Design and Realisation of Microwave Circuits, Part-4

Gunthard Kraus, DG8GB

Design and Realisation of Microwave Circuits, Part-4

VHF Communications 4/1997

7. MMIC AMPLIFIER

7.1. Preliminary Observations

At two points in our "Meteosat Converter" application project, we need at least 20dB amplification with simultaneous tolerable noise background:

a. Directly behind the "low-noise pre-amplifier" referred to in Section 6.2. of Part-, for the significant increasing of the signal level before the subsequent 1,700 MHz band pass, with simultaneous damping of the passive mixer.

b. Directly behind the mixer, or in the downstream intermediate-frequency stages to have a signal level sufficient for a relatively long coaxial cable connection from the converter (near the antenna) to the evaluation circuit in the building.

The use of a "universal amplifier module" from 50 Ohm technology is on offer here; to this end, MMIC's (microwave monolithic integrated circuits) are available in the most varied formats, and with varying limiting frequencies.

After a look through the relevant literature with application circuits, the choice finally fell on type INA 03814 from Avantek/HP, because of the current price and delivery situation. We find the following specification for this MMIC in the data sheet; 25dB amplification in the range up to 2 GHz, for noise factors of about 3dB. The S-parameter can also be found, on the Avantek data diskette, which can be supplied for Puff.

Here you merely need to convert the "touchstone file" (INA03184.S2P) into Puff format (which gives INA03184.dev). As mentioned, the conversion program you need for this is supplied with the Puff diskette. But don't forget to estimate the values for the "zero Hertz" frequency and add them to the *.dev file!

7.2. Draft Circuit and Simulation

To judge from what many semi-conductor manufacturers say, drafting a microwave amplifier circuit is child's play, and is exceeded in simplicity and operational reliability only by the drafting of a power supply circuit with a half-wave rectifier.

The input and output pins are simply connected using 50 Ohm microstrips, into which are inserted the DC blocking capacitors. The power supply DC voltage is fed in through a 47nH SMD choke. If, during the simulation, the same equivalent circuits are used for chokes and capacitors as for the low-noise preamplifier, you obtain a wonderfully smooth amplification curve between 1 and 2 GHz, with a peak value of app. 27dB. Fig.25 shows the MMIC amplifier simulation.

However, in reality the assembled circuit will initially oscillate at f = 1.4 GHz. Moreover, it is almost impossible to produce any attenuation by means of the well-known "calming measures" - e.g. using conducting foamed material, etc.

If we now make use of the experience gained in the drafting of the LNA and test the influence of the "earth feed-throughs", using Puff, we discover something really amazing: Even a total inductance of 0.3nH, applied for the feed-through of the two earth surfaces of the IC to the bottom earth plane, is sufficient to make the input reflection factor in the frequency range at 1.4 GHz markedly greater than 0dB (Fig.26 shows the ratios for 0.5nH). The circuit then has a negative input resistance. In these circumstances, it can't help but oscillate! We must therefore use all means to ensure that this inductance falls; Fig.27 shows how this is done.

a. A hole with the external diameter of the MMIC housing is drilled in the printed circuit board.

b. The top earth surfaces are now fed as a wedge into the layout until they are directly against the housing of the MMIC. c. The feed-throughs are started as close as possible to the housing. 5 holes are drilled on each side.

It can probably be easily recognised from the diagram that the question of the insulated earth surfaces for input, earthing and output has again been taken to heart (it's just that putting in the many 0.8 mm. rivets, with a "watchmaker's power riveting machine", specially created for this, as an extra, that really is a brute of a job!) The output and input microstrips require special consideration at 50 Ohms. They should have a "wedge-shaped point" , and then be reduced, right down to the width of the IC terminal lug, thus avoiding additional inductances and points of impact.

The already well-known 47nH choke for feeding in the operating voltage is here too as well, together with its 6mm. long feed and its joint built-in capacitance of 0.46pF and the 13 Ohm losses in the equivalent circuit diagram. The coupling capacitors in the input and output circuits are laid out, as usual, as two 0805-NP0 types in parallel, at 100pF. Firstly, this precisely fills out the width of the microstrips and thus reduces the reflections, and secondly, the joint inductance of 1.5nH and the serial loss resistance of 1 Ohm are halved by using one component.

Fig.28 displays the full results of the simulation, including the measured results for up to 4 GHz. Of course, immediately after the measurement the simulation was adjusted to the reality.

The onset of amplification at 3.6 GHz is actually brought about by the 6mm long microstrip and the 47nH choke, with the (assumed) 0.46pF capacitor! Here a short-circuit is obviously transformed to the amplifier output, and we must reduce the 0.46pF to 0.39pF for the Puff simulation to supply the same resonance frequency.

There is a quite simple reason why the minimum value in the actual assembled circuit is not as low as in the prediction. It is well-known that the losses for FR4 equipment increase more and more sharply from 1.5 to 2 GHz, and here we are always stimulating with a constant loss factor of 0.015. The assumed quality is therefore too high, and there is a discrepancy between reality and the Puff result.

Anyone who would like to have a more precise result would first have to measure the valid loss factor for this frequency, and then simulate for only a narrowly restricted frequency range for which this determined factor is approximately constant. For the same reason, the measured amplification in the range between 1.5 and 3.4 GHz is probably always somewhat lower than the theoretical progression, since - because of the poorer quality - as in every resonant circuit, the resonance curve of this "resonant circuit" must be flatter and broader.

One more little thing:
The S parameters of the INA 03184 are specified by the manufacturer only up to a maximum frequency of 4 GHz. But if, for example, an upper plot limit of 10 GHz is inadvertently entered, a bleeping noise is suddenly heard when the 4 GHz limit is exceeded, which in normal circumstances would be the acoustic signal for a computer fault. But the program has not crashed. It continues operating and bleeping, right up to the highest frequency entered, and you need only wait for the computer to quieten down. The diagram created using a network analyser (HP 8410A) shows the simulation behaviour of the circuit up to 12 GHz. Finally, Fig.30 shows the wiring diagram for the layout, which can certainly be analysed without problems.

(To be continued)