# Classy Amplifiers Simplify EMC Test

Feb. 1, 2009

Amplification isn’t a difficult concept. You start with a small RF or microwave signal and make it into a large one. Ideally, you wouldn’t be able to distinguish between the two signals except by their size. Indeed, a power amplifier makes the input larger, but in doing so adds distortion.

For many types of analog amplifiers, you can approximate the slightly nonlinear output by a power series. And, because good amplifiers introduce a relatively small amount of distortion, it’s common to use only the fundamental and third order terms of the series.1

where: V = input amplitude
O = output amplitude
G = gain
D3 = third order distortion
Substituting a trigonometric identity for cos3 results in

In this form, the gain reduction caused by the third order term is easily seen as well as a separate third harmonic term. Only the first order and third order parts of the coefficient of the fundamental frequency term are relevant.

TOI is another name for the third order intermodulation intercept point. For a signal that contains several frequencies, third order nonlinearity causes mixing. That is, new signal components are formed at frequencies that are the sums and differences of the original frequencies. This can be very important in a narrowband receiver where intermodulation components can fall within the passband even though the original signal frequencies do not.

Even for a single input frequency, Equation 2 shows that third order distortion creates a spurious signal at the third harmonic. That term is ignored in a discussion of power amplifier operation but would not be for a sensitive receiver.

Typically, the relationships between the parts of the fundamental term are plotted as power out vs. power in on a log-log graph. Figure 1 shows the performance of a hypothetical amplifier to demonstrate how Equation 2 works. If the amplifier were perfectly linear, D3 = 0, and the output in dBm is a fixed gain above the input. The straight line relating output to input has a slope of +1. The power of the third order term is shown as a straight line with slope +3.

Figure 1. Hypothetical Amplifier Example

As the input power is increased, beyond some point the output will begin to distort. At very high input power, the output will become saturated; that is, it will not increase further for increased input power.

Between the one extreme of almost perfect linearity at very low power levels and the other extreme of saturation, it’s common to refer to the 1- and 3-dB compression points. From Figure 1, these points are seen to be the output power 1 or 3 dB below the power that would be developed if the amplifier were perfectly linear.

All applications do not place the same emphasis on these performance measures. For example, traveling wave tube amplifiers (TWTAs) typically are operated in saturation so the amplifier’s power rating is the saturation power. In contrast, a solid-state power amplifier (SSPA) often is rated by its 1-dB compression power. Clearly, a 100-W TWTA would not work correctly in an application that required a 100-W SSPA. To achieve 100-W 1-dB power, you would need a 200-W or 300-W TWTA.2

By assuming that all the amplifier distortion is caused by the third order term, several relationships can be derived. The TOI is the point where the third order part and the linear part of the fundamental term have equal amplitudes. This occurs at

A large TOI implies a more linear amplifier, and it’s clear that V2 will be larger if D3 is smaller. For the Figure 1 amplifier, a voltage gain of 10 and a third order distortion of 0.02 were assumed. Substituting these values in Equation 3 gives V2 = 666.6 or approximately 41.25 dBm in a 50-? system. This is the value of TOI referred to the input. By definition, the straight line representing the third order power intersects the linear term at TOI.

It’s easy to see how quickly the third order power grows. Because the voltage is cubed, this term has little effect at small inputs. But, the +3 slope means that its power increases 100 times as fast as the input power. At a 25-dBm input, the third order output power is about 12.5 dBm compared to 45 dBm for the linear output. But, for a 10-dB increase at the input, it’s grown to 42.5 dBm while the linear output has only added 10 dB to 55 dBm.

For an amplifier that can be describ-ed by Equation 1, the 1-dB compression point occurs approximately 9.6 dB below TOI. Not all distortion in a real amplifier is caused by the third order term, and the shape of the transfer curve above the 1-dB compression point varies considerably.

For example, in Reference 2, a technical note comparing specific models of TWTAs and SSPAs, the author commented, “The SSPA transfer curve bends over sharply after the 1-dB compression point whereas the TWTA output power continues to rise, saturating about 3 dB higher.”

Nevertheless, TOI is a good indication of amplifier distortion. A power series treatment like Equation 1 is intended to model small amounts of distortion fairly well, but it doesn’t extend to saturation. In Figure 1, the difference in the power level between TOI and the 1-dB compression point is shown relative to input power because this is how both TOI and the 1-dB point are derived from Equation 2.

### Real Amplifiers

Having seen how the equations work for a hypothetical amplifier, how about trying a real one? As an example, consider the AR Worldwide Model 250W 1000A. This is a broadband SSPA rated to deliver 250 W minimum anywhere from 80 to 1,000 MHz with a typical ±1.5-dB flatness, maximum ±2.0 dB. It has a gain of at least 54 dB and develops 250-W minimum 3-dB compression power at 0-dBm input. Its 1-dB compression point is specified as 200 W minimum although the nominal 1-dB power is 255 W and the nominal 3-dB power is 310 W. TOI typically is 62 dBm, referred to output.

The minimum 3-dB compression values were used as a starting point for Figure 2. A 250-W 3-dB power means that the linear portion of the fundamental term would produce 3 dB greater power if there were no distortion. One point on the +1 slope output line must correspond to 57-dBm output and 0-dBm input. The minimum output power at the 3-dB compression point is 54 dBm or 250 W.

Figure 2. Real Amplifier Performance

Using a spreadsheet to model Equation 2, the D3 distortion coefficient was iterated until the distorted output level matched 54 dBm with 0-dBm input. The curved line represents the entire fundamental term while the straight line corresponds to only the linear part. Heavy lines indicate minimum specifications; thin lines show the maximum. The results and amplifier specifications are in dBm, but all the calculations are done in terms of voltage as per Equation 2.

Although many real amplifiers reference the maximum output power to 0-dBm input, no attempt was made to do this in Figure 1. However, other than the different ranges of input and output powers, the similarity that results from plotting Figures 1 and 2 to the same scale directly shows how the power series model applies.

Constructing Figure 2 was a compromise because the equations don’t tell the entire story. Starting from the 1-dB compression point didn’t produce the specified 3-dB compression power. Equally, starting from the 3-dB compression point, the 1-dB power was in error. And, from Figure 2, it’s obvious that TOI is only 9 dB above the 1-dB compression point, not the theoretical 9.63.

Nevertheless, 9 dB is a large span, and for this amplifier, there’s no way the transfer curve could be said to bend sharply after the 1-dB point. The performance of this real amplifier corresponds well enough to Figure 1 and Equation 2 for the model to make sense. The calculated power levels are within 0.5 dB of the specified performance whether the 1- or 3-dB compression is taken as the starting point.

It’s important to understand that the third order model is just a model and TOI is only one measure of performance. TOI can be quite high, even in those SSPAs that go into saturation very quickly after reaching the 1-dB compression point. Equation 2 deals with low-level distortion, and TOI is extrapolated from that. You can’t measure TOI because the amplifier saturates well before it is reached. Saturation and 3-dB compression aren’t necessarily well represented by the equation, nor are they intended to be.

### Test-Amplifier Attributes

Because test setups often are experimental in nature, the amplifier providing the test signal may accidentally drive into an open or short. Of course, if the DUT does not present a good 50-? load, the VSWR seen by the amplifier can be very high. Many types of output configurations and protection schemes are used to deal with the possibility of up to 100% reflected power.

The data sheet terminology to look for includes the words unconditional stability. Teseq quotes stability separately, specifying it as unconditional. For the Model 250W1000A, AR Worldwide lists mismatch tolerance and states “100% of rated power without foldback. Will operate without damage or oscillation with any magnitude and phase of source and load impedance.”

Nevertheless, there are trade-offs. Assuming that an amplifier does not oscillate regardless of the load impedance, the capability to withstand large amounts of reflected power then becomes a dissipation issue. John Dearing, product manager at Teseq, discussed the effect of bias design on output match tolerance:

“In Class A amplifiers, the transistors are conducting through 360° of the signal cycle. They draw maximum current and dissipate maximum power when the amplifier is in its quiescent state. This makes Class A amplifiers bigger, heavier, and less efficient. However, because they are designed to dissipate all of the heat when in a quiescent state, no matter how much of the power is reflected back from the load, there is no extra energy to dissipate, and the amplifier can continue to produce full power without risk of damage.”

As an example, the Teseq Model CBA 6G-050 Broadband Amplifier produces a typical 90-W output power and consumes something less than 800-VA supply power. All Teseq amplifiers are Class A including this one that weighs 25 kg. If you needed 90-W output but from a smaller, lighter weight amplifier, a Class AB or B amplifier might be a good choice.

Class AB amplifiers provide a small amount of quiescent bias current, which means that performance is similar to Class A for small signals. However, for larger signals, one transistor or the other in a push-pull pair will be turned off. Distortion can be small overall, but the power-handling capability will be much less than for Class A. Class B eliminates the quiescent bias current of Class A and AB but, as a result, introduces crossover distortion.

Ophir RF makes many types of RF and microwave power amplifiers including complete amplifier systems as well as separate modules and subsystems. Many of the amplifiers use a Class A/AB design, claimed to provide better linearity than Class AB or B while improving efficiency compared to a Class A amplifier. Ratings extend to 2,000-W saturated power at 30 MHz and 150-W linear power at 6 GHz.

Because both Class AB and Class B amplifiers are much more efficient than Class A, they can be smaller and lighter weight, but their outputs must be protected from excessive power dissipation. Mr. Dearing continued, “For Class AB and B amplifiers, foldback protection often is used. The reflected power is monitored and the amplifier gain controlled to maintain the reflected power below the safe level.

“While the load match is good or the demanded output power low, little differences among the amplifier classes are evident,” he explained. “However, once the VSWR increases beyond the safe limit, particularly when high power is demanded, Class AB and B amplifiers will start to fold back and produce less power.

“A related problem concerns modulation,” Mr. Dearing concluded. “If the foldback circuitry is fast enough to track the modulation, the amplifier gain will vary during the modulation cycle and the envelope will not
be correct.”

A common test requirement is to use 1-kHz modulation at an 80% level. As shown in Figure 3, this creates voltage peaks 1.8x the unmodulated carrier amplitude, corresponding to a peak power ratio of 3.2 or 5.1 dB. The amplifier needs to have a 5.1-dB power reserve if the test is to be performed correctly without limiting. It must be clear whether the reserve is in relation to saturation, the 3-dB compression point, or the 1-dB compression point.

Figure 3. 80% Modulated Carrier

Further insight was provided by Jason Smith, applications engineering supervisor at AR Worldwide, “Low-power SSPAs are designed to dissipate twice the heat generated at rated output power so they can withstand any mismatch including either an open- or short-circuit load without special protection circuitry.

“On the other hand, medium-power SSPAs, while also designed to withstand any mismatch indefinitely, have a foldback scheme to limit gain and control power levels,” he continued. “As an example, a 150-W amplifier of this type uses foldback to limit reflected power to no more than 150 W. For high-power SSPAs, foldback limits reflected power to 50% of the rated output. This means that no matter what the load, the amplifier still will produce at least 50% of the rated power.

“TWTAs also use foldback,” Mr. Smith explained, “but for amplifiers used to produce a continuous output, the reflected power is limited to 20% of rated level. Pulsed TWTAs implement foldback protection at 50% of rated power by reducing pulse width rather than gain. For both types of TWTAs, the protection scheme chosen maintains the most useful performance for the types of applications usually encountered.”

### Amplifier Technologies

According to AR Worldwide’s Mr. Smith, “Compared to TWTAs, SSPAs are more linear, exhibit fewer harmonics and less noise, tolerate greater load mismatches without need for considerable gain reduction, and have a much greater MTBF. Replacement costs for SSPA modules also are a fraction of the cost of replacement TWTs.”

Scott Behan, vice president of marketing at CAP Wireless, added, “SSPAs also exhibit fewer, if any, aging characteristics; have longer lifetime and reliability; require little, if any, warm-up time; and often provide flatter gain and power vs. frequency performance than a comparable tube amplifier. These characteristics make the SSPA a better candidate for EMC applications because it provides greater confidence that the user’s test setup performance has not drifted. EMI/EMC testing has a notorious reputation for wide variation in measurement repeatability so any reduction in variation is important.”

On that basis, why do people use anything other than SSPAs? One good answer is power. Although SSPAs continue to increase in power, only a few models approach the 10-kW level. Power greater than 10 kW, approaching 100 kW, is achievable with TWTAs.

Typically, high-power SSPAs consist of several modules and some means of combining their output power. Mr. Smith explained: “Quadrature hybrid couplers may form the first level of power combining, often followed by several stages of cascaded binary, so-called corporate, combining. Depending on the frequency of operation, these are constructed using wound transformers, microstrip, stripline, or waveguide. Because of the large number of amplifier modules combined to reach the highest power levels, radial combiners may be used in the final stages of the amplifier, allowing ultimate flexibility in the number of modules combined in a single stage with very low power loss.”

CAP Wireless’ Mr. Behan segmented the technologies his company uses by frequency range. Up to about 1 GHz, most broadband SSPAs use laterally diffused metal-oxide semiconductor (LDMOS) or MOS transistors in Class AB with feedback for linearization and achieve several hundred watts power output. Corporate combining is used for GaAs and GaN amplifiers up to 6 GHz with total power levels of 50 W.

Ideally, power combiners should not only present low loss in the forward direction, but also should provide isolation between modules. Isolation allows an amplifier to degrade gracefully if one or more modules fail. A radial power combiner is a separate structure that has one output port and many input ports fed from external amplifiers. Power from all the inputs adds simultaneously rather than in cascaded stages. The number of ports, overall bandwidth, and power level that can be achieved consistent with suitable insertion loss and isolation are among the design trade-offs.

A few companies are pursuing a similar radial combining technique but with the amplifier modules mounted inside the combining structure. Above 6 GHz, CAP Wireless uses its SpatiumSpatial Combining Structure.

As Mr. Behan described it, “A tapered center conductor transitions from the input SMA coax connector to a larger center conductor. Multiple antipodal finline antenna elements arranged radially gather the input microwave energy and transition the signals to several microstrip transmission lines. Each line feeds a monolithic microwave integrated circuit (MMIC) power amplifier. The amplifier outputs are launched back onto microstrip lines, which then couple to output antipodal finlines back into a coaxial waveguide where the fields coherently combine.”

The 6-GHz qualification apparently isn’t immovable as evidenced by the 2- to 20-GHz Giga-tronics Model GT-100A Microwave Power Amplifier based on the CAP Wireless Spatium Combining Architecture. The Model RM022020 20-W Amplifier from CAP Wireless, with the same 2- to 20-GHz bandwidth, also uses the Spatium technology. Figure 4 shows the typical 1-dB compression and saturated power levels for the RM022020 Amplifier from 2 to 18 GHz.

Figure 4. Spatium-Based Wideband Amplifier PerformanceSource: CAP Wireless

Giga-tronics’ Marketing Manager Leonard Dickstein commented on the suitability of this type of amplifier for EMI/EMC testing. A broadband amplifier eliminates the need to switch among several smaller bandwidth amplifiers as well as the correction factors required to normalize measurement results. In addition, removing the switches that otherwise would be required improves reliability, power output, and linearity.

Mr. Dickstein explained, “The more usual corporate combining used in power amplifiers involves several stages, each with associated losses. Loss after an amplification stage adversely affects the intercept points which, in turn, increases the intermodulation distortion. A spatially combined amplifier, because it minimizes this problem as well as AM-PM conversion, is well suited for amplifying signals with complex modulation or deep amplitude modulation.”

Wavestream is another company developing spatially combined power amplifiers. The PowerStream Grid Amplifier achieves higher output power than possible from a single MMIC in the 27.5- to 31-GHz Ka-band. Multiple amplifier chips are positioned in a waveguide. Each chip contains many separate amplifiers that all directly radiate power combined in the waveguide.

A similar PowerStream Deck Amplifier is closer to the Spatium idea since separate antennas are formed on cards that contain MMIC amplifiers operating in the 11- to 18-GHz Ku-band. Because the spatial combining is very efficient, overall power consumption and cooling requirements are reduced. This design provides higher power than the Grid amplifier but at lower frequencies. Both amplifier types are intended for use in satellite applications.

### Summary

Obviously, many factors are involved in the selection of an RF or microwave power amplifier for EMI/EMC testing. It must produce the required output power over the specified frequency range. But, does the power rating mean the 1- or 3-dB compression points or saturation? And, does the output power result from a 0-dBm input as often is the case, or is some other input level stated?

Because of the wide range of impedance values a DUT may present, the amount of power reflected could be large. Unconditional stability guarantees that the output won’t oscillate regardless of the load. However, the output devices must be able to dissipate the reflected power, and this is determined by the amplifier’s classification.

If you need the full power output to be produced into any load, then a Class A amplifier is the best choice. Perhaps foldback to some lower maximum output power is acceptable because that condition will only occur infrequently in your application. Under those circumstances, a Class AB or B amplifier would work well and have higher efficiency and lower cost than the Class A equivalent. And, output requirements much above a few kilowatts will favor TWTAs over SSPAs.

As is the case when choosing most types of electronic test equipment, the best selection criteria are those associated with your particular application. Arrange to borrow a demo unit after you’ve narrowed the field of candidates and run tests that directly relate to the kind of work you do. Although there are several performance aspects to consider, no doubt you will find a suitable RF or microwave amplifier from the wide range of available products.

### References

1. Notes on Intermodulation Distortion, http://web.engr.oregonstate.edu/~karti/ece621/lec08_01_16.pdf.
2. Van Fleteren, S., Traveling Wave Tube vs. Solid State Amplifiers, DJM Electronics

February 2009