by Scott Behan, Senior Marketing Manager – Qorvo
Solid-state technology has been touted as being the replacement for vacuum tubes since Shockley, Bardeen and Brattain invented the transistor at Bell Labs. While this statement rapidly became a truism for digital and small signal circuits, vacuum electron devices (VEDs) have stubbornly maintained their hold on high-power RF amplification for nearly 70 years. The maturity and synergistic implementation of two technologies, Gallium Nitride (GaN) transistor/Monolithic Microwave Circuit (MMIC) and high-power spatial combining techniques promise to pry loose another finger from VED’s market grip.
Traveling Wave Tube (TWT) amplifiers (TWTAs) are a common form of VED used to provide microwave power amplification with broad band (multi-octave) capability and/or at power levels to multi-kilowatts. TWTA performance, for the most part, is now achievable using solid-state methodology by combining several GaN devices that have 50-100 watts each. It is fairly straightforward to calculate, and by combining 16 parts, one could expect to achieve greater than one kW without much difficulty.
This high-power capability, or even just the potential to enable the design and implementation of a high-power solid-state amplifier, has caused the few remaining TWTA suppliers to rally behind their cause and advertise the flaws (as they view them) attributable to a GaN-based implementation.
Conversely, proponents of solid-state solutions emphasize the issues associated with TWTAs (real or perceived), while emphasizing the positive attributes of the solid-state solution.
While I will admit that my personal bias leans toward implementation of solid-state solutions, it appears clear that for many applications, there is a trade space that must be evaluated to determine which, if either, technology has a clear advantage for the specific purpose. With an eye toward these trades, I address some of the above noted perceptions.
Reliability is perhaps the easiest of all to address, from either side. For either technology, reliability is highly influenced by the environment in which the product is stored and operated. From the TWTA perspective, TWTAs have been used as the high-power system of choice in space applications from the earliest days of space flight. The systems must have a high reliability and availability, and their continued use in the space environment lends credence. GaN-based amplifier solution providers point out the extensive part longevity supported by testing of substantial numbers of die and wafers at various temperatures. TWTs don’t have the volume or cost points to support this type of testing.
Tubes would appear perhaps to offer advantages in reliability at extremely elevated temperatures, as the cathode requires an elevated temperature (800-1200°C) just to stimulate electron emission. As noted in Figure 3, MTTF numbers for GaN devices for junction temperatures much above 300°C start dropping below 10,000 hours. Although the tube itself may enjoy elevated temperature operation, its associated high-voltage power supply, which is usually solid-state, becomes the limiting factor.
TWTA lifetimes are generally defined by depletion of the cathode (wearout). For a TWT in continuous operation, this could be 50,000 to 100,000 (1) hours or more. Solid-state devices have predictable longevity based upon junction temperature, with lifetimes exceeding 10,000,000 hours at junction temperatures of less than 200°C. The composite lifetime of a solid-state amplifier comprised of several (20 for instance) high-power devices can be roughly approximated to be that lifetime at the junction temperature, divided by the number of devices. So the 20-element amplifier with less than 200°C junction temperatures could be expected to have a lifetime somewhere around 500,000 hours, or approximately a five to ten times improvement over the tube. The challenge is to achieve low junction temperatures in the application.
Since the improvement in lifetime with decreasing junction temperatures for a solid-state amplifier is dramatic, thermal management plays a key role in maximizing reliability.
Tubes are not without their own contributors to degradation in reliability. Passive storage and continuous operation in hot standby often contribute to reduced operating life.
As mentioned above, TWTA cathodes must be hot to operate, and solid-state amplifiers enjoy a much longer life if they are kept cool. In reality, both technologies incorporate components whose lifespans are extended by operation at cooler temperatures. It is incumbent on the designer to implement a thermal management method appropriate to the product requirement, with the additional demand on the TWT designer that the cathode must be maintained at an elevated temperature to properly function. From this perspective, the implementation of the tube presents an additional challenge, providing thermal isolation between the cathode and other circuitry. The solid-state designer only has to remove the heat, and in general, the more efficiently, the better.
Current GaN devices are manufactured such that the heat is most effectively removed via conduction from the bottom side of the die. This facilitates mounting them to a surface, which can be cooled in any number of ways, depending on the application. Junction temperature is readily calculated based on the thermal characteristics of each interface material, and from this, device MTTF and system MTBF established.
GaN has significant operational advantages over other semiconductor technologies in that the operating junction temperature and the 1,000,000 MTTF junction temperatures are significantly (typically > 75°C) higher than that of silicon or gallium arsenide (GaAs).
So ultimately, which technology is better thermally? If the amplifier is in an environmental condition where the junction temperatures can be assured of consistent operation below 250°C, a power-combined GaN solution will probably provide longer life than a comparable TWTA. If not, and provided the TWT that may be exposed to higher temperatures can be significantly thermally separated from its control and power electronics, the TWT may be the preferred solution.
GaN RF device development has been steadily progressing for about 20 years, since the late 1990s. Since that time, processes and products have continued to improve capability and mature in quality, resulting in high yields of high performance parts. Several GaN fabs worldwide offer captive, merchant, and open foundry capability. The USAF has recently certified Qorvo GaN processes at Manufacturing Readiness Level (MRL) 9, indicating full technology and process maturity.
EOL and DMSMS
End of Life (EOL) and diminishing manufacturing sources and material shortages (DMSMS) are both valid concerns, both driven by capitalistic requirements to provide a return on investment (ROI) to shareholders. Although semiconductor companies have not typically been subject to DMSMS in their supply chain, their need to maintain their fabs at capacity for maximum utilization requires constant review of product plans, with a goal to serve those markets and opportunities that will provide the maximum ROI. Along with this, technologies, fabrication geometries, and processes are constantly being improved to meet the increasing technological requirements at reduced cost. As product designs age, volume typically drops, and at some point in time, the required processes and equipment for a given design must be replaced with improved capability. At that point, a semiconductor product using the obsolete processes may no longer be commercially available or viable as a product, and it is declared EOL. Fortunately, this cycle usually takes 15-20 years and occurs because technology has improved to the state that legacy equipment is no longer cost effective. But, with an appropriate power amplifier design based on a platform technology such as the Spatium® power combiner, new devices can be rapidly implemented to replace the obsolete part, enabling equal or better performance at lower cost and higher reliability.
Tube manufacturers typically have just the opposite problem. Offering product in relatively low volume, particularly compared with commercial semiconductors, they must leverage their supply chain to take advantage of similar industries that might use the same raw material in sufficient volume so as to create demand. In the past few years, a DMSMS issue for Tungsten filament, a key component of many TWTs, has arisen due to the large scale replacement of incandescent light bulbs by compact fluorescent and now LED illumination. Insufficient demand for Tungsten filament, and particularly the variant required by tube manufacturers, caused the supply to vanish, and cost to increase. The U.S. government has established the filament as a critical requirement and has largely funded an organization to continue low-rate production to support critical needs (2).
The high volume manufacturing capacity argument distinctly leans in favor of the semiconductor industry. TWTs require certain processes, such as vacuum bake-out and burn-in of the cathode that require extensive elapsed time and cannot be effectively accelerated. This effectively limits the throughput. With the high-power microwave TWT market estimated at approximately $600M USD, and even an absurdly low estimate of ASP of $20,000 US per tube calculates out at 30,000 TWTs annually.
This quantity is dwarfed by the billions of RF semiconductor amplifiers produced each year by each of several RF and microwave semiconductor manufacturers. Although most of these products are not high-power, and may be intended for commercial applications, manufacturers successfully leverage this capacity to produce large quantities of high-power GaN MMIC parts with extremely high consistency. This capacity, coupled with industry standard circuit assembly techniques, enables a high-volume scalable capacity that TWTA suppliers can’t match.
It is inconclusive whether TWTAs typically need to be powered on periodically to prevent “gas up” (7)(8), and thus are not particularly suitable for long term storage for use as spares. CPI believes the risk of improper handling and operation outweighs the benefits of periodic testing (8), but acknowledges that microleaks are a known potential cause of failure during storage.
Typical solid-state amplifiers may be stored indefinitely in a benign, dry environment with extremely high likelihood of proper operation when they are removed from storage and operated.
Due largely to the requirement for the TWT cathode to reach an appropriate temperature to produce electrons, most TWTs require at least some time, typically a few minutes, from the time prime power is applied until the units are operational. After initial power application, this time frame can be reduced by “r” leaving the TWTA in “standby,” although this maintains the cathode at temperature and continues to use up cathode life. Additionally, the cathode temperature can be reduced to a temperature where electron emission is no longer excited, but hot enough so that it can quickly be brought up to operational temperatures. Both of these conditions require maintenance power.
GaN-based amplifiers can be powered up with millisecond delays, any delay primarily due to charging of capacitors in power conditioning circuits. Solid-state amplifiers typically exhibit some degree of performance variation as the circuits and elements achieve thermal equilibrium.
TWTA Single-point Failure versus Solid-state Graceful Degradation
TWTAs, because of their individual high-power capability, often create an opportunity for a single-point failure location in the relevant system. Because of this, TWTAs are often configured in an N+1 redundant configuration where a spare TWTA is maintained in “hot” standby (cathode below critical temperature, standby mode to maintain life of the cathode of the spare). This has a cost disadvantage because the additional TWTA must be accounted for in the configuration, but it also can potentially enable field repair of a properly designed system by swapping out the non-functional TWTA while the system remains operational.
Since most GaN implementations of a high-power amplifier will require the combination of multiple devices, a single device failure potentially only results in reduction in power. This may be an acceptable or preferable alternative to maintaining a complete spare. If the complete redundancy is preferred, (to facilitate service, for example), it is generally not necessary, nor is it preferable to keep the spare in a hot standby condition. The spare, while being instantly available, does not suffer any degradation of life since is not operational.
High-voltage Power Supply
Anecdotally, the high-voltage power supply (HVPS) is often identified as being a significant contributor to the failure rate of TWTAs. The power supply must be designed to operate at multiple kilovolts. Insulation materials, dielectrics and conductor spacings must be appropriate, and not susceptible to aging, cracking, or other deleterious effects. One advantage the high-voltage operation does provide is that the corresponding current that is supplied is typically much less than one amp, thus small diameter conductors can be effectively used and I2R losses are minimal.
The low (typically less than 50VDC) power supplies used by GaN and other solid-state power amplifiers are commercially common and often readily available off the shelf. Since the voltages are low, they do not represent the same potential safety hazard as the high-voltage supplies.
VEDs are often touted as operating at higher efficiencies than solid-state, even with GaN, can achieve. A brief survey of TWTA performance from various suppliers (3)(4) appears to show DC/RF efficiencies of 35-40% for C through Ku band CW TWTAs with approximately 10% bandwidth. Octave to octave+ band CW TWTAs have efficiencies of 20 to 30%.
Narrow band (10%) GaN devices intended for CW operation in Ku band of 50 to 60 Watts with efficiencies of 30% are available (5). GaN C Band MMICs with ~18% bandwidth have greater than 40% efficiency (6). Octave and Octave+ GaN MMICs have published efficiencies exceeding 20%.
It is true that some amount of combining loss, even if small, inevitably drops the overall efficiency. A typical efficient spatial combining system, such as the Spatium® combiner, has about 0.5 dB of combining loss, or 95% combining efficiency. Another point in favor of the tube efficiency is their typically high gain (50-60 dB), versus a multiage MMIC of 30 dB. The gain difference requires an additional drive amplifier to be coupled to the solid-state solution, further degrading the system efficiency by the total amount of driver DC power consumption. However, even with these contributors toward solid-state inefficiency, it is difficult to argue a blanket statement that one technology has a distinct advantage in efficient operation over the other.
This debate regarding capabilities, benefits, performance, and cost of solid-state versus vacuum devices continues. I suspect as long as there continue to be unique uses and requirements, there will continue to be differing approaches to implement a solution. Ultimately, the market will choose, based on economics. Even if solid-state solutions could technically satisfy all of the needs, I believe it unlikely that solid-state manufacturers will find it prudent to address certain markets and applications, given the exploding need for RF parts and the finite capacity to produce them. But for those applications where the demand is ultimately going to be driven by cost and volume, solid-state will win out over VEDs. For everything in between, it will be up to the designer to determine the optimum solution.
About the author:
Scott Behan is senior marketing manager with Qorvo’s High Power RF Systems group. He has more than 30 years of experience in high-power amplifier design and applications, holds several patents and pending patents in high-power microwave amplifier circuits, and possesses a broad knowledge of related systems and subsystem implementation. He earned a BSEE from Worcester Polytechnic Institute and has published several articles on microwave amplifiers and associated applications.