Amateur Radio Archive

Modular Microwave Breadboard Svstem

• Home Made Energy: Does It Work?
• Blood Boiling Karate Drills And Exercises

One potential disadvantage of the drop-in assembly methods used for the modspiitter and tow-band microcircuits is that there s no easy way to test the individual circuits within these hybrids With the former design method of placing each circuit in its own package, this was not a problem ir- the new desigr however the circuits are hioden away inside a final large package and it may not even be possible to gain access to both ports of all of the circuits A new technique was needed to test the circuits during development and in production

Fortunately, during development of the microcircuits, the HP 83040 modular microcircuit package was also being developed This microwave modular breadboard package consists of a series of end blocks with built-in coaxial-to-microstnp launches and center bodies of various lengths for circuits to be tested. Included is all of the hardware needed to connect the various blocks. The end blocks and center bodies in this scheme all share a common mechanical interface, which is genderless so that they can be connected in any order and in any quantity. Fig 1 shows examples of modular breadboard parts.

Using this sysiem in designing the multistage low-band microcircuit each separate circuit (or circuit element) was developed and tested in the final microstrip environment Then the individual blocks were assembled into a full circuit. The entire low-band microcircuit was developed using such a scheme, and the initial fully working breadboard was built using il

Advantages of the HP 83040 modular microcircuit package include:

■ It offers microcircuit breadboard packaging that is very cost-effective compared to custom packaging and saves time because it is available off the shelf.

■ All microstrip circuitry is measured directly in microstrip envi-

Fsg. 1. Examples of microwave breadboards using the HP 83040 modular microcircuit package.

ronment.

■ It provides true zero-length through calibration since the end blocks can be directly connected together Deem bedding can be used to make precise microwave measurements.

■ Entire microcircuits can be measured and assembled in any combination of subcircults.

Stan Bischof

Development Engineer Network Measurements Division

100% of the time, so no power is lost, as it is in a single balanced mixer, which throws away power during half of its cycle.

This type of mixer can be described as a biphase modulator. Ideally, il performs a time-domain multiplication of the KF signal with a square wave at the frequency of the LO drive. In the frequency domain, this yields a convolution of the RF signal spectrum with the spectrum of the square wave drive, which contains the fundamental and odd harmonics of the drive frequency. Of the many terms in the convolution, the principal ones are the fti|.- fLa and fRF + fLO terms. The fttJ. + f[j0 term is filtered out so that the primary term left is ihe — fm term, which is exactly what is wanted for the IF.

The eight diodes are grouped into two quads, which are mounted on the top and bottom of the mixer substrate in a symmetric pattern (Fig. 5).

Fig. ü shows some of the details of the RF balun. Ideally, as described above, the RF path is switched by the LO drive such that the signal is directed to the IF output in either 0° or 100° phase orientation. Before this happens, the signal is split into two halves, relying on symmetry to ensure that they are equal in magnitude and phase. This is accomplished by two transitions. First, the signal is transformed by a step transition from microstrip line to a coupled suspended microstrip format consisting of topside and bottom-side electrodes in a symmetric arrange ment on the lop and bottom of the substrate. Then, on each side, two 100-ohm lines are split off in another sfep transition. This leaves a total of four leads, two on each side of the substrate. Each lead is connected to a diode swritch. The four diodes in each quad are always connected to the IF output in pairs and the IF impedance is 50 ohms, so the RF input port, in principle, is a good 50-ohm match. In practice, of course, none of the transitions is perfect and

Upper Diode Quad

_
	31—*--"nlX
		---— — -------

Lower Diode Quad

Ftg. 5. Diode quads in the triple balanced mixer

Lower Diode Quad the diodes are not ideal switches.

Each balun line is a quarter wavelength long at the lower frequency of 5400 MHz. It is important that all the lines be as close to the same length as possible so that the phases match. At 5400 MHz, 90" of phase shift corresponds to roughly 4400 micrometers in this medium. Thus, to get phase match of, say, ±5°, the paths must match within £ 250 ¿¿m. or about one circuit thickness.

Fig. 7 shows the LO balun, which is similar in concept to the RF balun. It, too, is formed of a microslrip-to-sus-pended transition followed by a split into two paths of suspended balanced microstrip line. In this case, however, the impedances are different. Since the LO provides a large enough signal to turn the diodes on fully, each of the two arms of the balun is effectively terminated in two series diodes. Since each diode has a large-signal impedance of about seven ohms, two diodes in series represent around 15 ohms. For an adequate match, the balun arms are used as quarter-wave transformers to make each arm look like about 100 ohms at its input. To transform 15 ohms to 100 ohms takes V 1 5X1 00 or approximately 40 ohms. Hence each arm is designed to have an impedance of 40 ohms.

Because of symmetry, the RF and IF paths in the LO balun do not, in principle, load down the ends of the balun arms, so they need not be considered in the above analysis. As in the RF path, there are four leads, which are properly terminated. To maintain symmetry, the arms are carefully constructed to have the same length. A balun is used at the IF port to provide a single-sided output to drive a single-sided filter. The high output level of the mixer allows the gain of the IF amplifier to be 10 dB less than in previous designs. This lower gain gives a corresponding improvement in I he broadband noise of the low-band microcircuit.

IF Amplifier

The IF amplifier of the low-band microcircuit is a three-stage FET amplifier with 30 dB of gain and 24 dHm of output power. The design is a lower-gain version of ihe IF amplifier used in the HP 8753A/B/C spectrum analyzers.

Fig. 6. Details of the RF balun in the triple-balanced mixer

The IF amplifier makes possible a +17-dBm instrument power specification, which is needed for the high-power version of the HP 8360, while maintaining the harmonic performance.

Quasi-Elliptic Low-Pass Filters

Throughout the low-band microcircuit, a number of low-pass filters with very high performance are needed. They need sharp cutoffs and very broad stop bands. At the same time they must be low in cost, repeatable, and as small as possible. To do Ibis on microstrip. a new filter was developed. The design uses spiral inductors and radial transmission lines, and has stop-band zeros to extend the stop band. In most cases, the filters are less than a quarter wavelength long at the cutoff frequency, so they are quite small both physically and electrically.

One filter of this type, used in the RF path, is designed to pass 5.4 GHz. Requirements for this filter were that it attenuate the second, third, and fourth harmonics by 40 dB or more, and that it cut off as rapidly as possible above 5.4 GHz. This low-pass filter is a microwave implementation of a simple Chebychev eleventh-order, 0.1-dB ripple, L-input ladder filter, with one major difference, as described below. As such, it consists of Ihe microwave equivalents of six series inductors and five shunt capacitors. The inductors are all realized using spiral coils of high-impedance line. The capacitors are realized using radial transmission lines.

This circuit differs from the conventional Chebychev design in its behavior outside of ihe passhand. While a Chebychev design has all its transmission zeros at infinity, this design places transmission zeros at finite frequencies within the desired stop band. This is accomplished by using the radial transmission lines as shunt resonators. Because these transmission zeros are at finile frequencies, equal ripple is not achieved in the stop band, as it is with a true elliptic filter. Therefore, this filler can he described as a quasi-elliptic low-pass filter.

The reason for using a quasi-elliptic luw-pass filter can be seen by looking al the response of a typical distributed

Fig. 7. Details of the LO balun In the triple-balanced mixer

filter. While the passband can be made to look nearly ideal, the distributed elements lead to secondary- responses, which create secondary pass bands. This effect differs from filter to filter, but typically at around three times the cutoff frequency there is another passband where the response is not much below the response in the main passband. The quasi-elliptic tow-pass filter places zeros within and near this secondary passband and is able to keep the response low- at much higher frequencies than would otherwise be possible.

The inductor design is shown in Fig. 0. It consists of an inner bonding pad, a length of high-impedance line arranged Ln a spiral, and a bond wire connecting the inner end of the spiral to another bonding pad.

The ground plane is cut away from beneath the inductor out to a distance of one substrate thickness. This raises the line impedance, or equivaiently. removes some of the shunt capacitance of the spiral. This accomplishes tw:o things. First, since the line impedance is higher, the line need not be as long to realize a given inductance. Second, since the shunt capacitance is lower, the self-resonant frequency of the inductor is higher, rendering the filter response useful to higher frequencies.

These spiral inductors are designed using an empirical formula based on the particular geometry and material used. The quasi-elliptic low-pass nature of the filter is not influenced by the inductors, but is rather set by the radial transmission line used as capacitors. Each of the five shunt capacitors is realized by a set of two radial transmission lines of the type shovvn in Fig. 9.

Like a shunt stub, a radial transmission line resonates at some frequency to transform its open end into a short circuit, thereby forming a transmission zero for the line it is connected to. While the shunt stub transmission line resonates at the frequency where its length is a quarter wavelength, the radial transmission line resonates at a frequency where its length is roughly one sixth of a wavelength, Thus the radial transmission line is more compact. A key advantage of a radial transmission line for this application is that it is pointed at one end, so 11 forms a simple, well-defined connection, unlike a typical shunt transmission line.

The length of each radial transmission line is chosen for

Radius

Radial Transmission Line

Fig. 9. Radial transmission line resonator used in the quasi-el-hptio tow-pass //iters.

Radial Transmission Line

Fig. 9. Radial transmission line resonator used in the quasi-el-hptio tow-pass //iters.

its resonant frequency, and the angle is chosen for the parallel-plate capacitance needed to realize the required shunt capacitance for the Chebvchev filter. As Fig, 9 shows, all ten radial transmission lines are different. This provides twice as many zeroes to work writh and avoids the situation of two nearly identical resonators being connected together Using one resonator at each node also works, but this halves the number of usable zeroes.

Since this filter is filtering a narrow-band signal, the transmission zeros are clustered in narrow bands around the harmonics of the RF (i.e.. at ID.8, 16,2, and 21.6 GHz). Fig. 10 shows a full quasi-elliptic low-pass filter.

Modsplitter

The modsplitter makes a number of contributions to overall system performance. Fig. 11 shows its block diagram. The modsplitter provides the LO input to the low-band microcircuit. Extending this frequency to 7.8 GHz, higher than in previous designs, allows the low-band microcircuit to improve spurious performance. The higher drive frequency moves the 2/1 spurious response [the 2fRp - fjxt mixer spur) out of band, where it can be filtered. One of the contributors to low-band video feedthrough is eliminated by locating the modsplitter's pulse modulator after the low-band output. This eliminates the possibility that

Ground Plane Relief

Bond Wire

Bond Wire

Fig, 8. Details of the spiral inductor used in the quasi-elliptic tow-pass fitters.

Fig. 10. Fifteen-pole quasi-elliptic tow-pass filter

0 0