules usually do not require any absorber foam inside. Microwave absorber is also not required in the space among the shielded boxes.


The tracking-generator module location is shown on Fig.10. The tracking generator has the same depth (240mm) and width (220mm) as the spectrum analyzer, so that the two boxes can be stacked easily one over another. The height of the tracking generator is only 32mm, since all modules are located in a single plane. The bottom of the box is simply a piece of 1mm thick aluminum sheet, bent in the form of an "U". The cover is a similar "U" made from 0.6mm thick aluminum sheet.

Fig.10 - Tracking-generator module location.



6. Alignment and checkout

The tracking generator is a much simpler piece of test equipment than the spectrum analyzer. Correspondingly the alignment and checkout should be much simpler too. Of course, the spectrum analyzer should be available at all times to supply the signals of both VCOs. Before connecting all tracking-generator modules together, it makes sense to make a few simple checks on each module separately.


In all modules it makes sense to check the DC bias points of all semiconductors. Most simple errors can be detected in this way. Next some RF checks can be performed. Both buffer amplifiers should be checked for the output RF power level when driven by the corresponding VCO in the spectrum analyzer.

The VCO needs a fine tuning of the frequency coverage, since the microstrip board was originally designed for a slightly lower frequency in the spectrum analyzer. Of course, the operating frequency of the VCO should be monitored while carefully shortening the base and collector strips. The VCO should achieve the nominal operating frequency of 2.1GHz at a tuning voltage of 3...3.5V. Further, the VCO should operate without dropouts through the whole range of tuning voltages 0...+5V. Some trimming of the center finger (with the BB833 varactor) may also be required.

In the PLL logic, the operation of the crystal oscillator should be checked first. Both 1kohm trimmers should be set initially with the sliders on the respective "hot" ends, to allow the widest span of the output voltage Vf. Without any input signal Vf should reach almost +5V. If a frequency above 70.7MHz is fed to the input (for example a grid-dip meter coupled through a small wire loop), the Vf output should drop to zero.

Using the already tested modules, the 2.1GHz PLL can be assembled and tested. While the PLL is locked, the DC voltage on the LOCK test point should drop to less than 0.2V. Both 1kohm trimmers in the PLL logic are then set for a frequency coverage of +/-60MHz around the nominal frequency of 2100MHz. While testing the PLL, the main output of the VCO should be terminated properly on a matched load.

After the alignment of the PLL is completed, the remaining modules of the tracking generator can be wired together. The latter do not include any alignment points, but their operation should be verified. The spectrum analyzer should be set to scan the whole frequency band 0...1750MHz with the widest (4MHz) IF filter. While connecting the output of the tracking generator to the input of the spectrum analyzer, the whole trace should raise almost to the full scale. The trace will not be perfectly straight. The required fine adjustments to flatten the response will be described later.

The next step is to test the tracking generator with narrower IF filters. There will probably be no change with a 700kHz IF bandwidth. However, the response will probably drop with narrower IF filters. The response should be brought back to its original value by adjusting the crystal oscillator in the PLL logic. If the range of the trimmer is insufficient, a parallel capacitor may be added or the trimmer may be replaced with an adjustable inductor. In the worst case, the crystal in the PLL logic and/or the 60MHz crystal in the spectrum analyzer may have to be replaced.

In order to check the operation of the AGC, the Y input of the oscilloscope (used as the display of the spectrum analyzer) is temporarily connected to the AGC line. Although the AGC voltage may span 0...+8V, it should remain in the range +1...+3V in a correctly-operating tracking generator. A too high or too low AGC voltage may suggest what is actually wrong with the tracking generator. While testing the AGC voltage, the main output of the tracking generator should be terminated to a matched load. If the main output is left unterminated, the AGC voltage may move out of the required +1...+3V range.

The main checkout of the tracking generator is completed at this point. However, several minor adjustments may be necessary both in the tracking generator and in the spectrum analyzer to optimize the operation. In particular, the frequency response of the spectrum analyzer should be flattened as much as possible. Without any adjustments, the response of the tracking generator connected directly to the spectrum analyzer may deviate by as much as +/-5dB.

Amplitude variations of the frequency response of the spectrum analyzer are mainly caused by the first mixer and its termination impedances. Additional impedance matching may flatten the overall response and/or move dips to compensate upward bumps in the response. Some dips and bumps move quickly while changing the length of the cable connecting the first mixer to the cavity filter. Small pieces of copper foil, soldered to the output microstrip of the wide-band VCO module, are usually very efficient in improving the impedance matching of the LO mixer port.

A substantial mismatch, especially above 1GHz, is also caused by the two step attenuators with toggle switches. Above 1GHz it is therefore recommended to use the aux -10dBm output to avoid at least one of the two step attenuators. A better solution is to connect external, fixed 10dB microwave attenuators (with SMA connectors) to both the tracking-generator output and the spectrum-analyzer input.


After all described fine adjustments, the frequency response of the tracking generator and spectrum analyzer connected together should be within +/-2dB from 100kHz to 1.6GHz. The decay above 1.6GHz is caused by the low-pass on the input of the spectrum analyzer. An accuracy of +/-2dB is reasonable even for much more expensive professional spectrum analyzers.

Of course, since the response of the spectrum analyzer and tracking generator can be measured easily, the resulting error can simply be subtracted from the real measurement result. Most manufacturers therefore offer an electronic storage/normalizer unit to be connected between the spectrum analyzer and the oscilloscope display. An electronic storage/normalizer (using cheap integrated circuits) was also developed and built for the described spectrum analyzer and tracking generator.

Original full-size drawings


Harmonic Converter for Spectrum Analyzers


1. Extending the frequency coverage of spectrum analyzers

RF spectrum analyzers are usually designed as scanning receivers with a high first intermediate frequency. If the first IF is set above the input frequency range, unwanted mixing products can be removed with a low-pass filter. Similar designs are therefore used also in other test equipment, like signal generators.

A high first IF implies that the circuits inside the test equipment should operate at relatively high frequencies. However, there are several technological constraints limiting the first IF to below about 5GHz. The most important constraint is probably the accumulation of phase noise of the oscillators used in the different up-conversions and down-conversions.

Practical spectrum analyzers therefore have the first IF in the range between 2GHz and 3GHz. For operation above 3GHz the input low-pass filter is replaced by a tunable (YIG) bandpass filter. Even without any input filtering, a spectrum analyzer is still able to provide useful information by carefully considering all mixing products with both the fundamental oscillator frequency and its harmonics.

For operation at very high frequencies (in the millimeter-wave range), commercial spectrum analyzers are usually supplied with external (waveguide) harmonic mixers, as shown on Fig.1. Of course the spectrum analyzer should provide both a LO output and an IF input to connect an external mixer. Some harmonic mixers may even require a bias adjustment to optimize the conversion efficiency at a particular harmonic.

Fig.1 - Extending the SA coverage with a harmonic mixer.

In a homemade spectrum analyzer one cannot probably afford an expensive YIG-tuned preselector. Although there may be differences in the conversion efficiencies of different mixers, harmonic mixing almost always provides useful results. The only remaining problem is the identification of the different mixing products when a tuned preselector is not available.

A very efficient solution is an additional harmonic converter, including a harmonic mixer and an adjustable local oscillator, as shown on Fig.2. The adjustable conversion oscillator allows an additional degree of freedom to shift the harmonic-mixer responses. The latter allows identification and convenient separation of the different mixer responses even without a tuned input preselector.

Fig.2 - Extending the SA coverage with a harmonic converter.

The same external (harmonic) converter can also be used to extend the frequency coverage of a tracking generator as shown on Fig.3. In fact the tracking generator itself is no longer needed, since the first LO output of the spectrum analyzer is used as the tracking generator. The additional (harmonic) converter is then used to translate the 2...3.7GHz band back to 0...1.7GHz, where the basic spectrum analyzer is receiving.

Fig.3 - Extending the tracking-generator coverage with a (harmonic) converter.

In the case of a typical spectrum analyzer with a 2GHz first IF this approach allows an additional coverage of 2...3.7GHz besides the usual range of 0...1.7GHz offered by conventional tracking generators. However, two constraints have to be met. First, the additional converter has to be tuned precisely to the SA first IF. Second, sufficient filtering has to be provided so that the converter oscillator does not enter into the SA IF. A separation buffer amplifier for the SA LO signal is required for the same purpose.

A simple, tunable harmonic converter for spectrum analyzers will be presented in this article. Although this converter was originally designed for the spectrum analyzer shown in [4] or [5], it will operate with any spectrum analyzer having the first IF between 2GHz and 3GHz. The detailed block diagram of the harmonic converter is shown on Fig.4.

Fig.4 - Harmonic converter detailed block diagram.



Besides a harmonic mixer the converter includes a tunable oscillator covering approximately the same frequency range as the VCO inside the spectrum analyzer (2.1...3.85GHz). The latter allows selecting the best mixing harmonic to obtain a spurious-free response and a good conversion efficiency at the same time. Of course the frequency coverage should include the SA first IF to allow the extension of the range of the tracking generator.

The (IF) output of the harmonic converter includes a wide-band amplifier and several low-pass filters. The amplifier is required to compensate for the high conversion loss of the harmonic mixer as well as poor noise figure of the following spectrum analyzer. Low-pass filtering has to be performed in several stages to remove all unwanted mixing products as well as to prevent the harmonics of the conversion oscillator from entering the basic spectrum analyzer.

Finally, the harmonic converter includes all required power supplies, since it is intended as a stand-alone unit. The VCO and output amplifier require a +8Vdc supply obtained with a 7808 regulator. A DC/DC converter with an additional stabilizer is used for the VCO tuning voltage 0...30V.

2. Harmonic mixer

The circuit diagram of the harmonic mixer is shown on Fig.5. In fact the harmonic mixer does not differ much from other mixers, like the first and second mixers in the spectrum analyzer shown in [4] or [5]. The dual schottky diode BAT14-099 produces useful harmonics up to at least 30GHz while driven with a large (+13dBm or 20mW) LO signal in the frequency range 2.1...3.85GHz.

Fig.5 - Harmonic mixer.



Since a flat response across the whole microwave frequency range is desired, particular attention has to be made to suppress any parasitic resonances of the circuit. The input signal is therefore coupled using very small 0402 SMD capacitors. Since the width of a single capacitor is 0.5mm, three such capacitors are installed in parallel to match the 1.5mm width of the 50ohm microstrip line.

Unwanted resonances of the balun made from UT-085 semirigid cable are suppressed by four 68ohm SMD resistors. The 10ohm and 22ohm resistors are also used to suppress circuit resonances. Finally, the whole shielded module requires a piece of microwave absorber (antistatic foam) under the cover to suppress cavity resonances.

The harmonic mixer does not include any attenuator on the input. Since it is difficult to build good attenuators for frequencies above 10GHz with conventional SMD resistors, it is recommended that external SMA attenuators are used when required. The SMA connectors themselves limit the frequency range of the harmonic mixer to about 26GHz. Although there is no input filter, waveguide transitions were found to be excellent high-pass filters when required.

3. Wide-band VCO

The design of a suitable wide-band VCO was already shown in [1], [2] or [3]. The circuit diagram on Fig.6 has just a small modification. A 10uF capacitor is added to the tuning voltage line for additional filtering, since inside the harmonic converter the VCO frequency is only set manually wih a precision Helipot potentiometer.

Fig.6 - Wide-band VCO.



The main output of the VCO (+13dBm) drives the harmonic mixer directly. The auxiliary output (about -5dBm) is made available on the back panel. The latter may be used to check the conversion frequency with a counter or as a 2...3.85GHz signal source for other purposes.

For operation with the spectrum analyzer described in [4] or [5] or other spectrum analyzers with the first IF around 2GHz, BB833 varactors should be used in the VCO to allow a frequency coverage from about 2GHz to about 3.85GHz. For spectrum analyzers with the first IF above 2.5GHz, BB857 varactors are recommended (VCO range 2.4GHz to 4.6GHz).

4. Output amplifier

The circuit diagram of the output amplifier is shown on Fig.7. The amplifier is designed around an INA10386 MMIC that provides a flat gain of more than 20dB and a noise figure better than 4dB across the whole frequency range up to 1.75GHz. The INA10386 provides just the correct amount of gain to compensate for the mixer conversion loss and SA noise figure.

Fig.7 - Output amplifier.



The output amplifier is necessarily installed in its own shielded enclosure, since unwanted signal filtering is very critical. Besides the low-pass filter inside the harmonic mixer module (cutoff about 1.75GHz), two additional low-pass filters with cutoff frequencies of about 2.8GHz are installed before and after the INA10386 amplifier. To suppress higher-order responses of the microstrip circuits as well as enclosure cavity resonances, microwave absorber foam is also needed under the cover of the amplifier module.

The INA10386 MMIC requires DC decoupling capacitors on both the input and output. Since the harmonic converter is usually used at relatively wide SA resolution bandwidths (100kHz or more), the IF range can be limited to 100kHz on the lower end. Even in the latter case finding suitable coupling capacitors is not simple. The best results were found with a parallel connection of a low-loss, 100pF NP0 0805 size capacitor that is soldered on the PCB first. Then a lossy 100nF (Z5U ceramic) and larger (1206) capacitor is soldered across the smaller 100pF capacitor.

5. Power supply

Like the spectrum analyzer described in [4] or [5] or the companion tracking generator described in [6], the harmonic converter requires an unstabilized +12Vdc supply. The latter is stabilized to +8V with a 7808 regulator to supply the VCO and output amplifier. The 7808 regulator is wired exactly in the same way as in the spectrum analyzer or tracking generator. The harmonic converter usually remains on all of the time to minimize frequency drifts, so a turn-on switch is usually not required.


The BB833 (or BB857) varactors inside the wide-band VCO require much higher tuning voltages up to +30V. Further, the tuning voltage has to be well stabilized and filtered, to avoid frequency drifts and excessive phase noise. Both unwanted effects are further multiplied by the order of the harmonic inside a harmonic converter!

The tuning voltage supply is shown on Fig.8. The circuit includes a flyback DC/DC converter with the transistors BC308 and BD139. The rough voltage regulation provided by the two 18V zener diodes and BC548 feedback transistor is far from being sufficient for a tuning voltage. The resulting +37V is further stabilized down to +30V with a uA723 precision regulator. Due to the low current drain at +30V, no external transistors are required around the uA723. However, the uA723 still requires some feedback resistors to set the output voltage, over-current protection and a capacitor for frequency compensation.

Fig.8 - Stabilized DC/DC converter.



The actual VCO tuning voltage is obtained through a precision 50kohm Helipot potentiometer from the well-regulated +30V supply. The potentiometer is installed on the front panel and is equipped with a revolution-counting dial knob. Since there is no tuning-voltage linearizer, the frequency scale is quite nonlinear. The latter may even be an advantage in a harmonic converter, where fine frequency adjustments are usually required at the upper end of the frequency coverage while using high-order harmonic mixing at input frequencies above 10GHz.

6. Construction tips

With the exception of the DC/DC converter, the harmonic converter only includes microstrip circuit boards. The harmonic mixer and output amplifier boards are shown on Fig.9 while the wide-band VCO has already been described in [1], [2] or [3]. All microstrip boards are etched on 0.8mm thick, double-sided FR4 glassfiber-epoxy laminate. The DC/DC converter is built on a single-sided board as shown on Fig.10, etched on 1.6mm thick FR4 laminate.

Fig.9 - Microstrip circuit boards (150dpi).


Fig.10 - DC/DC converter PCB (30X80, 150dpi).

All three microstrip boards are installed in shielding boxes made from 0.5mm thick brass sheet just like the modules of the spectrum analyzer [4] or [5] or companion tracking generator [6]. All three brass boxes require a microwave absorber (antistatic foam) under their covers to suppress cavity resonances.


The harmonic converter module location is shown on Fig.11. The harmonic converter has the same depth (240mm) as the spectrum analyzer [4] or [5]. The width is 100mm while the heigth is only 32mm, since all modules are located in a single plane. The bottom of the box is simply a piece of 1mm thick aluminum sheet, bent in the form of an "U". The cover is a similar "U" made from 0.6mm thick aluminum sheet.

Fig.11 - Harmonic-converter module location.



The 7808 regulator is bolted to the back plate for heatsinking purposes. The two 470uF electrolytic capacitors, VK200 RF choke and 1N5401 diode are simply soldered between the +12V supply connector and the leads of the 7808 regulator. The overall current drain amounts to about 150mA.

7. Alignment and checkout

A harmonic converter is a much simpler piece of test equipment when compared to a spectrum analyzer or companion tracking generator. The described harmonic converter has no tuning points. Of course, the correct operation of the power supply including the DC/DC converter should be checked first.


The most demanding module is certainly the wide-band VCO. The latter should be checked for the frequency coverage as well as output-signal level. The latter should not drop below +10dBm on any frequency. Of course, if the spectrum analyzer published in [4] or [5] is being built at the same time, it makes sense to install the better VCO inside the spectrum analyzer and use the remaining VCO for the harmonic converter.

The spectrum analyzer should be sensitive enough to display the noise of the harmonic-converter output amplifier even with an input attenuator setting of 10dB. This corresponds to a spectrum-analyzer noise figure of about 20dB (without the input attenuator) that is a typical value for most spectrum analyzers.


Finally the first LO output from the spectrum analyzer is connected to the harmonic-converter. Additional attenuators may be required to bring the input signal level of the latter to about -10dBm. When the harmonic converter is tuned to the exact first IF of the spectrum analyzer, the setup can be used as a tracking generator.

Original full-size drawings


Storage-Normalizer for Spectrum Analyzers


1. Displays for spectrum analyzers

Spectrum analyzers usually produce large quantities of information that can be properly represented only on a fully graphic display. The latter can be a simple oscilloscope CRT display, a raster-scan CRT or a dot-matrix LCD. Many older spectrum analyzers were equipped with delicate and difficult-to-use CRTs with a special storage screen. Newer spectrum analyzers have digital video memories and usually provide a hardcopy option, either with a built-in printer or through a general computer interface.

Unfortunately, new display technologies do not necessarily mean improvements of the overall instrument performance. In particular, in the case of radio-frequency spectrum analyzers, all engineers appreciate the good old analog oscilloscope displays. New spectrum analyzers, including the highest-priced models, all include digital displays that have both too low resolution and too slow display update.

Of course, the best solution is to have both analog and digital display technologies available in a single instrument. An older spectrum analyzer or a homemade instrument with an analog oscilloscope display can easily be upgraded with an additional storage-normalizer unit. Since the latter can be switched in and out as required by the particular measurement, it allows the advantages of both display technologies in a single instrument.

A storage-normalizer is particularly useful while using a tracking generator as shown on Fig.1. In the latter case, the response of the tracking generator, spectrum analyzer and other test equipment is stored in digital memory during system calibration. The response of the device-under-test (DIRECT) can then be quickly compared with the stored system response (MEMORY). Since most spectrum analyzers provide a logarithmic video signal (usually 10dB per division), a simple subtraction of the system response from the measured trace (DIFFERENCE) removes any system influence on the measured device response.

Fig.1 - Principle of operation of a storage-normalizer.

2. Storage-normalizer design

A simple storage-normalizer is described in this article, to be used with the spectrum analyzer described in [1] or [2] or other similar instruments. Since the narrowest resolution bandwidth of the above-mentioned spectrum analyzer is only 10kHz, very slow frequency sweeps are usually not used. A digital storage is therefore not required for faster frame refresh at very slow sweep times.

The design of the storage-normalizer is much simplified since the display always operates at the same sweep frequency as the spectrum analyzer. All required clock and address signals are generated by simple and fast hardware logic in place of a much slower microcontroller. Fast A/D and D/A converters and a sufficiently large memory allow high-quality images to be stored in the video memory.

The video bandwidth of the spectrum analyzer ([1] or [2]) amounts up to 500kHz without any video filtering. To avoid any video degradation, a sampling frequency of 1MHz was selected. A sampling frequency of 1MHz is far too high for most microcontrollers. Therefore a simple control logic with 74HCxxx circuits generates all required clocks and addresses.

The dynamic range of the logarithmic detector amounts to almost 100dB. Inexpensive 8-bit A/D and D/A converters therefore allow a resolution of about 0.4dB for the storage-normalizer. A resolution of 0.4dB is comparable to the accuracy of the 10-stage logarithmic detector presented in [1] or [2].

The A/D and D/A converters as well as the video memory are shown on Fig.2. The A/D converter TDA8703 is a 8-bit flash A/D converter that allows sampling frequencies up to 30MHz. Flash A/D converters include a comparator for each step of their transfer function, in total 255 comparators in a 8-bit A/D converter. Their main advantage is very fast operation. Also, flash A/D converters do not require a critical sample-and-hold circuit on the analog input.

Fig.2 - A/D and D/A converters and video memory.



The TDA8703 has three-state outputs to be connected directly on an 8-bit data bus. The O/UF (over/underflow) is an additional output that goes high whenever the input-signal level is too low or too high, in other words outside the normal operating range of the A/D converter. While testing the circuit of the storage-normalizer it may be useful to connect a LED from the O/UF output to ground, of course through the 220ohm current-limiting resistor.

The 62256 is a 32768-byte static CMOS RAM. At a sampling frequency of 1MHz it can store about 32ms of video signal. The storage-normalizer is intended to be used at a sweep speed of 2ms/division for a total sweep time of 20ms, resulting in 20000 samples or 20000 bytes to be stored in the video memory. The 62256 has eight common input-output pins D0...D7, that can be tri-stated and connected directly to an 8-bit data bus.

The TDA8702 is a fast 8-bit D/A converter. It includes an internal 8-bit latch on the digital inputs. Data transfer is triggered by the DAC-CLK, so that the same data bus can also be used for other purposes. The TDA8702 has two complementary analog outputs. Both outputs are used in the storage-normalizer to further reject noise and voltage variations on the +5V supply line.

The A/D and D/A converters as well as the memory require several steering signals. The clock/address generator is shown on Fig.3. All clocks run at 1MHz, but have different pulsewidths and phases. All clocks are obtained from a 8MHz crystal oscillator driving a 8-bit shift register (74HC164). The shift register is connected as a frequency divider by 8. The different clocks are obtained from the 74HC164 outputs through a few NAND gates (two 74HC00).

Fig.3 - Clock/address generator.

The A/D converter TDA8703 performs one conversion for each ADC-CLK pulse. The outputs of the TDA8703 are then enabled by the /ADC-CE signal. Writing to the memory is enabled with /WR while reading from the memory is enabled by /OE. Finally the byte from the memory is copied into the D/A converter TDA8702 with DAC-CLK. When all clocks are disabled, the output Q0 of the 74HC164 increments the address counter.

The address counter includes four synchronous counters 74HC161. The TRIG pulse coming from the spectrum analyzer signals the start of the sweep. After the TRIG pulse is "cleaned" in a D-flip-flop (one half 74HC74), it resets the address counter. The address counter supplies a 15-bit address to the memory. The last counting stage (Q3 of the last 74HC161) is used to stop the address counter until the next trigger pulse.

Writing the video data to memory is enabled manually by depressing the WRITE key. The write command is synchronized to the display sweep with a D-flip-flop (second half 74HC74). In this way one or more whole sweeps are written to the memory regardless of the WRITE-key contact bouncing. The WRITE command enables the /WR clock. The latter remains in its inactive (high) state when writing is not enabled.

The digital memory requires an analog interface as shown on Fig.4. The latter is used to adjust the A/D and D/A signal levels to the values used between the spectrum analyzer and oscilloscope display (20dB/V, impedance 680ohm). Further, the analog interface includes a circuit with operational amplifiers to compute the difference between the direct video signal and memory content.

Fig.4 - Analog interface.



The analog interface includes a quad operational amplifier MC33074. The first op-amp (pins 5,6,7) is a voltage follower for the input signal. The A/D converter gets part of this signal through a resistive divider. The second op-amp (pins (8,9,10) amplifies the output of the D/A converter to the same level as available directly from the spectrum analyzer. The third op-amp computes the signal difference. Finally, the fourth op-amp (pins 1,2,3) operates as a voltage follower for the DIFFERENCE-OFFSET control voltage.

The four operational amplifiers run on a +12V supply, since the input and output video-signal voltages usually range between 1V and 6V. A stable +7V reference voltage is provided by the uA723 regulator. On the other hand, the A/D and D/A converters as well as the memory require a +5V supply voltage. The overall power supply is shown on Fig.5 providing +5V for the digital memory +12V for the analog interface.

Fig.5 - Power supply.



3. Storage-normalizer assembly and alignment

The wiring of the storage-normalizer modules is shown on Fig.6. The spectrum analyzer provides four signals: video (Y-deflection), triggering (TRIG), blanking (BLANK) and sawtooth (X-deflection). The storage-normalizer only affects two signals: the video signal is being processed while the TRIG signal is used for synchronization of the address counter.

Fig.6 - Storage-normalizer wiring diagram.



The storage-normalizer is built on two printed-circuit boards except for the power supply from Fig.5. The latter is built around the 7805 regulator and +12V supply connector. The digital part of the storage-normalizer (A/D and D/A converters, memory and clock/address generator) are built on a double-sided printed-circuit board with the dimensions of 120mmX80mm as shown on Fig.7. The analog interface is built on a single-sided printed-circuit board with the dimensions of 80mmX40mm as shown on Fig.8. Both circuit boards are etched on 1.6mm-thick glassfiber-epoxy laminate FR4.

Fig.7 - Digital-memory circuit board (150dpi).


Fig.8 - Analog-interface circuit board (150dpi).

The storage-normalizer module location is shown on Fig.9. The storage-normalizer has the same depth (240mm) as the spectrum analyzer [1] or [2]. The width is 100mm while the height is only 32mm, since all modules are located in a single plane. The bottom of the box is simply a piece of 1mm thick aluminum sheet, bent in the form of an "U". The cover is a similar "U" made from 0.6mm thick aluminum sheet.

Fig.9 - Storage-normalizer module location.



The 7805 regulator is bolted to the bottom plate for heatsinking purposes. The two 470uF electrolytic capacitors, VK200 RF choke and 1N5401 diode are simply soldered between the +12V supply connector and the leads of the 7805 regulator. The overall current drain amounts to about 150mA.


Although the storage-normalizer is a low-frequency video circuit, the analog interface requires a few alignments to make best use of the available dynamic range of both A/D and D/A converters. Of course the storage-normalizer should perform some useful function without any alignment (with both trimmers and potentiometer in central position) and this can be checked immediately with the oscilloscope display.

In the storage-normalizer the correct input signal level to the A/D converter should be set first. For this purpose the spectrum analyzer is tuned to represent a wide and high peak across the whole display in the DIRECT mode. Next the storage-normalizer is switched to DIFFERENCE and the WRITE key is kept depressed. The GAIN trimmer is then adjusted to obtain a straight line on the oscilloscope display.

If a straight line can not be obtained, the input signal level is probably outside the range required by the TDA8703. This can be readily checked with a LED connected to the O/UF output. The most efficient counter-measure is to correct the DC offset of the video amplifier inside the spectrum analyzer. Since the LED is no longer required after the video signal levels in the spectrum analyzer and storage-normalizer are adjusted correctly, the LED is not installed on the front panel.

After the trimmer GAIN is adjusted correctly, the storage normalizer is switched alternatively in the DIRECT and MEMORY modes. The trimmer MEMORY OFFSET is then adjusted so that there is no observable difference between the DIRECT and MEMORY displays. Finally, the potentiometer DIFFERENCE OFFSET is a front-panel command and is set accordingly to the type of measurement that is performed: spectrum analyzer alone, tracking generator or different devices under test with insertion loss or gain.


While using the described simple storage-normalizer one should however understand that the latter has no information about the settings of the spectrum analyzer. If the central frequency, sweep width or sweep time are changed, the information stored in the storage-normalizer probably becomes useless. The whole system should be then calibrated again, writing new information in the storage-normalizer memory.

Original full-size drawings


Marker Counter for Spectrum Analyzers


1. Amplitude and frequency display

Spectrum analyzers are used to measure the frequency and amplitude of radio-frequency signals. Both quantities are at least roughly displayed on the CRT screen. When the spectrum analyzer only sweeps across a narrow frequency band, some additional frequency information is required to locate the narrow segment in the whole frequency range covered by the spectrum analyzer. Since the most popular amplitude display is 10dB/div, some additional accuracy is highly desired at least in a few interesting points on the CRT screen.

Older spectrum analyzers were equipped with a mechanical frequency dial. The dial accuracy was limited by the precision potentiometer used to set the current through the tuning coil of the YIG oscillator. The YIG-oscillator frequency is exactly proportional to the DC magnetic field, set by the current through the tuning coil.

The delicate and unreliable mechanical dial was quickly replaced by a digital milliampere-meter with three or four digits. The accuracy of this kind of frequency display is severely limited by the hysteresis of the core of the YIG magnet. It is sufficient to tune the spectrum analyzer first to zero, then move to the highest frequency and return back to zero to observe frequency errors as large as +/-20MHz!

An accurate frequency display only became possible recently with the availability of true microwave synthesizers including microwave prescalers. There are still many new spectrum analyzers on the market, including models with microprocessor control, that do not have a true frequency readout. Since only the current through the tuning coil is measured, their frequency readout relies on the linearity of the YIG tuning response.

All spectrum analyzers can be equipped with an external frequency counter. Since the circuit of a spectrum analyzer includes many frequency conversions, the frequencies of all local oscillators have to be measured and the result added to or subtracted from the last IF value. Further, the gates of the frequency counters need to be synchronized with the spectrum-analyzer sweep to measure the frequencies in a known point of the display.

The task of the frequency counter can be simplified with a tracking generator. The latter performs all required frequency summations and/or subtractions to obtain a signal on the exact frequency of operation of the spectrum analyzer. While using a tracking generator, the operating frequency of the spectrum analyzer can be measured with any frequency counter, provided that the spectrum analyzer is switched to zero span.

Modern spectrum analyzers with microcomputer-supported graphical displays can measure and display the frequency of one or more arbitrary points on the screen. These points are highlighted by well-visible markers on the screen. The frequency and amplitude of the signal at each marker are then displayed in numerical format as well.

The design of a suitable frequency counter displaying both frequencies and amplitudes of all markers allows many degrees of freedom and is almost independent on the type of the basic spectrum analyzer as well as auxiliary equipment (tracking generator, storage-normalizer). For a single-marker frequency measurement, a counter synchronized to the spectrum-analyzer sweep is sufficient. On the other extreme, a microcomputer-supported display is only limited by the performance of the microprocessor used to draw the picture.

A simple marker counter will be described in this article, designed to operate together with the spectrum analyzer [1] or [2] and corresponding tracking generator [3]. The tracking generator allows a simple frequency counter with a single input and a single prescaler. Of course, the gate of the counter is synchronized to the spectrum-analyzer sweep.

The block diagram of the marker counter for spectrum analyzers is shown on Fig.1. The circuit performs three functions: in the selected point on the screen it measures and numerically displays the frequency and amplitude of the signal and at the same time draws a marker on the oscilloscope display.

Fig.1 - Marker counter for spectrum analyzers.

The frequency counter includes three units: prescaler, counter and time base. The prescaler divides the input frequency by 256, allowing the use of a relatively slow counter. The counter module includes a decoder and driver for the four-digit, 7-segment LCD. The time base module provides a reset to the counter, opens the gate and stores the result for the LCD decoder/driver. A variable delay of the TRIG pulse coming from the spectrum analyzer selects the horizontal position of the marker where all measurements are done.

The video-signal sampler is triggered at the same time as the frequency counter. The video-voltage sample is stored in a capacitor and displayed on a digital voltmeter. The scale of the latter is calibrated directly in decibels. An additional front-panel command allows an arbitrary offset of the decibel scale.

Finally, the same MARKER pulse drives the circuit that actually adds the marker to the video signal. The marker module includes a 455kHz oscillator. The output of the latter is added to the Y-VIDEO signal, adding a short vertical line at the marker position on the oscilloscope display

Since spectrum analyzers are sensitive radio receivers, the marker counter is installed in its own shielded case. Metal shielding is sufficient in most cases, even with a short whip antenna installed directly on the input connector of the spectrum analyzer. Of course, if interference from the marker counter or tracking generator is suspected, either or both can be simply turned off. The circuits of the marker counter are therefore designed not to disturb the Y-VIDEO and TRIG signals in the power-off state.

2. Prescaler

The tracking-generator output frequency ranges almost from DC up to a few GHz. A high-speed prescaler is required to measure very high frequencies. On the other hand, a prescaler slows down the operation of the counter. Therefore, the divider modulo of the prescaler is not arbitrary.

In the case of a spectrum analyzer, there is not much time to measure the frequency. The frequency measurement should be shorter than the time of one complete sweep, usually around 20ms. The resolution of the frequency counter is therefore limited to 100kHz. The latter requires 2.56ms when using a divide-by-256 prescaler. A better frequency resolution could only be obtained by averaging several measurements in consecutive sweeps of the spectrum analyzer.

An additional difficulty is represented by the zero on the frequency scale. The operation of the tracking generator is quite uncertain at very low frequencies. The prescaler is even less reliable at very low input frequencies. Accurate frequency measurements around zero can only be obtained by measuring the frequencies of all variable oscillators in the spectrum analyzer and then computing the final result according to the conversions used in the spectrum analyzer.

The lower frequency limit for the tracking generator shown in [3] is about 100kHz. The latter is a very low frequency for ECL prescalers. Inexpensive prescalers for TV receivers are only designed to operate above about 70MHz. The extension of the prescaler frequency range to very low frequencies requires careful input signal processing.

Inexpensive prescalers U664 and U891 were first tried in the described circuit. Their upper frequency limit is about 1.6GHz and is very sensitive to chip tolerances and ambient temperature. It can be somewhat improved using a higher-than-nominal supply voltage. Unfortunately both the U664 and U891 will hardly operate correctly below 20MHz, probably due to some capacitive coupling inside the prescaler chip itself.

Fortunately prescalers from other manufacturers did not show the same behavior at very low input frequencies. Much better results were obtained with the SDA4212. The latter operates up to 1.8GHz at room temperature. On the other hand, its lower frequency limit can be extended to less than 1MHz with suitable input signal processing. Both limits compare favourably with the requirements of the spectrum analyzer described in [1] or [2].

The circuit diagram of the whole prescaler is shown on Fig.2. The input amplifier INA03184 is mainly required at low frequencies to transform the input sinewave into a squarewave for the ECL prescaler. Even more important is the limiter with the BAT14-099R schottky quad. The latter both shapes the signal and prevents saturating the input of the SDA4212.

Fig.2 - Prescaler.



The SDA4212 prescaler has a standard pinout. The U891 could be used in the same socket, except for limiting the frequency coverage from 20MHz to 1.6GHz. The output signal of the ECL prescaler is amplified by the BFP183 transistor to TTL logic levels.

Most ECL prescalers allow some selection of the divider modulo with pin 5. A divider modulo of 256 was selected since it is available in most prescalers. Dividing by 256 provides an output frequency up to 7MHz at an input of 1.8GHz. Since the following counter allows input frequencies up to 16MHz, the same circuit could directly count both VCO frequencies with faster prescalers like the uPB1505.

The prescaler module requires a single supply voltage +5V. All components are SMD types except for the SDA4212. Experiments have shown that the SDA4212 operates better while inserted in a good-quality IC socket rather than being soldered directly to the circuit board. It seems that the socket contacts provide some additional impedance matching for the SDA4212.

3. Counter

A digital counter is the most obvious component of a digital frequency meter. Of course the content of the counter has to be memorized and displayed at the end. Most frequency counters are equipped with 7-segment LED displays. The latter have a large current consumption leading to generation of large amounts of heat and radio-frequency interference. Both of the latter are highly undesirable in a sensitive piece of test equipment like a spectrum analyzer.

On the other hand, liquid-crystal displays have a low power drain and generate little radio-frequency interference. Multiplexed LCD modules with integrated drivers and controller are certainly the easiest to use. Unfortunately, due to the LCD multiplexing, their contrast is poor and it is relatively difficult to get such displays with large characters.

The described frequency counter is therefore designed around an unmultiplexed LCD with four large digits as used in table clocks. A suitable counter, memory, 7-segment decoder and LCD driver is available in a single integrated circuit ICM7224IPL. With the latter, a number of 40xx or 74HCxxx integrated circuits or a programmed microcontroller are no longer necessary.

The connection of the ICM7224IPL is shown on Fig.3. The counter section is controlled through the inputs COUNT and /COUNT-INHIBIT. The COUNT input is driven by the prescaler. The /COUNT-INHIBIT input enables or stops the counter. On the other hand, any level transitions on the /COUNT-INHIBIT input will not advance the counter inside the ICM7224IPL as long as the COUNT input does not toggle. The content of the counter is copied to the memory with a low pulse on /STROBE. Finally, the counter is reset to zero with a low pulse on /RESET.

Fig.3 - Counter.



Since the LCD is not multiplexed, each segment on the display requires its own connection to the ICM7224IPL. The latter is also driving the backplane (BP) electrode common to all segments. In this way the display driver always applies an alternate voltage of a few tens or hundreds Hz to the LCD. The LCD frequency is provided by an internal oscillator inside the ICM7224IPL.

The ICM7224IPL and the LCD have a few unconnected pins, like the decimal points of the LCD. The ICM7224IPL could even drive a fifth digit (segment AB5) that is not available in the LCD used in this project. On the other hand, the /CARRY output allows the concatenation of several ICM7224IPL and corresponding displays. The LZBIN input and LZBOUT output allow correct leading-zero blanking when several ICM7224IPLs are concatenated. In the described circuit the LZBIN input is not connected, since it is kept in the correct logic state by an internal resistor.

The ICM7224IPL requires a single supply voltage +5V. At +5V supply the counter section is able to operate correctly up to 15...25MHz. In order to save space on the front panel of the marker counter, the ICM7224IPL is installed below the LCD. The single-sided printed-circuit board requires two jumpers (Vdd and BP) below the ICM7224IPL.

4. Time base

Since the marker counter has to be synchronized to the spectrum-analyzer sweep, the design of its time base slightly differs from conventional frequency counters. Besides the synchronization with the spectrum analyzer, the time required by the frequency measurement itself should be considered. If the spectrum-analyzer sweep is reasonably linear, the counter measurement will correspond to the frequency in the center of the measurement interval.

Besides steering the counter gate, memory and reset, the time base also provides MARKER pulses to other circuits to sample the video signal and draw a marker on the corresponding point on the screen. Of course the MARKER pulse should be generated exactly in the center of the counter gate pulse fed to /COUNT-INHIBIT.

The circuit diagram of the time base is shown on Fig.4. The TRIG pulse from the spectrum analyzer resets the counter and triggers the first monostable (first half of first 74HC4538). The latter produces a variable delay to adjust the marker position. After the first monostable times out, the second half of the same 74HC4538 produces a microsecond pulse.

Fig.4 - Time base.



The microsecond pulse starts the time base by resetting the flip-flop in the second 4518 (pins 10, 11 and 15). The flip-flop releases the time-base divider including a 74HC4040 and another 4518, driven by a 10MHz crystal oscillator. The divider produces either 31.25kHz or 3125Hz according to the desired counter gate time for 1MHz or 100kHz resolution.

The decade counter in the second 4518 (pins 2, 5 and 6) waits in the state "8". The incoming clock will move the decade counter to 9, 0, 1, 2, 3, 4, 5, 6, 7 and back to 8. When the state "8" is reached, the flip-flop is triggered and the time-base divider is blocked. The output /COUNT-INHIBIT is enabled between states 0 and 7 for a total of 8 clock periods or 256us (resolution 1MHz) or 2.56ms (resolution 100kHz). After the counter gate is closed, a ten microsecond pulse /STORE is generated to copy the counter content in the display memory (second half of second 74HC4538).

In the middle of the counter-gate period, on the transition from state 3 to state 4 of the decade counter, the first half of the second 74HC4538 is triggered to produce the MARKER pulse. The MARKER pulse width is set to about 30 microseconds to produce a well-visible marker on the oscilloscope screen as well as drive the video-signal sampler.

The time base requires a single supply voltage +5V. The latter also requires a wire jumper under the second 74HC4538 on the single-sided printed-circuit board. All control signals for the counter (/RESTE, /COUNT-INHIBIT and /STORE) are available on a single connector including +5V and ground. The TRIG input is protected with diodes and resistors also in the case when the marker counter is powered off.

5. Sampler

While the amplitudes of all displayed signals can be read easily from the spectrum-analyzer screen, an additional digital display of the signal amplitude at the marker position is always welcome. The spectrum analyzer presented in [1] or [2] has an analog Y-VIDEO output providing 20dB/V. Of course the Y-VIDEO output changes a lot during a single sweep across the screen. To measure the signal level at the marker position, the Y-VIDEO signal has to be sampled at the appropriate point in time.

The circuit diagram of the video-signal sampler is shown on Fig.5. The video signal is first fed through a simple low-pass (22kohm, 68pF) to the voltage follower with the first CA3140 op-amp. The CA3140 avoids loading the video-signal line in all cases: when the sampler is operating and in the power-off condition. The 7V5 zener diode is used to protect the MOS input of the CA3140.

Fig.5 - Sampler.



The sampler includes a transmission gate (built with the MOS transistors contained in the 4007) and a 33nF capacitor. The CMOS transmission gate is turned on only for a very short period by the MARKER pulse, amplified by the 2N2369 transistor to CMOS logical levels. The 33nF capacitor then holds the sampled voltage, followed by another MOS voltage follower with the second CA3140 op-amp.

The CA3140 operational amplifiers have many desirable features. Besides a very high impedance of the MOS inputs, their input and output voltage ranges go down to the negative supply rail. The latter feature allows a single positive supply +12V for the op-amps. The CA3140 also withstands large overloads on the inputs, up to +8V above the positive supply rail. The latter prevents damage to the op-amp in the power-off condition.

The 4007 was selected as the sampling switch because of repeatability. More common CMOS transmission gates like the 4016 or 4066 may have very different electrical performances when supplied by different manufacturers.

The output voltage can be displayed in many different ways. The simplest solution is to use a digital-voltmeter module (DVM) with a similar non-multiplexed LCD as used in the frequency counter. While DVM modules are cheap and easily available, the corresponding user manual usually requires to have the power supply isolated from the measured voltage.

The requirement for the isolated supply for the DVM module can be overcome easily by studying the internal circuit diagram of the DVM. Most DVM modules with a 3-1/2 digit LCD are built around the integrated circuit ICL7106CPL as shown on Fig.6, including those with an unmarked, bare chip bonded directly to the printed-circuit board. Some additional components may be included in some DVM modules to drive the decimal points on the display.

Fig.6 - DVM.



The nominal DVM-module supply voltage is a 9V battery connected between +BAT and -BAT. Internally the ICL7106CPL includes a precision regulator for the 2.8V reference voltage between +BAT and COMMON. Besides being used as a source for the reference voltage between REFLO and REFHI, the regulator also provides a virtual ground called COMMON for all the analog circuits inside the ICL7106IPL. One of the voltmeter inputs (usually INLO) is connected directly to COMMON already inside the DVM module.

The sampler circuit board includes a 7808 regulator to supply the DVM module with +8V on +BAT, while -BAT is connected to ground. In this case COMMON is held at about +5.2V. While designing additional circuits around the DVM module, one should consider that the internal regulator inside the ICL7106CPL can only sink current from COMMON (load between COMMON and +BAT).

DVM modules are usually designed for a full-scale sensitivity of +/-200mV. A resistive voltage divider is required to measure higher voltages, so that the voltage between INPUT and COMMON remains within the +/-200mV range. If the resistive divider pulls current from COMMON, then an additional load from COMMON to +BAT is required, like the 10kohm resistor on Fig.5, to prevent any current being sourced from COMMON.

Since the Y-VIDEO signal is logarithmic 20dB/V, it makes sense to calibrate the DVM scale directly in dB (or tenths of dB). A 0.1dB change in the signal level is represented by a 5mV change in the Y-VIDEO signal. The latter should be divided by 50 to match the DVM steps of 0.1mV (3-1/2 digit full scale +/-200mV). The exact divider ratio is set by the 10kohm trimmer SCALE. The front-panel command OFFSET allows adding an arbitrary offset to the sampled Y-VIDEO value.

The sampler requires an unstabilized supply of +12V for both CA3140s and the 4007. The DVM and OFFSET potentiometer are supplied by the 7808 regulator. The Y-VIDEO input is designed to represent a high-impedance load both during power on and power off.

6. Marker

Numerical frequency and amplitude displays are of little value if it is not known to which signal they refer to. The simplest way to indicate the measured signal is to draw a well-visible marker in the corresponding place on the spectrum-analyzer display. The marker can be drawn in many different ways depending on the accessible signals driving the cathode-ray tube or other display device: vertical (Y) deflection only, both vertical and horizontal (XY) deflection and/or beam intensity (Z).

In the most general case only the vertical (Y) deflection is available. The marker pattern should therefore be added to the Y-VIDEO signal. Of course the marker pattern should be selected in such a way that it is not confused with the typical patterns displayed by the spectrum analyzer. At the same time, the marker pattern should not corrupt much of the information contained in the Y-VIDEO signal.

The marker circuit shown on Fig.7 generates a short vertical line as the marker pattern. The latter extends both above and below the actual trace generated by the spectrum analyzer. The line pattern is generated by adding a few oscillations of a sinewave 455kHz oscillator to the Y-VIDEO signal.

Fig.7 - Marker.



The 455kHz oscillator includes a BF961 MOS transistor as the active device and an IF transformer (white core or AM2) as the selective feedback. The MOS transistor allows both a fast startup of the oscillator as well as a fast shutdown after the power is removed. The 470ohm resistor across the oscillator supply speeds up the shutdown.

The oscillator output is added to the Y-VIDEO signal through the secondary winding of the IF transformer. The circuit is designed to obtain a pattern height of about one division or 0.5Vpp while using a standard IF transformer for 455kHz. When the marker circuit is powered off, the 470ohm resistor loads the primary of the IF transformer so that the circuit has no effect on the Y-VIDEO signal.

The oscillator is turned on or off by the 2N2369 transistor. The marker pattern height is set by the supply voltage with the 10kohm trimmer. Since the duty cycle of the oscillator operation is very low, its supply current comes from the 220uF capacitor. The latter is then slowly recharged through the rest of the cycle through the trimmer and 1.5kohm resistor. Since the +12V supply is not stabilized, the marker height may change slightly with the supply voltage.

7. Assembly of the marker counter

The marker counter is built in the same way as the corresponding spectrum analyzer [1] or [2] and tracking generator [3]. The +5V supply comes from a 7805 regulator as shown on Fig.8. To save some space on the front panel, a three-position switch is used to select the counter resolution (1MHz or 100kHz) or to turn the marker counter off.

Fig.8 - Power supply.



The marker counter includes a single RF printed-circuit board for the prescaler, built with SMD components and shielded in a box of 0.5mm brass sheet with the dimensions of 30mmX60mmX30mm. The prescaler circuit board is shown on Fig.9 and is etched on single-sided, 0.8mm thick glassfiber-epoxy laminate FR4.

Fig.9 - Prescaler circuit board (30X60, 150dpi).

All other circuit boards only carry low-frequency circuits built with standard components with wire leads. The circuit boards are shown on Fig.10 and are all etched on single-sided, 1.6mm thick glassfiber-epoxy laminate FR4. The single-sided boards require a few wire jumpers: two jumpers under the ICM7224IPL in the counter module and one jumper under the second 74HC4538 in the time base. The DVM module is usually available already built, so no special circuit board is required.

Fig.10 - Other (low-frequency) circuit boards (150dpi).

The marker counter module location is shown on Fig.11. The marker counter has the same depth (240mm) and width (220mm) as the spectrum analyzer [1] or [2]. The height is set to 42mm by the available DVM module and counter LCD. The bottom of the box is simply a piece of 1mm thick aluminum sheet, bent in the form of an "U". The cover is a similar "U" made from 0.6mm thick aluminum sheet.

Fig.11 - Marker-counter module location.



The 7805 regulator is bolted to the back plate for heatsinking purposes. The two 470uF electrolytic capacitors, VK200 RF choke and 1N5401 diode are simply soldered between the +12V supply connector and the leads of the 7805 regulator. The overall current drain amounts to about 80mA. The empty space in the box of the marker counter could be used for further additions like a computer interface.


The marker counter is a relatively simple piece of equipment that should operate without any tuning. Some settings are self evident, like the marker height and width. If no marker can be obtained, the tap on the primary of the IF transformer may be too far from the center of the winding and the oscillator may not work at all.

The accuracy of the counter is defined by the 10MHz crystal oscillator. While checking the frequency counter, both the lower and upper frequency limits have to be tested. If the frequency counter displays strange symbols on the LCD at particular counts, then a 15nF capacitor has to be added from /STORE to ground (see Fig.4).

Finally, the scale of the DVM has to be set correctly so that the display is calibrated in decibels. The easiest way is to use the SCALE trimmer in the sampler module and a calibrated step attenuator between the tracking-generator RF output and spectrum-analyzer RF input.


The connection of the decimal points of both LCDs may not be simple, since the active LCD segments should only receive an AC voltage. Any DC component might damage the LCD through electrolytic processes. Some cheap DVM modules have jumpers that connect the decimal points to the COMMON pin of the ICL7106CPL. Due to the lower amplitude, the contrast of the decimal points may not be as good as that of the regular seven segments. In a similar way, the decimal points of the counter display may be turned on by connecting them to ground through a suitable capacitor.

Original full-size drawings


LCD Oscilloscope for Spectrum Analyzers


1. Spectrum-analyzer project 2007 update

Since the development of the wide-band VCO almost 10 years ago, the whole spectrum-analyzer project with all related accessories: tracking generator, harmonic converter, storage-normalizer, marker counter and accessories developed by other experimenters (Darko S57UUD) have been published in many different places: magazines "VHF-Communications", "AMSAT-DL Journal", "CQ ZRS" and the book "Beacon 99" (project Phare, It is reasonably believed that the the project has been successfully reproduced in hundreds of units.

Since the spectrum-analyzer project is still very interesting for many radio amateurs and other radio and electronic experimenters, I decided to republish my original articles in English on my personal web page. Unlike printed magazines, the web allows to publish many color pictures of all interesting details of the project in addition to PDF, PCB and software files. Last but not least, updates to the project are really simple on the web.

In this article a simple and inexpensive LCD oscilloscope to be used as a display for the spectrum analyzer will be presented. Although a small LCD screen is unable replace a good analog oscilloscope, a LCD may be very useful in field measurements under strong daylight conditions, for battery operation or simply when the available oscilloscope is required for a different measurement at the same time.


Before describing the LCD oscilloscope, some necessary updates to the spectrum-analyzer project will be presented. As expected and also according to the feedback, the most difficult part to reproduce is the wide-band VCO. This difficulty is in part due to component tolerances and in part due to assembly (soldering) tolerances. In particular, the BB833 varactor diodes were found to have rather wide manufacturing tolerances both in the capacitance range and in the series (loss) resistance affecting the Q of the varactor.

As already explained in the VCO article, two different printed-circuit boards are proposed for the wide-band VCO. Starting with unknown components, one should always assemble the "narrow-stripe" wide-band VCO first (PCB#1). The "wide-stripe" PCB#2 should only be used if the frequency coverage obtained with PCB#1 is insufficient.

If output power of the wide-band VCO drops in the middle of its frequency range between 2.5GHz and 3GHz or the VCO simply stops oscillating at all in this frequency range, excessive losses in the BB833 varactors are suspected. Besides poor varactor Q (try getting better BB833 varactors from a different source), the interdigital feedback filter may be misaligned due to soldering tolerances. In the latter case, a rather simple solution is to try to trim the length of the open end of the central finger of the interdigital filter.

Although the wide-band VCO design was also published in a well-respected and widely-known professional magazine "Microwave Journal" already back in 1999 [1], it took several years for the professionals to fully understand the operation and capabilities of this VCO design. Multiple-resonator-multiple-varactor microwave VCO modules only became available recently on the professional market. Finally, there is a valid and inexpensive alternative to the YIG oscillator! Of course, these inexpensive commercial VCO modules can be used in place of the described wide-band VCO.

After using the described spectrum analyzer and all related accessories for many years, some long-term effects were noted as well. The worst seems to be the corrosion caused by outgassing chemicals from the antistatic foam used as a microwave absorber in high-frequency shielded modules. A higher conductivity foam makes a better microwave absorber, but unfortunately more chemicals cause more corrosion as well! Fortunately, the microstrip circuits of the spectrum analyzer continue to work correctly with little if any degradation, although they really look ugly...


A less-frequent problem is a long-term failure of ATF35176 or similar HEMT devices. If the voltage between drain and source is kept too high, the drain current slowly decays. This decay may be very slow, just a few percent per week, but it is cumulative, irreversible and continues down to zero leading to a total failure of the circuit!

In this project, HEMTs are used in the buffer stages of wide-band VCOs and in the buffer amplifiers inside the tracking generator. The described failure can be detected as a rise of the drain voltage. If the drain voltage rises above +3V...+3.5V, the decay will speed up and the HEMT will have to be replaced soon, possibly with better devices.

Some electronic parts became obsolete or hard to get. The most difficult seems to be the INA10386 MMIC amplifier that has no direct replacement. Worst of all, defective factory rejects are shipped for some obsolete parts, like the uA723 voltage regulator. Some uA723 regulators have a very noisy voltage reference with very large 1/f "popcorn" noise. The latter may be large enough to disturb even large-signal circuits like the video amplifier inside the spectrum analyzer.

Besides getting better uA723 regulators, there is a simple solution to this problem. The internal 7V voltage reference, available on pin 6, should be filtered before use. If the latter is fed through a resistor to pin 5 (non-inverting input), then a single electrolytic capacitor from pin 5 to ground (pin 7) solves the problem.

2. LCD oscilloscope for spectrum analyzers

LCD modules are probably the most popular displays today, ranging from simple numerical displays with few single-color digits to large, graphical, high-resolution, full-color computer monitors. A medium-resolution, single-color graphical LCD module is required in an oscilloscope display for a spectrum analyzer. Of course, a simple solution was sought for the described spectrum-analyzer project.

Medium-resolution graphical LCD modules may have different interfaces. The simplest LCD modules have no built-in controller. These modules only have shift registers associated with the columns and rows of the display. The user has to provide a continuous data flow to refresh and multiplex the display content. This requires a powerful microprocessor with lots of memory or in other words a complicated circuit with many chips.

LCD modules with a built-in controller providing the refresh and multiplex of the display are much simpler to use. The most popular alphanumeric controller is certainly the Hitachi HD44780, that even has some very limited graphical capabilities. Older graphical LCD modules use the Toshiba T6963 controller with an external 8kbyte RAM (usually 6264). These modules also require a negative-voltage supply for the LCD and a high-voltage AC source for the electro-luminescent backlight.

Recent graphical LCD modules use the Samsung KS0107 and KS0108 chips. The KS0107 is the clock generator, scans 64 rows and may drive multiple KS0108 chips. The KS0108 drives the columns and includes storage for up to 64x64 dots. The microprocessor simply writes into the RAM inside the KS0108 chips. These modules usually include one or two 7660 chips to generate the required negative Vee LCD supply of -5V or -10V on-board the LCD module. Finally LEDs are used for the backlight so that the module can be operated from a single +5V supply.


A 128x64 graphical LCD module is used in the described LCD oscilloscope. This module includes one KS0107 and two KS0108 chips. All three controller chips are usually bonded directly to the printed-circuit board and covered with drops of black resin.

The LCD oscilloscope is built around a PIC 16F876A microcontroller. The latter includes an A/D converter and steers the graphical LCD module directly. The circuit diagram of the LCD oscilloscope is shown on Fig.1.

Fig.1 - LCD oscilloscope for spectrum analyzers.





The LCD oscilloscope requires two signals from the spectrum analyzer: the analog video (including blanking) and the trigger. Due to the limited resolution of the LCD, the analog signal is oversampled. Each column on the LCD is computed from eight consecutive samples and the last sample from the previous column. The A/D sampling rate is set so that the whole sweep (128 columns or 1024 samples) corresponds to about 20ms or in other words the fastest rate of the described spectrum analyzer.

Additional inputs to the microcontroller are provided for two switches and one pushbutton. One switch is used to turn on the grid while the other selects a 80dB (4V) or 40dB (2V) full-scale range. The scale is adjusted by two trimmers defining the reference voltages of the A/D converter. The pushbutton activates a MIN/MAX memory that is very useful when observing wide-deviation slow FM signals or infrequent pulsed signals on the spectrum analyzer.


All outputs of the PIC 16F876A have series 1.8kohm damping resistors to reduce the amount of radio interference radiated by the microcontroller. The same resistors are also used to operate the data bus to the LCD module bidirectionally, simplifying the programming of the PIC 16F876A. In particular, the RC0-7 pins are always programmed as outputs. In order to read the status of the KS0108 controllers, the DB7 line is also fed back to the RB1 input.

The display-refresh rate is limited by three factors: the A/D-conversion speed, the PIC computing power and the wait cycles required by the KS0108 controllers. In the described LCD oscilloscope, all three factors are of the same order of magnitude limiting the refresh rate to about 35Hz...40Hz and thus providing a live picture of the input signal. A minimum clock frequency of 20MHz is required for the PIC 16F876A for this purpose.

A 24MHz clock for the 16F876A was therefore selected to allow the spectrum-analyzer sweep time to remain at 20ms at all times with some safety margin. This is consistent with the resolution of the display (128 columns) and available filter bandwidths and spans of the spectrum analyzer. Slower signals can be observed with the MIN/MAX function, using 256 bytes of the internal RAM inside the 16F876A. The minima and maxima are simply accumulated as long as the MINMAX pushbutton is kept depressed.


The same 256 bytes of internal RAM are also used for 128 minima and 128 maxima storage during normal operation without the MIN/MAX function. The A/D routine runs under interrupts triggered by an internal timer inside the PIC 16F876A. The main program is an endless loop computing and refreshing the LCD content from the intermediate storage. Yet another interrupt is used for the trigger function.

The described LCD oscilloscope is fully compatible with all described accessories: tracking generator, storage-normalizer and marker counter. The 40dB scale is in fact intended for reflection measurements with the tracking generator and an additional, external directional coupler. The oversampling of the video signal allows a proper display even of the 455kHz marker.


The LCD oscilloscope includes its own +5V regulator (78L05). The latter also provides power to the backlight LEDs through two 10ohm current-limiting resistors. Current-limiting resistors and other protection components are also provided on all inputs to the module.

3. 230V mains power supply

Since the described LCD oscilloscope is small and simple, the remaining space behind the LCD module is used for a 230Vac mains power supply, providing 12V for the LCD oscilloscope, spectrum analyzer and all accessories. The circuit diagram of the power supply is shown on Fig.2.

Fig.2 - 230V mains power supply.



The discrete-component regulator has many advantages over integrated circuits: very low drop-out voltage, grounded heatsink of the power device and last but not least, no additional protection diodes are required in the case of (parallel) 12V-battery operation! Since the spectrum analyzer is sensitive to low-frequency magnetic fields, it is imperative to use a toroidal-core transformer in the power supply.


4. Assembly of the LCD oscilloscope

The first thing to do is to check the availability and size of suitable LCD modules with the KS0107/KS0108 controller chips. Do not forget to check the presence of the 7660 Vee generator as well as the type of backlight available. Last but not least, check the connections of the available LCD module!

The printed-circuit board was developed for Wintek LCD modules. Other manufacturers may use a different pin-out requiring to swap some wires. In particular, the Vdd (+5V) and Vss (GND) connections may be swapped! CS1 and CS2 may be active-low or active-high, since each KS0108 actually has three such inputs (two inverted) and just one is brought out to the connector.


The LCD oscilloscope is built on two printed-circuit boards as shown on Fig.3. The PIC 16F876A and associated components are installed on a single-sided printed-circuit board with the dimensions of 80mmX60mm. The power-supply regulator is built on a single-sided printed-circuit board with the dimensions of 80mmX40mm. Both circuit boards are etched on 1.6mm-thick glassfiber-epoxy laminate FR4.

Fig.3 - Printed-circuit boards (150dpi).

The LCD oscilloscope has the same depth (240mm) as the spectrum analyzer [2] or [3]. The width is 110mm while the height is set to 70mm considering the size of the available LCD module. All modules are located in a single plane. The bottom of the box is simply a piece of 1mm thick aluminum sheet, bent in the form of an "U". The cover is a similar "U" made from 0.6mm thick aluminum sheet.


Besides the DIN connector, two additional BNC connectors for the video+blanking and trigger signals are provided on the back panel to connect an additional, high-quality, analog oscilloscope. Although the LCD oscilloscope is protected from over-voltages on all inputs even in the power-down state, the LCD oscilloscope should be powered at all times to avoid loading the outputs of the spectrum analyzer.


Besides the LCD module, two switches (grid on/off and scale 40dB/80dB), the MINMAX pushbutton and the LCD-contrast potentiometer are located on the front panel. The latter could be omitted, since LCD modules from many different manufacturers were found to operate best at the maximum voltage. This means that the Vo input is simply tied to the Vee output through the 82kohm resistor on the printed-circuit board and no additional potentiometer is required.


The described LCD oscilloscope will display correctly input voltages between about +0.7V and +5.5V due to the emitter follower in front of the A/D converter. Therefore the trimmers inside the video amplifier of the spectrum analyzer [2] or [3] may need to be readjusted first, using a reliable analog oscilloscope. The trimmers inside the storage-normalizer and marker counter may need some minor readjustments as well.

Finally, the 80dB scale of the described LCD oscilloscope can be adjusted with the two trimmers providing the reference voltages to the A/D converter. Please note that the operation of the LCD oscilloscope is slightly different between the 80dB mode and the 40dB mode. In the 80dB mode, the trace is always visible and saturates on the bottom or top of screen. In the 40dB mode, the trace runs out of the screen and only the central part of the original 80dB scale is displayed.

Datasheets, PCB files & PIC software


* * * * *

< Prev   Next >
Who's Online
Buy microwave components on-line

© 2017 Greek Microwave Group - Official Site

Get The Best Free Joomla Templates at