By Bruce Carter, Texas Instruments

For decades, successful RF designs have employed discrete transistors. And this still holds true today, especially in cost sensitive designs. But if performance, rather than cost, is the driving factor of an RF design, op amps more than hold their own against transistors. Op amp-based RF circuitry is easier to design and has less risk associated with it. Furthermore, "tweaking" of the RF design in the lab can be almost entirely eliminated.

For high performance RF equipment, using high-speed op amps has some distinct advantages over discrete transistors:

• When discrete transistors are used, the transistor's bias and operating point interact with the stage's gain and tuning. With op amps, the bias point is independent of gain and tuning.
• Op amp stages can operate over a wide range of frequencies because there are no inductors to worry about.
• When transistors are used, parameter drift and beta variation must be taken into account over the system's operating temperature range. When op amps are used, drift is almost entirely eliminated because gain is determined in stable, passive components.

Choosing the Op Amp
When selecting op amps for RF applications, a decision must be made about whether to use voltage or current feedback amplifiers. Voltage feedback amplifiers are limited by gain vs. bandwidth issues imposed by their internal compensation capacitor. This limits their use in RF circuits to high gain, relatively low frequency circuits.

Current feedback amplifiers, on the other hand, are often believed to have no limitation of gain vs. bandwidth. Supposedly, they are usable to almost their specified −3 dB frequency at just about any gain. This is only partially true. While there is no "−20 per decade" slope to their gain/bandwidth curve, current feedback amplifiers most certainly are bandwidth-limited at higher gains. Internal parasitics do take a toll on the bandwidth!

Current feedback amplifier stability is determined largely by the feedback resistor's value. This value, or narrow range of values, is often fairly low, in the order of a few hundred ohms at most. This limits the useful voltage gain range to a relatively low 10V or 12V in a single stage. Nevertheless, current feedback amplifiers are the component of choice for broadband amps.

Broadband Amplifiers
For this article, the THS3202 was chosen for its wide bandwidth and fast slew rate. A THS3202 evaluation board makes a convenient platform on which to construct the circuits presented here. Key questions are: Just how much gain is an op amp-based circuit capable of, and over what frequency range? The circuit in Figure 1 was used to answer these questions. It was first configured as a single stage, consisting of device A of a dual package, a 301W feedback resistor, a 16.5W gain resistor, and a 49.9W back termination resistor. This configuration produces an amplifier voltage gain of 20 and a stage voltage gain of 10 when connected to a 50W monitoring device.

Note the simplicity of this circuit compared to traditional RF circuitry. Provide the op amp, termination and decoupling components, and two resistors and the circuit is done! The 301 W (Rf) and 16.5 W (Rg) resistors are all that are required to set the stage gain. In fact, one of the strong points of an op amp-based design is that the stage's gain can be set precisely by the resistors alone. This circuit produces the lower amplitude curve shown in Figure 2.

The voltage gain of the op amp stage itself is 20, but this is cut in half by the back termination resistor's action in combination with the load. The RF amplifier's −3 dB point is about 390 MHz. If a flat gain over frequency is required, this circuit is usable only to about 200 MHz. Input and output VSWR values are better than 1.01:1 for most of the bandwidth, only degrading to about 1.1:1 near 200 MHz. S21 is −75 dB over most of the bandwidth, only degrading to −50 dB near the bandwidth limit.

You might be wondering if more gain could be coaxed from the stage by lowering the gain resistor (Rg) even more. The answer is yes, but there is a practical limit. Remember that the feedback resistor (Rf) is a large determining factor for current feedback amplifier stability. Remember also that Rg has to drop proportionally more. You can see that it would not be long until the Rg value becomes impractically small.

Lab tests were done with various Rf values. The results indicated that there is no advantage to making Rf smaller than 200W. Below that, peaking starts to occur, regardless of the Rg value, and becomes worse and worse as the resistance is made lower and lower. This is exactly what one would expect, because a designer working with current feedback amplifiers cannot make Rf a short.

More gain requires cascading multiple stages of THS3202 op amps. Since the THS3202 is a dual device, a two-stage RF amplifier is easy to implement at very little additional cost.

To convert the amplifier to two stages, the feedback resistor is lowered to 200W and the gain resistor lowered to 10.5W. A second stage is then connected to the first, using the same values for the feedback and gain resistors. Isolation is accomplished by using inter-stage termination resistors. The optional 39 pF capacitor provides peaking to compensate for some high frequency roll-off. Unfortunately, it also creates a capacitive load on the first amplifier output. This increases the first stage's tendency to peak as seen in the upper curve of Figure 2. This indicates a tendency toward instability, and it manifests itself by poorer IP3 performance. If maximum IP3 performance is a must, the designer should delete the capacitor and live with less bandwidth from the stage. The other S parameters for this circuit are similar to the single op amp's case.

At this point, it is important to mention signal levels. All of the above points in the direction of impressive performance, but if these benefits apply only to very small signals they are hardly benefits at all!

The signal level a designer can pass through an op amp is determined by its input and output voltage rails, as described by the device's data sheet. The rails form a set of high and low voltage "hard clipping" points for the signal as it passes through the operational amplifier. Consequently, the −1 dB compression point occurs very soon after any voltage rail limitation has been breached. The wise designer will not attempt to squeeze that last dB from the stage because the hard clipping points may produce substantial harmonics.

For a THS3202 stage, the amplifier's output can swing ±3.2V, therefore the output of the stage can swing ±1.6V. This corresponds to 14 dBm of output power.

IF Amplifiers
The gain circuit shown in Figure 1 can easily be cascaded with SAW (Surface Acoustic Wave) filters to form high-performance intermediate frequency (IF) stages. The only design consideration is the insertion loss of the filter, which may not be a constant value from part to part, or from batch of parts to batch of parts. If precise gain is needed from the stage, the designer may need to include a trim resistor in one or both stages. This trim adjustment will not affect the stage's tuning except for a slight effect on its upper frequency limit.

It was a simple task to cascade the dual wideband RF amplifier with SAW filters. We used an EVM board from Sawtek (www.sawtek.com), which greatly simplified prototyping. A short SMA-to-SMA cable connected the two boards. It is important to place the SAW filter element after the gain stage so that noise generated in the op amp circuitry is filtered with the same response as the signal. If the gain stage were placed after the SAW filter, the amplifier stage's broadband noise response—instead of a filtered, narrow band response—would be passed to the next stage.

As for applications, 70 MHz and 140 MHz IF amplifiers find use in cellular telephone base stations and satellite communications receivers. A Sawtek 854660 filter was selected for 70 MHz and a Sawtek 854916 filter was selected for 140 MHz. These filters require input and output inductors and operate with standard 50W input and output. A 70 MHz SAW filter produced the upper response curve and the 140 MHz filter produced the lower response curve shown in Figure 3.

These response curves are almost identical in shape to the curves provided by Sawtek! Although narrow band response is shown in Figure 3, there was virtually no harmonic content when the broadband response was examined. In other words, the amplifier is providing gain, but not adding undesirable harmonic content. The insertion loss of the 70 MHz SAW filter is about 7 dB. The insertion loss of the 140 MHz SAW filter is only 8 dB, but the gain circuit itself starts to roll off at this frequency, accounting for the rest of the loss. Careful examination of the lower passband shows the slight roll-off due to the broadband stage characteristic.

Conclusions
Although inexpensive RF design continues to be the exclusive domain of transistors, there is a class of RF applications where performance, not cost, is the driving factor. These applications stand to benefit tremendously from the excellent RF performance op amps can provide. By freeing the designer from the interrelated tasks of calculating biasing, gain, decoupling and peaking, op amps simplify RF design, even for novices. Troublesome components such as inductors and variable peaking capacitors are eliminated, and circuit gain depends on relatively stable resistors instead of the transistor's widely variable parameters. This makes RF design with op amps repeatable in production and eliminates trimming and aligning test stations when the product is manufactured. Equipment maintenance is also simplified, as no periodic adjustments are needed to maintain top-level performance.

Bruce Carter is an Applications Engineer for Texas Instruments. He can be reached by e-mail at r-carter5@ti.com or by phone at (214) 480-4176.