A. General
1. What is the difference between Ldmos and Vdmos? How to choose between the two?
2. Is there a glossary for terms such as ft, gmax, gain, efficiency etc?
3. A very simplified explanation of transistor parameters for the non technical.
4. How do I compare published dc data of one manufacturer to another?
5. Are Polyfet transistors depletion or enhancement mode? Nchannel? Gate material?
6. How tight should the transistor mounting screws be?

B. Design
1. How do I bias a Mosfet transistor?
2. Are Mosfets appropriate for broadband high power rf amplifier designs?
3. Should a transistor always be biased in Class A for best linearity?
4. What is the difference between linear and non-linear simulations?
5. High Power broad band simulation example - TB-160
6. How to stabilize high gain rf power mosfet transistors?
7. Are there example designs from Polyfet that might match my needs?
8. Can we expect good power match by conjugate matching small signal S11 and S22 data from S parameters?
9. Can we expect good power match by conjugate matching to published Zin and Zout data?
10. What are the design considerations to achieve low noise figures?
11. Should I be using a 4:1 or a 9:1 transformer?
12. When should I be adding ferrites to transformers? Is powdered iron the right choice?
13. Designing for wide temperature variations.
14. What are some wideband operation considerations?
15. What is the best way to mount a surface mount transistor?
16. Design by load pulling data versus using large signal impedance data
17. How does high power combining work?
18. Precautions in using high gain ldmos transistors.
19. When and what are the advantages in employing push-pull versus single-ended designs.
20. What type of Power meter/sensors should be used?
21. How do device manufacturers test the VSWR of a LDMOS or VDMOS transistor?
22. What are the different ways to combine amplifiers to achieve higher power ?


A. General
1 DMOS transistor stands for double diffuse metal oxide silicon transistor. Back in the 60s, photolithography techniques limited gate channel lengths to no smaller than 10u. In order to have shorter channel lengths to increase ft, device engineers invented the dmos transistor whereby the gate channel length is not determined by the width of the printed gate, but rather by the difference in diffussion depths between the Pbody diffusion and the Source diffusion. Channel lengths of 2.0u or less became feasible.
The 'V' in VDMOS stands for vertical. The current flow from source (top of the die) to drain(bottom of the die) is vertical - thus the term vertical.
The 'L' in LDMOS stands for lateral. The current flow is horizontal fom source to drain - thus the term lateral.
Click here for a pictorial description between the two technologies.

Return to top

2 A very complete glossary of RF terms can be found in the article Rftopics. It also covers other topics such as where to source rf components, design considerations, thermal considerations, general circuit techniques, cautions on using rf transistors, etc.

3 MOS TRANSISTOR ANALOGY - Imaging the MOS transistor as a tap in your sink. Let the water flow be electricity. The following analogy applies
Drain - Drain in the sink where the water flows into.
Source - From where the water comes to the tap Gate -
Gm - The controlling valve of the tap - If Gm is high, it takes only a small amount of turn on the valve to affect a large change in the amount of water flow.
Idsat - The maximum amount of water flow.
Vt - The minimum amount of tap opening to begin water flow.
Idss - Drip that you don't want to have.
Igss - Water leak coming out of the valve.
Bvdss - The maximum water pressure you can put behind the tap before it burst.
Vdson - Number of tap turn openings to achieve high water flow. The fewer tuns the better.

4 This article describes how to relate gm to coss and how best to compare transistors between different manufacturers.

5 Polyfet transistors are N Channel Enhancement mode, silicon gate DMOS transistors. Gate channel lengths are about 1.0u. Polyfets are designed to have minumum drain to gate (Miller) feedback capacitances to permit high frequency of operation. The parasistic NPN bipolar is minimized, creating a rugged transistor capable of withstanding high VSWR.

Adherence to the following recommendations will assure optimum performance from Polyfet MOSFET transistors.
1) Tapped holes in heatsinks should be free of burrs and have a minimum depth of 0.25 inches.
2) Suitable length screws should be used with a 4-40 UNC/2A thread. A washer should be used to spread the joint pressure.
3) For transistors up to 80 Watts Pout, the heatsink should be a minimum of 0.120 copper or 0.200 Al. For transistors of greater levels of Pout, thickness should be increased proportionally.
4) Flatness and finish of mounting surfaces is critical. Flatness should be 0.0008 inch or better and finish should be minimum of 16.
5) A sparing use of evenly distributed heatsink compound on the transistor flange is recommended. Suitable brands of heatsink compound are: Dow Corning 340, Eccotherm TC-5 (E&C) and Wakefield 120.
6) The screws through the flange holes should first be finger tight then tightened to 0,6 to 0,75 Nm to achieve the published thermal resistance between the device and the heatsink.
7) When a transistor is removed from a heatsink, the joint pressure will almost certainly have distorted the flange. Grinding or lapping of the flange according to the information above is necessary to restore proper conditions for mounting.

Return to top

B. Design
1. Biasing -- The simplest way to bias a RF Mosfet transistor is to apply a bias voltage consistent with the desired quiescent current through a gate resistance of 1 - 10 K ohms. See article for detail description. If the amplifier is to be operated over large temperature extremes, then some kind of temperature sensing circuitry must be incorporated to maintain a nearly constant quiescent current over temperature. The easiest way is to use small signal diodes connected in the bias path so as to increase the bias voltage at low temperatures and reduce the bias voltage at high temperatures. The junction voltage potential of a diode follows this function and can be successfully used to compensate the Mosfet. With the advent of inexpensive digital technology, an alternate method of temperature compensation may be accomplished by monitoring the temperature and programming a ROM look up table to generate a gate voltage that will maintain a nearly constant quiescent current. Being digital in nature, the temperature tracking is considerably more precise than a simple diode approach.

2. Mosfet vs Bipolar -- Mosfets technology is vastly superior to bipolar technology for the design of broad band amplifiers at the low to medium frequencies because at these frequencies the input impedance of the device is very high and a simple termination of the gate will determine the gain of the device over many decades of frequency. At higher frequencies, the input impedance of the Mosfet becomes quite similar to the bipolar counterpart. However, Mosfets and especially LDMosfets have much superior gain at higher frequencies making them still the device of choice. In fact Mosfet technology has become the dominant technology for the generation of RF power over the past decade.

3. Class of Bias --- Theoretically class A operation should result in the best linearity of a device. However one must look at the mechanism of the non-linearity to determine if class A operation will always result in superior performance. The answer lies in the transconductance curve of the Mosfet device. Looking at a typical Mosfet device the transconductance rises from VTO at an exponential rate until the narrow channel effect starts to choke off the current increase and the transcondance starts to decrease. The curve is typical of all narrow channel devices whether Vmos or LDMos devices. Since the transcondance is constantly changing, class A operation does not always result in the best intermodulation distortion or linearity unless the current swing is extremely small which will result in very low efficiency. In many cases, class AB operation can result in superior linearity with increased efficiency.

Return to top

4. Linear vs Non Linear simulation -- Linear simulation is derived when a circuit with active devices is operated at such a low power that the simulated measurements are no longer power dependent. This simulation can be achieved by two methods. First the circuit uses a non linear model that is normalized to a quiescent current condition and power levels used in the simulation is so low as not to change the data. Another linear simulation is to use tabular data to describe an active device and generate simulation data based on the data file for the active device. Usually the data file is in S parameter format. Other formats have been used in the past at lower frequencies. If the non-linear model and the linear data file agree, both simulations will yield the same measurement data. In the case of using a non-linear model with a non linear simulator the simulation results are generally very close to actual amplifier performance. The non linear simulator will provide gain compression, power output, efficiency and harmonic power levels. With somewhat less accuracy, intermodulation distortion can be measured but not with the same accuracy as the single tone measurements due to the fact that to obtain accurate results, the device model would have to track an actual device closer than 5%. Five percent accuracy is generally acceptable for gain compression and efficiency measurements. Non linear simulators generally are more costly, but are really the only choice if large signal performance simulation is desired. Polyfet uses standard Berkeley Spice format for its non linear models. S parameters and Spice models can be downloaded from the shortform catalog section of the website or from the model files section of the catalog. Example simulation files can be downloaded for use as templates to create other new designs.

5. RF amp Simulation -- TB-160 for ADS and TB-160 for Microwave office shows a good correlation between simulation results and actual measurements. This is a 30-510Mhz high power broadband amplfier using the ldmos LR401 transistor.

6. Stabilizing transistors -- Due to the very high gain exhibited by all Mosfet devices, and keeping in mind that the device should be unconditionally stable at all frequencies, the device must be resistively terminated at all frequencies even outside the operational band of interest. Two common approaches to Mosfet stabilization are series and parallel gate resistance and negative feedback. Either one can be used or a combination of both for circuit stabilization. Each technique has advantages and drawbacks. Simply resistive can lower gain below the desired limit and drain to gate feedback can affect the rise and fall times of the pulse response of the circuit. In addition, if drain to gate feedback in employed, care must be given to ensure during power up, a surge in gate voltage may occur due to the charging of the required blocking capacitor. These methods are described in a talk given at MTT 1997. Linear simulation with S parameters can assist the stability of a given circuit configuration although the worst case stability condition may occur at power levels above the quiescent current.

7. TB Notes -- Polyfet provides many example amplfier designs to assist customers in building their own. Documentations include circuit scheamatics, layout, pictures of amplfiers and measured data. Polyfet is able to provide autocad files to make the required PCBs.

Return to top

8. Conjugate Matching S para? ---Contrary to some early articles on Mosfet circuit design, Polyfet has found that small signal impedance parameters are not suitable for the design of a large signal device. It is not to say that if a suitable circuit design exists, the insertion of an S parameter file will generate simulation results similar to the small signal response of the amplifier, but the design of the circuit cannot be accomplished with small signal data. In power amplifier design the output match is not the conjugate of the output impedance of the device. The output match is primarily a load impedance that takes into account the supply voltage swing available, the desired power and device saturation current. Likewise, the input impedance is a direct function of the output load impedance so one must start at the load line and work back toward the input match. The published S parameter data is measured with a load of 50 ohms and typical power load lines are usually between 2 and 6 ohms.

9. Conjugate Matching Zio? --Using published large signal data, Zin - Zout, is a good starting point for circuit design. It must be remembered that the data was obtained in a narrow band circuit and one can expect some variations in performance due to the fact that the termination impedance at the harmonic frequencies is totally different from a narrow band circuit and a broad band circuit. Since most device designs utilize class AB operation, considerable harmonic voltages are present on the drain of the device and the termination at the harmonic frequencies can affect the actual performance considerably. Using a large signal simulation, can fine tune the circuit before actual physical construction.

10. Noise figure, much like load line matching, is not determined by small signal S parameters. Very similar to power matching, noise matching is achieved with presenting the input of the device with a load impedance that results in the optimum noise figure. Presently Polyfet does not provide noise figure impedance data since the majority of high power devices are used in applications where the system noise figure is determined in the early driver stages or even in the frequency source. However, there are applications for using a high power device for a wide dynamic range preamplifier. Input noise figure impedance contours may be generated manually or by load pull techniques, but the current technology does not allow the generation of noise figure data on high power devices to be estimated by simulation.

Return to top

11. 4:1 or 9:1 Matching -- Most broadband amplifier applications that involve octave or more bandwidth, use broad-band transformers to achieve an impedance close to the desired level. Reactive matching is used to fine tune the match for the application. Coaxial transmission line transformers have been chosen as the broad-band matching tool of preference because of the inherently wide bandwidth and the ease of duplication. It is even possible in some cases to print the transformers on a printed circuit board eliminating the need to expend the time and material for construction. One limitation of coaxial transmission line transformer is the limitation of choice of transformation ratios. Transformation ratios of 4:1 and 9:1 are most common. Other ratios are possible, but are generally not practical at higher frequencies due to the complexity of the structure and the resulting parasitic inductance of the connections. The decision to use a 4:1 rather than a 9:1 is determined by power level and bandwidth. The 4:1 transformer by physical design exhibits the broadest bandwidth and the 9:1 can achieve somewhat more than twice the impedance ratio for high power applications. A subtle difference between common 4:1 and 9:1 transformers is that the 4:1 is generally of the constant delay type that has inherently wider bandwidth. The common 9:1 design that is actually an auto-transformer with the primary connected back to the secondary which limits the electrical length to less than 1/8 wavelength at the highest frequency of operation. This limits the lower frequency of operation.

12. Ferrite materials have electrical properties suitable for use in RF power amplifiers. Ferrite materials maybe used to extend the low frequency response of the transformers by increasing the shunt inductance. It also allow the use of smaller matching structures to achieve acceptable performance or to improve the stability of an amplifier by providing a choking function. Depending on the specific application, the material and the permeability of the material must be chosen to achieve optimum results. The most common application is the drain choke that is used to supply the drain. Powered iron cores are the best choice because there is considerable DC current and the powered iron has a very high saturation current level. For impedance transformers, both ferrite and powdered iron material maybe used. Generally ferrite material has higher permeability than powdered iron so they are more useful in the design of multi-octave frequency applications. When using ferrite material, caution must be taken in choosing a material that exhibits a curie temperature much higher that the expected operating temperature. This limits the choice to lower permeability material as the high permeability material generally has a lower curie temperature. Powered iron permeability is generally much lower than ferrite, but has successfully been used in transformers in the 100 - 1000 MHz frequency range. It has a very high curie temperature and can extend the low frequency performance of a transformer an octave or more.

13. Wide Temp Designs -- There are two primary issues that must be addressed in the operation of power amplifiers over extended temperature ranges. First is the gate bias circuit must compensated for the change in VTO over the desired temperature range. Second, since the Mosfet device has a positive temperature of the silicon material, power output will decrease with increase flange temperatures and sufficient power margin must be designed into the product to ensure both junction temperature compliance and performance margin.

Return to top

14. Wideband Designs - If there is one golden rule in the design of wide band power amplifier circuits, it is to design the broadest possible load line and select a device to operate into the resulting load. Trying to select a device and then design a multi-octave matching circuit may end up with an unrealizable design. This may sound like an over simplification, but since most if not all multi-octave design involve transformers, the transformer ratios are very limited that have proven useful for RF applications, so there is a limitation in practical broad band matching networks. One can spend an extremely long time trying to accomplish the impractical. This is a very good application for non linear circuit simulation.

15. Mounting Surface Mount Devices -- Ideally if a surface mount device can be soldered directly to a heat sink, heat dissipation becomes a non problem. In the case when the transistor is mounted on a pcb, adequate heat transfer can be made through plated via holes. An application note provided in the industry describes how to mount such devices in detail.

16. Load Pull vs. Zin Zout. In recent years, automated load pull technology has replaced the time consuming laborious effort of using break apart test fixtures to obtain a device's large signal device impedance. Using computed automation, optimum input and load impedance maybe determined for best gain, efficiency, or lowest ACP. The only limitation with such measurement techniques is that due to the fact the tuning networks tend to be a quarter wavelength long, automated load pulling is delegated to frequencies above 500 MHz. Polyfet has pioneered in the development of large signal impedance data using computer simulation of non-linear models. This technique allows the generation of approximate large signal data down to the MHz range. The extraction of this data is somewhat time consuming since it is best accomplished by manual tuning as opposed to just using the optimizing function available in most simulators. Multi-octave designs are best accommodated by using computer generated large signal impedance data and narrow band high frequency designs >500 MHz are best designed using load pull techniques.

Return to top

17. Combining High Power Modules - High power modules maybe combined for higher powers provided the following guidelines are followed. The first rule in successful combining is to choose a combiner that has good isolation between ports. There are two basic type of combiners. These are inphase and quadrature. The inphase combiners have the most bandwidth and can cancel even order harmonics, but since the amplifiers are effectively in phase, a mismatch in the load impedance causes each amplifier to experience the same condition. The quadrature type of combiner has the attribute that splits up the load impedance to different levels to no matter what the load impedance is. Each module sees a different load and usually one module can deliver sufficient power. In four way combining, sometimes both combiners can be effectively used. Two pairs of quadrature combined stages can be combined with inphase combiners to suppress even order harmonics. Another general rule is to tune the modules to gain and phase match to within .5 dB and 20 degrees maximum. If the gain or phase match is above those limits, the power is not effectively added and some of the power is dissipated in isolation resistors that never gets to the load. The third issue is the combiner should exhibit at least 15 dB isolation between all ports. To ensure stable operation the amplifier modules must be stable at all frequencies, since the isolation performance figures pertains only to the bandwidth combiner, not all frequencies that the amplifier may have considerable gain. If all of the above rules are followed, then one should be able to successfully combine modules for higher power.

18. Precautions in using LDMOS transistors - LDMOS transistors have very high gain and can be unstable at low frequencies. Care must be taken during the design stage and initial start up to prevent the transistor from oscillation and self destruction.
To prevent oscillation, we suggest a RC feedback from Drain to Gate. The capacitor should be a chip capacitor of about 0.01 ufd and the resistor of about 100-200 ohms. Other forms of gain reduction technique is a series RC from gate to source. The capacitor value is about 0.01 ufd and the resistor is about 10-20 ohms. Placing a 2-5 ohm in series with the gate will also lower the gain of the transistor. The input balun is usually loaded down with ferrite to reduce oscillation possibilities. The output inductor from drain to Vdd is usually wound around a 10 ohm resistor creating a lossy inductor to prevent oscillations. When on the bench, start with a lower Vdd then designed. Bring up the Idq slowly and increase Rf input slowly. Use a small power supply to begin with so there is not enough power to burn out the transistor.
Other suggestions
Some precautions are necessary when designing with RF power transistors in order to assure optimum reliability. A review of the important ones may save you some costly burnouts. Initial Amplifier Turn-on - The life or death of your transistor in a newly completed amplifier design may depend on how it is first tested. Always begin at a low Drain voltage and minimum drive. Watch the drain current, power output and the spectrum analyzer carefully. Both Id and Pout should come up smoothly. Tune the amplifier while watching the spectrum analyzer for spurious response. Spurious responses below 1 or 2 Mhz are the most lethal. Any spurious response that will occur in an amplifier can be seen at a very low drain voltage. Good starting voltages are 9 volts for 12 volt amplifiers and 15 volts for 24 to 28 volt amplifiers. Operate the transistor within Specifications - Do not sacrifice reliability to achieve a little extra power output or gain by excluding specifications. Voltage breakdowns, maximum drive, and power dissipation are very important; but, three other specifications are most often abused. Load VSWR - Although MOSFETs are not as sensitive to VSWR as bipolar transistors, some caution should be applied. In general, the higher frequency of operation, the higher VSWR tolerance of the device. Most manufacturers specify a maximum load VSWR for safe operation. The maximum permissible VSWR is less at lower frequencies. Example: a transistor which is safe with and VSWR load at 1 Ghz maybe capable of withstanding only a 10:1 VSWR at 100 Mhz. Also the maximum VSWR specifications on a data sheet is usually for a transistor-circuit combination. Some circuits present higher VSWR's to the transistor than others. Frequency Range - A transistor should be used within its intended frequency range if possible. If a transistor is used at a lower frequency it will be more fragile and more susceptible to oscillations. If operated at higher frequencies, lower gain and poorer efficiencies can be expected. Spikes and Power Supplies - The source of DC power for an RF amplifier is very important. Spikes and high voltage kill transistors! For example, an automotive electrical system which supplies a nominal 12 volts usually runs at 13.6 volts and can go to 16 volts or have spikes even higher. Watch your supply voltage carefully, get rid of the spikes and make sure the transistor you use can handle what is left. An often overlooked supply voltage problem occurs in the electronically regulated power supply. Many of the finest laboratory power supplies are sensitive to radiated RF. While the RF amplifier is being tuned or being subjected to a load mismatch, the power supply may suddenly add several volts on its own. Always watch your voltmeter to make sure this does not happen to you.

Return to top

19. Push Pull vs. Single Ended -- Lee Max of CTC, INC., a technological leader at that time, invented the application of push-pull RF devices in 1977. The original idea was by operating an RF amplifier in push-pull effectively operates the devices in series, and quadruples the output load impedance for an equivalent power output. The broad band efficiency and cancellation of the second harmonic distortion in addition to higher power capabilities makes this a viable topology today. Generally if the desired bandwidth of a proposed amplifier approaches an octave, push-pull allows operation with minimal second harmonic distortion. It also simplifies the cascading of stages because driving an input of a device with high levels of second harmonic tends to decrease efficiency. In high power applications, greater than 100watts, it helps to distribute the heat and allows the design of practical 300 watt devices. The issues that should be addressed during the initial design phase are what is the frequency range and desired power output. If the frequency range is greater that a half an octave or the power is greater than 100 watts, a push pull part is the most desirable choice.

Return to top

20. Power Measurement Using Non-Thermal Devices
There are three primary methods for measuring RF power. They are diode voltage detectors similar to the Bird 43 series of wattmeters or RF volt meters similar to the Booton 9200B. Both of these devices read RF voltage and their meter scale is calibrated to read in dBm or Watts. Thermocouple and thermistor devices similar to the HP/Agilent 438/432 series respond to heating power and read actual RMS power less mount conversion losses.
Theory of Diode/Voltage Detector Operation
Let's examine the diode and voltage detectors. Since diode and votlage detectors are voltage sensitive devices, they are calibrated by using the known relationship between RMS and peak voltage. This relationship is only true for distortion free sine waves or amplifiers and or transmitters with harmonic filters installed. Their usefulness is limited to measuring CW signals. Both conventional AM and SSB (two tone) signals in addition to harmonic distortion will cause unpredictable results. The measured voltage is a function of the addition and subtraction of the magnitude and phases of the carrier sidebands.
Diode/Voltage Detector Limitations
This is a fundamental limitation of using diode devices to measure RF power. Both thermocouples and thermistor devices use heating as a method of measuring power. This is true RMS power and although the implementation of the devices is slightly different, both are unaffected by harmonic and the number of signals or sidebands in the RF signal present.
The reason for the measurement error when using diode type detectors is that the complex voltages generated by vector addition of all the signals present distorts the signal from a pure sine wave, and the relationship between RMS (true power) and peak voltage is no longer valid. To illustrate this point if one is trying to measure the power level of a signal with two tones and one tone is 20 dB less that the first tone, the thermal type wattmeters will read 1% higher than with one tone. This is because the power in the second tone is really one-hundredth that of the primary tone. A diode type power meter would read the voltage. Now a tone 20 dB down from the reference tone will have an amplitude of 10% of the reference tone so depending on the relative phase of the two tones the resultant peak to peak voltage could result and an error of +/- 10% in voltage or +/- 20% in power. This is a possible error of almost +/- 1 dB and compared to an error of just .043 dB error for the thermal measurement method.
In general, the diode type of power meter is accurate and useable for any amplifier testing that is CW with the harmonic level below -50 dBc. If this is not the case, then thermal type meters should be used to avoid power measurement errors. Diode or voltage detection type power measurement should not be used where harmonics and multi-tone signals similar but not limited to SSB are present.

Return to top

21. VSWR Testing

There are three common methods in use to day to test for VSWR ruggedness.
First maybe used at virtually all RF frequencies from HF through the GHz range. A simple resistive attenuator is placed at the output of the device's test fixture. The value is determined by the VSWR to be tested. We will discuss the calculation on how to arrive at the correct attenuation value later. See note 1. A variable phase shifter is attached to the output of the attenuator. The attenuator provides the VSWR and the phase shifter provides the ability to test at all phase angles. At low to mid UHF frequencies, the phase shifter is implemented using a dual ganged butterfly capacitor with two inductors connected in such a manner that by adjusting the capacitors the network generates phase angles from a short to an open circuit and all the values in between. At higher frequencies (greater than 500 MHz), a line stretcher is more commonly used as the Q is higher and the physical size is reasonable. At high VSWR levels, caution should be used in selecting a line stretcher that can handle the current or its internal sliding contacts will be damaged in a short period of use.
A word of caution about using attenuators to generate VSWR loads. In most cases the attenuators especially of low attenuation values (less than 6 dB) are designed to only handle their rated power with a 50 ohm termination. By placing a near infinite VSWR load that a phase shifter produces, will cause the peak voltage and current to double in either the shunt or series resistive elements depending on the phase angle of the VSWR. This is going to increase the power dissipation of these elements to four times their normal power dissipation with a 50 ohm load. When using attenuators to produce high VSWR loads, the maximum power applied should be one quarter (1/4) of their terminated (50 ohm) rating. Failure to limit the power will cause the attenuator to fail or drift out of specification due to overheating the resistive elements.
Another method is to use a length of shorted coaxial cable. The length is selected empirically such that the return loss is equal to the VSWR desired. This cable is connected to the output of the amplifier. Depending on the loss of the cable and the frequency of the test, a slight adjustment of the test frequency will shift the reflection 360 degrees. This accomplishes testing at all phase angles by shifting the test frequency a couple of percent. This method is easy to configure because it uses just a piece of coaxial cable and a short termination. The disadvantage of such a method is a different length of cable would be required for each test frequency and generally the cable length for moderate VSWR loads is quite long (as much as 100 feet). This method has some practical use for very high power testing as the power is dissipated along the length of the cable.
In any of the above tests, the DUT is set up in a standard power measurement arrangement. A test reference output power output is measured in a pre-determined test condition. All power is removed and the adjustable VSWR load is applied to the DUT output. DC and RF input power is applied. The phase is slowly rotated 360 degrees for 15 -30 seconds. After the test, the load is replaced and the DC and RF power is applied. The device passes the test if the power does not drop more than .1-.2 dB after the test.
The above procedures only test for minimal degradation for a limited amount of time and no attempt is made to maintain a constant power into the VSWR. A product passes if it survives. For applications requiring developing rated output power into a VSWR, both circuit topology and device size (power and voltage ratings) must be considered.

Note 1.
VSWR to Return Loss Calculation: Return Loss in dB = 20 Log 10(VSWR -1/VSWR+1)
Example: VSWR 3:1 = 20 Log 10(3-1/3+1) = -6.02 dB
Now to select an attenuator remember that the return loss of an attenuator is twice the attenuator's loss due to the input power being attenuated though the attenuator and the reflection is also attenuated on the way back to the source. The correct value of attenuator to use to achieve a 3:1 VSWR is a 3 dB attenuator.

Return to top

22.Amplifier Combining Techniques

High power modules may be combined for higher powers provided the following guidelines are followed. The first rule in successful combining is to choose a combiner that has good isolation between ports. There are three basic types of combiners. They are in-phase, 180 degree and quadrature (also known as 90 degree hybrid).   The in-phase combiners have the most bandwidth but do not cancel even or odd order harmonics. The 180 degree combiner is capable of reasonable bandwidth and possesses the ability to cancel even order harmonics. In both the in-phase and 180 degree combiners, the amplifiers are effectively either operated in phase or out of phase. A mismatch in the load impedance causes each amplifier to experience the same phase angle of the VSWR.   Therefore the ability of the amplifiers to deliver substantial power into a VSWR is limited to what a signal stage can deliver.

The quadrature type of combiner has the attribute that splits up the load impedance presented to each module. No matter what the combined amplifier’s load impedance is, each module sees different load impedances and normally one module can deliver sufficient power to drive the load at near full power.  This results in a pair of modules that will deliver a more nearly constant power at all phase angles of a mismatched load. A side benefit of a broad band quadrature combiner is the capability of canceling the third harmonic. This is beneficial when driving low pass harmonic filters.

In four way combining, sometimes different types of combiners can be effectively used together. An example is a push-pull amplifier module (180 degree) combining followed by quadrature combining.  This module can be further combined with in-phase combining to even higher powers, sometimes exceeding 1 kilowatt or more. The 180 push-pull design cancels the even order harmonics, the quadrature combining cancels third harmonic distortion and dramatically improves the ability to drive nominally high VSWRs. Further combining can be accomplished with in-phase techniques to very high power levels. In the above examples all common types of combining are utilized.

Another general rule is to tune the modules to gain and phase match to within +/- .5 dB and 20 degrees maximum. If the gain or phase match is above these limits, the power is not effectively added and some of the power is dissipated in isolation resistors and is never propagated to the load.

The third issue is, the combiner should exhibit at least a15 dB isolation between all ports.  To ensure stable operation the amplifier modules must be stable at all frequencies, since the isolation performance figures pertain only to the bandwidth of the combiner, and not all frequencies where the amplifier may have considerable gain.

In narrow band applications it is especially important to insure the combiner has a broader isolation bandwidth than the out of band gain of the amplifier. It is not uncommon for amplifiers to exhibit considerable gain over a relatively high bandwidth due to the low Qs in the matching networks.

All of the combining techniques mentioned above may be fabricated in lumped components, microstrips or broadside coupled striplines. Usually the printed microstrip and stripline designs are implemented at higher frequencies and lumped designs with or without ferrite loading are found in HF through VHF frequencies where the physical size of printed designs is impractical.

If all of the above rules are followed, then one should be able to successfully combine modules for higher power.